From 9b47faeaa1136ead696cec2308e29792e609c7e6 Mon Sep 17 00:00:00 2001 From: Courtney Armour Date: Thu, 28 Apr 2022 11:19:57 -0400 Subject: [PATCH 01/26] initial compare_models script --- R/compare_models.R | 111 +++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 111 insertions(+) create mode 100644 R/compare_models.R diff --git a/R/compare_models.R b/R/compare_models.R new file mode 100644 index 00000000..203ca853 --- /dev/null +++ b/R/compare_models.R @@ -0,0 +1,111 @@ +#' Get Difference + +# calculate the difference in the mean of the metric for the two groups +# sub_data: subset of the performance data for just two groups +# group_name: name of column with group variable +# metric: metric to compare +get_difference <- function(sub_data,group_name,metric){ + # get the mean metric value for each group + means <- sub_data %>% + dplyr::group_by(.data[[group_name]]) %>% + dplyr::summarise(meanVal = mean(.data[[metric]]),.groups="drop") %>% + dplyr::pull(meanVal) + # find the difference in the mean between the two groups + abs(diff(means)) +} + +#' Shuffle Group + +# shuffle the groups across the data +# sub_data: subset of the performance data for just two groups +# group_name: column name to shuffle +shuffle_group <- function(sub_data,group_name){ + # get teh group labels + group_vals <- sub_data %>% + dplyr::pull( {{ group_name }} ) + # shuffle the group labels + group_vals_shuffled <- base::sample(group_vals) + + #assign shuffled groups to group column + data_shuffled <- sub_data %>% + dplyr::mutate( !!group_name := group_vals_shuffled) + + return(data_shuffled) +} + +#' Permute P-Value + +# data: the concatenated performance (from what output?) +# metric: metric to compare, either AUC or cv_metric_AUC +# group_name: column with group variables to compare +# group_1: name of one group to compare +# group_2: name of other group to compare +permute_p_value <- function(data, metric, group_name, group_1, group_2, nperm){ + # check that the metric and group exist in data + assertthat::has_name(data,metric) + assertthat::has_name(data,group_name) + # check that group_1 and group_2 exist in the data + #assertthat::has_name(data %>% distinct({{group_name}}),{{ group_1 }}) + + # subset results to select metric and group columns and + # filter to only the two groups of interest + sub_data <- data %>% + dplyr::select({{ metric }},{{ group_name }}) %>% + dplyr::filter( .data[[group_name]] == {{group_1}} | .data[[group_name]] == {{group_2}}) + + # observed difference: quantify the absolute value of the difference + # in metric between the two groups + metric_obs <- get_difference(sub_data,{{group_name}},{{metric}}) + + # shuffled difference: quantify the absolute value of the difference + # in metric between the two groups after shuffling group labels + metric_null <- replicate(nperm,get_difference(shuffle_group(sub_data,group_name),group_name,metric)) + + # n: number of shuffled calculations + n <- length(metric_null) + # r: replications at least as extreme as observed effect + r <- sum(abs(metric_null) >= metric_obs) + + # compute Monte Carlo p-value with correction (Davison & Hinkley, 1997) + p_value=(r+1)/(n+1) + return(p_value) +} + +# Wrapper to do all comparisons +# data: the concatenated performance (from what output?) +# metric: metric to compare, either AUC or cv_metric_AUC +# group_name: column with group variables to compare +compare_models <- function(merged_data,metric,group_name,nperm=10000){ + + # identify all unique groups in group variable + groups <- merged_data %>% + dplyr::pull( {{group_name}} ) %>% + unique() + + # create a table with all possible comparisons of groups + # without repeating pairings + p_table <- tidyr::expand_grid(x=1:length(groups), + y=1:length(groups)) %>% + dplyr::filter(x < y) %>% + dplyr::mutate(group1 = groups[x], + group2 = groups[y]) %>% + select(-x, -y) %>% + group_by(group1,group2) %>% + summarize(p_value = permute_p_value(merged_data,metric,group_name,group1,group2,nperm), + .groups = "drop") + + return(p_table) + +} + +df <- tibble(samp=c("a","a","b","b"),val=c(.2,0.3,0.8,0.9)) +get_difference(df,"samp","val") +shuffle_group(df,"samp") +permute_p_value(df,"val","samp","a","b",nperm=10) +compare_models(df,"val","samp",nperm=10) + +#### TESTS #### + +testthat::expect_equal(get_difference(tibble(AUC=c(0.5,0.8), + type=c("a","b")), + "type","AUC"),0.3) From 2634dc710a3acdf2ea1c1bcd994aae0cba6da25e Mon Sep 17 00:00:00 2001 From: Courtney Armour Date: Thu, 28 Apr 2022 13:19:32 -0400 Subject: [PATCH 02/26] typo --- R/compare_models.R | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) diff --git a/R/compare_models.R b/R/compare_models.R index 203ca853..c9d9df4e 100644 --- a/R/compare_models.R +++ b/R/compare_models.R @@ -1,4 +1,4 @@ -#' Get Difference +# Get Difference # calculate the difference in the mean of the metric for the two groups # sub_data: subset of the performance data for just two groups @@ -14,8 +14,7 @@ get_difference <- function(sub_data,group_name,metric){ abs(diff(means)) } -#' Shuffle Group - +# Shuffle Group # shuffle the groups across the data # sub_data: subset of the performance data for just two groups # group_name: column name to shuffle @@ -33,8 +32,7 @@ shuffle_group <- function(sub_data,group_name){ return(data_shuffled) } -#' Permute P-Value - +# Permute P-Value # data: the concatenated performance (from what output?) # metric: metric to compare, either AUC or cv_metric_AUC # group_name: column with group variables to compare @@ -98,14 +96,16 @@ compare_models <- function(merged_data,metric,group_name,nperm=10000){ } +#### TESTS #### df <- tibble(samp=c("a","a","b","b"),val=c(.2,0.3,0.8,0.9)) get_difference(df,"samp","val") shuffle_group(df,"samp") permute_p_value(df,"val","samp","a","b",nperm=10) compare_models(df,"val","samp",nperm=10) -#### TESTS #### - +#### testthat::expect_equal(get_difference(tibble(AUC=c(0.5,0.8), type=c("a","b")), "type","AUC"),0.3) + + From d580d4d9d6f27aa8490568883c644dcd368cb1dc Mon Sep 17 00:00:00 2001 From: Courtney Armour Date: Fri, 29 Apr 2022 19:01:28 -0400 Subject: [PATCH 03/26] add compare_model tests --- R/compare_models.R | 150 ++++++++++++++++++++------- tests/testthat/test-compare_models.R | 50 +++++++++ 2 files changed, 162 insertions(+), 38 deletions(-) create mode 100644 tests/testthat/test-compare_models.R diff --git a/R/compare_models.R b/R/compare_models.R index c9d9df4e..77d6218d 100644 --- a/R/compare_models.R +++ b/R/compare_models.R @@ -1,10 +1,27 @@ -# Get Difference -# calculate the difference in the mean of the metric for the two groups -# sub_data: subset of the performance data for just two groups -# group_name: name of column with group variable -# metric: metric to compare +#' Average metric difference +#' calculate the difference in the mean of the metric for the two groups +#' +#' @param sub_data subset of the merged performance data for just two groups +#' @param group_name name of column with group variable +#' @param metric metric to compare +#' +#' @return numeric difference in the average metric between the two groups +#' +#' @export +#' @author Courtney Armour, \email{armourc@@umich.edu} +#' +#' @examples +#' \dontrun{ +#' df <- tibble(condition=c("a","a","b","b","c","c"), +#' AUC=c(.2,0.3,0.8,0.9,0.85,0.95)) +#' sub_df <- df %>% filter(condition %in% c("a","b")) +#' get_difference(sub_df,"condition","AUC") +#' } get_difference <- function(sub_data,group_name,metric){ + if(!is.numeric(sub_data %>% dplyr::pull(metric))){ + stop("The specified metric is not numeric, please check that you specified the right column.") + } # get the mean metric value for each group means <- sub_data %>% dplyr::group_by(.data[[group_name]]) %>% @@ -14,12 +31,28 @@ get_difference <- function(sub_data,group_name,metric){ abs(diff(means)) } -# Shuffle Group -# shuffle the groups across the data -# sub_data: subset of the performance data for just two groups -# group_name: column name to shuffle +#' Shuffle the values in the groups column +#' +#' @param sub_data subset of the performance data for just two groups +#' @param group_name column name to shuffle +#' +#' @return `sub_data` with the `group_name` column values shuffled +#' @export +#' @author Courtney R Armour, \email{armourc@@umich.edu} +#' +#' @examples +#' \dontrun{ +#' set.seed(123) +#' df <- tibble(condition=c("a","a","b","b","c","c"), +#' AUC=c(.2,0.3,0.8,0.9,0.85,0.95)) +#' sub_df <- df %>% filter(condition %in% c("a","b")) +#' shuffle_group(sub_df,"condition") +#' } shuffle_group <- function(sub_data,group_name){ - # get teh group labels + if(!(group_name %in% colnames(sub_data))){ + stop("The group_name does not exist in the data.") + } + # get the group labels group_vals <- sub_data %>% dplyr::pull( {{ group_name }} ) # shuffle the group labels @@ -32,22 +65,44 @@ shuffle_group <- function(sub_data,group_name){ return(data_shuffled) } -# Permute P-Value -# data: the concatenated performance (from what output?) -# metric: metric to compare, either AUC or cv_metric_AUC -# group_name: column with group variables to compare -# group_1: name of one group to compare -# group_2: name of other group to compare -permute_p_value <- function(data, metric, group_name, group_1, group_2, nperm){ + +#' Calculated a permuted p-value comparing two models +#' +#' @inheritParams compare_models +#' @param group_1 name of one group to compare +#' @param group_2 name of other group to compare +#' +#' @return numeric p-value comparing two models +#' @export +#' @author Begüm Topçuoğlu, \email{topcuoglu.begum@@gmail.com} +#' @author Courtney R Armour, \email{armourc@@umich.edu} +#' +#' @examples +#' \dontrun{ +#' df <- tibble(model=c("rf","rf","glmnet","glmnet","svmRadial","svmRadial"), +#' AUC=c(.2,0.3,0.8,0.9,0.85,0.95)) +#' set.seed(123) +#' permute_p_value(df,"AUC","model","rf","glmnet",nperm=100) +#' } +permute_p_value <- function(merged_data, metric, group_name, group_1, group_2, nperm=10000){ # check that the metric and group exist in data - assertthat::has_name(data,metric) - assertthat::has_name(data,group_name) + if(!(metric %in% colnames(merged_data))){ + stop("The metric does not exist in the data.") + } + if(!(group_name %in% colnames(merged_data))){ + stop("The group_name does not exist in the data.") + } # check that group_1 and group_2 exist in the data - #assertthat::has_name(data %>% distinct({{group_name}}),{{ group_1 }}) - + if(!(group_1 %in% (merged_data %>% dplyr::pull(group_name)))){ + stop("group_1 does not exist in the data.") + } + if(!(group_2 %in% (merged_data %>% dplyr::pull(group_name)))){ + stop("group_2 does not exist in the data.") + } + # subset results to select metric and group columns and # filter to only the two groups of interest - sub_data <- data %>% + sub_data <- merged_data %>% dplyr::select({{ metric }},{{ group_name }}) %>% dplyr::filter( .data[[group_name]] == {{group_1}} | .data[[group_name]] == {{group_2}}) @@ -57,7 +112,8 @@ permute_p_value <- function(data, metric, group_name, group_1, group_2, nperm){ # shuffled difference: quantify the absolute value of the difference # in metric between the two groups after shuffling group labels - metric_null <- replicate(nperm,get_difference(shuffle_group(sub_data,group_name),group_name,metric)) + rep_fn <- select_apply("replicate") + metric_null <- rep_fn(nperm,get_difference(shuffle_group(sub_data,group_name),group_name,metric)) # n: number of shuffled calculations n <- length(metric_null) @@ -69,11 +125,35 @@ permute_p_value <- function(data, metric, group_name, group_1, group_2, nperm){ return(p_value) } -# Wrapper to do all comparisons -# data: the concatenated performance (from what output?) -# metric: metric to compare, either AUC or cv_metric_AUC -# group_name: column with group variables to compare + +#' Compute all pairs of comparisons +#' calculate permuted p-value across all pairs of group variable. +#' wrapper for `permute_p_value` +#' +#' @param merged_data the concatenated performance data from `run_ml` +#' @param metric metric to compare, must be numeric +#' @param group_name column with group variables to compare +#' @param nperm number of permutations, default=10000 +#' +#' @return a table of p-values for all pairs of group varible +#' @export +#' @author Courtney R Armour, \email{armourc@@umich.edu} +#' +#' @examples +#' \dontrun{ +#' df <- tibble(model=c("rf","rf","glmnet","glmnet","svmRadial","svmRadial"), +#' AUC=c(.2,0.3,0.8,0.9,0.85,0.95)) +#' set.seed(123) +#' compare_models(df,"AUC","model",nperm=100) +#' } compare_models <- function(merged_data,metric,group_name,nperm=10000){ + # check that the metric and group exist in data + if(!(metric %in% colnames(merged_data))){ + stop("The metric does not exist in the data.") + } + if(!(group_name %in% colnames(merged_data))){ + stop("The group_name does not exist in the data.") + } # identify all unique groups in group variable groups <- merged_data %>% @@ -97,15 +177,9 @@ compare_models <- function(merged_data,metric,group_name,nperm=10000){ } #### TESTS #### -df <- tibble(samp=c("a","a","b","b"),val=c(.2,0.3,0.8,0.9)) -get_difference(df,"samp","val") -shuffle_group(df,"samp") -permute_p_value(df,"val","samp","a","b",nperm=10) -compare_models(df,"val","samp",nperm=10) - -#### -testthat::expect_equal(get_difference(tibble(AUC=c(0.5,0.8), - type=c("a","b")), - "type","AUC"),0.3) - +# df <- tibble(samp=c("a","a","b","b"),val=c(.2,0.3,0.8,0.9)) +# get_difference(df,"samp","val") +# shuffle_group(df,"samp") +# permute_p_value(df,"val","samp","a","b",nperm=10) +# compare_models(df,"val","samp",nperm=10) diff --git a/tests/testthat/test-compare_models.R b/tests/testthat/test-compare_models.R new file mode 100644 index 00000000..933e8dbe --- /dev/null +++ b/tests/testthat/test-compare_models.R @@ -0,0 +1,50 @@ + +# get_difference +test_that("get_difference works",{ + expect_equal(get_difference(data.frame(AUC=c(0.5,0.8),type=c("a","b")), + "type","AUC"), + 0.3) + expect_error(get_difference(data.frame(AUC=c(0.5,0.8),type=c("a","b")), + "type","type"), + "The specified metric is not numeric, please check that you specified the right column.") +}) + +# shuffle_group +test_that("shuffle_group works",{ + set.seed(2022) + df <- tibble(condition=c("a","a","b","b"),AUC=c(0.2,0.3,0.8,0.9)) + expect_equal(shuffle_group(df,"condition"), + tibble(condition=c("b","b","a","a"),AUC=c(0.2,0.3,0.8,0.9))) + expect_error(shuffle_group(df,"group"), + "The group_name does not exist in the data.") +}) + +# permute_p_value +test_that("permute_p_value works", { + set.seed(2022) + df <- tibble(model=c("rf","rf","glmnet","glmnet","svmRadial","svmRadial"),AUC=c(0.2,0.2,0.2,0.2,0.2,0.2)) + expect_equal(permute_p_value(df,"AUC","model","rf","glmnet",nperm=10), + 1) + expect_error(permute_p_value(df,"auc","model","rf","glmnet",nperm=10), + "The metric does not exist in the data.") + expect_error(permute_p_value(df,"AUC","group","rf","glmnet",nperm=10), + "The group_name does not exist in the data.") + expect_error(permute_p_value(df,"AUC","model","RF","glmnet",nperm=10), + "group_1 does not exist in the data.") + expect_error(permute_p_value(df,"AUC","model","rf","logreg",nperm=10), + "group_2 does not exist in the data.") +}) + +# compare_models +test_that("compare_models works",{ + set.seed(2022) + df <- tibble(model=c("rf","rf","glmnet","glmnet","svmRadial","svmRadial"),AUC=c(0.2,0.2,0.2,0.2,0.2,0.2)) + expect_equal(compare_models(df,"AUC","model",10), + tibble(group1=c("glmnet","rf","rf"), + group2=c("svmRadial","glmnet","svmRadial"), + p_value=c(1,1,1))) + expect_error(compare_models(df,"auc","model",100), + "The metric does not exist in the data.") + expect_error(compare_models(df,"AUC","group",100), + "The group_name does not exist in the data.") +}) From 7ad628f94d7949b593a1f407bba22721d20420fc Mon Sep 17 00:00:00 2001 From: Courtney Armour Date: Mon, 2 May 2022 12:05:56 -0400 Subject: [PATCH 04/26] fix failed tests --- R/compare_models.R | 8 +++---- tests/testthat/test-compare_models.R | 31 +++++++++++++++++++++------- 2 files changed, 27 insertions(+), 12 deletions(-) diff --git a/R/compare_models.R b/R/compare_models.R index 77d6218d..44effd80 100644 --- a/R/compare_models.R +++ b/R/compare_models.R @@ -167,12 +167,12 @@ compare_models <- function(merged_data,metric,group_name,nperm=10000){ dplyr::filter(x < y) %>% dplyr::mutate(group1 = groups[x], group2 = groups[y]) %>% - select(-x, -y) %>% - group_by(group1,group2) %>% - summarize(p_value = permute_p_value(merged_data,metric,group_name,group1,group2,nperm), + dplyr::select(-x, -y) %>% + dplyr::group_by(group1,group2) %>% + dplyr::summarize(p_value = permute_p_value(merged_data,metric,group_name,group1,group2,nperm), .groups = "drop") - return(p_table) + return(as.data.frame(p_table)) } diff --git a/tests/testthat/test-compare_models.R b/tests/testthat/test-compare_models.R index 933e8dbe..e5c0a867 100644 --- a/tests/testthat/test-compare_models.R +++ b/tests/testthat/test-compare_models.R @@ -12,9 +12,13 @@ test_that("get_difference works",{ # shuffle_group test_that("shuffle_group works",{ set.seed(2022) - df <- tibble(condition=c("a","a","b","b"),AUC=c(0.2,0.3,0.8,0.9)) + df <- structure(list(condition=c("a","a","b","b"), + AUC=c(0.2,0.3,0.8,0.9)), + .Names = c("condition","AUC"), + row.names = c(NA,-4L), + class = "data.frame") expect_equal(shuffle_group(df,"condition"), - tibble(condition=c("b","b","a","a"),AUC=c(0.2,0.3,0.8,0.9))) + data.frame(condition=c("b","b","a","a"),AUC=c(0.2,0.3,0.8,0.9))) expect_error(shuffle_group(df,"group"), "The group_name does not exist in the data.") }) @@ -22,7 +26,11 @@ test_that("shuffle_group works",{ # permute_p_value test_that("permute_p_value works", { set.seed(2022) - df <- tibble(model=c("rf","rf","glmnet","glmnet","svmRadial","svmRadial"),AUC=c(0.2,0.2,0.2,0.2,0.2,0.2)) + df <- structure(list(model = c("rf","rf","glmnet","glmnet","svmRadial","svmRadial"), + AUC = c(0.2,0.2,0.2,0.2,0.2,0.2)), + .Names = c("model","AUC"), + row.names = c(NA,-6L), + class = "data.frame") expect_equal(permute_p_value(df,"AUC","model","rf","glmnet",nperm=10), 1) expect_error(permute_p_value(df,"auc","model","rf","glmnet",nperm=10), @@ -38,13 +46,20 @@ test_that("permute_p_value works", { # compare_models test_that("compare_models works",{ set.seed(2022) - df <- tibble(model=c("rf","rf","glmnet","glmnet","svmRadial","svmRadial"),AUC=c(0.2,0.2,0.2,0.2,0.2,0.2)) + df <- structure(list(model = c("rf","rf","glmnet","glmnet","svmRadial","svmRadial"), + AUC = c(0.2,0.2,0.2,0.2,0.2,0.2)), + .Names = c("model","AUC"), + row.names = c(NA,-6L), + class = "data.frame") expect_equal(compare_models(df,"AUC","model",10), - tibble(group1=c("glmnet","rf","rf"), - group2=c("svmRadial","glmnet","svmRadial"), - p_value=c(1,1,1))) + structure(list(group1=c("glmnet","rf","rf"), + group2=c("svmRadial","glmnet","svmRadial"), + p_value=c(1,1,1)), + .Names = c("group1","group2","p_value"), + row.names = c(NA,-3L), + class = "data.frame")) expect_error(compare_models(df,"auc","model",100), "The metric does not exist in the data.") expect_error(compare_models(df,"AUC","group",100), "The group_name does not exist in the data.") -}) +}) \ No newline at end of file From 37a822b0eab1de6d17ec211baea72768aeb8ee72 Mon Sep 17 00:00:00 2001 From: "github-actions[bot]" <41898282+github-actions[bot]@users.noreply.github.com> Date: Mon, 2 May 2022 16:18:51 +0000 Subject: [PATCH 05/26] =?UTF-8?q?=F0=9F=8E=A8=20Style=20R=20code?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- R/checks.R | 2 +- R/compare_models.R | 142 +++++++++++++++------------ tests/testthat/test-compare_models.R | 140 ++++++++++++++++---------- 3 files changed, 168 insertions(+), 116 deletions(-) diff --git a/R/checks.R b/R/checks.R index 0ce412af..08a2a67f 100644 --- a/R/checks.R +++ b/R/checks.R @@ -274,7 +274,7 @@ check_outcome_value <- function(dataset, outcome_colname) { stop( paste0( "A binary or multi-class outcome variable is required, but this dataset has ", - num_outcomes, " outcome(s): ", paste(outcomes, collapse = ", ") + num_outcomes, " outcome(s): ", paste(outcomes, collapse = ", ") ) ) } diff --git a/R/compare_models.R b/R/compare_models.R index 44effd80..c949a598 100644 --- a/R/compare_models.R +++ b/R/compare_models.R @@ -7,25 +7,27 @@ #' @param metric metric to compare #' #' @return numeric difference in the average metric between the two groups -#' +#' #' @export #' @author Courtney Armour, \email{armourc@@umich.edu} -#' +#' #' @examples #' \dontrun{ -#' df <- tibble(condition=c("a","a","b","b","c","c"), -#' AUC=c(.2,0.3,0.8,0.9,0.85,0.95)) -#' sub_df <- df %>% filter(condition %in% c("a","b")) -#' get_difference(sub_df,"condition","AUC") +#' df <- tibble( +#' condition = c("a", "a", "b", "b", "c", "c"), +#' AUC = c(.2, 0.3, 0.8, 0.9, 0.85, 0.95) +#' ) +#' sub_df <- df %>% filter(condition %in% c("a", "b")) +#' get_difference(sub_df, "condition", "AUC") #' } -get_difference <- function(sub_data,group_name,metric){ - if(!is.numeric(sub_data %>% dplyr::pull(metric))){ +get_difference <- function(sub_data, group_name, metric) { + if (!is.numeric(sub_data %>% dplyr::pull(metric))) { stop("The specified metric is not numeric, please check that you specified the right column.") } # get the mean metric value for each group means <- sub_data %>% dplyr::group_by(.data[[group_name]]) %>% - dplyr::summarise(meanVal = mean(.data[[metric]]),.groups="drop") %>% + dplyr::summarise(meanVal = mean(.data[[metric]]), .groups = "drop") %>% dplyr::pull(meanVal) # find the difference in the mean between the two groups abs(diff(means)) @@ -43,25 +45,27 @@ get_difference <- function(sub_data,group_name,metric){ #' @examples #' \dontrun{ #' set.seed(123) -#' df <- tibble(condition=c("a","a","b","b","c","c"), -#' AUC=c(.2,0.3,0.8,0.9,0.85,0.95)) -#' sub_df <- df %>% filter(condition %in% c("a","b")) -#' shuffle_group(sub_df,"condition") +#' df <- tibble( +#' condition = c("a", "a", "b", "b", "c", "c"), +#' AUC = c(.2, 0.3, 0.8, 0.9, 0.85, 0.95) +#' ) +#' sub_df <- df %>% filter(condition %in% c("a", "b")) +#' shuffle_group(sub_df, "condition") #' } -shuffle_group <- function(sub_data,group_name){ - if(!(group_name %in% colnames(sub_data))){ +shuffle_group <- function(sub_data, group_name) { + if (!(group_name %in% colnames(sub_data))) { stop("The group_name does not exist in the data.") } # get the group labels - group_vals <- sub_data %>% - dplyr::pull( {{ group_name }} ) + group_vals <- sub_data %>% + dplyr::pull({{ group_name }}) # shuffle the group labels group_vals_shuffled <- base::sample(group_vals) - - #assign shuffled groups to group column - data_shuffled <- sub_data %>% - dplyr::mutate( !!group_name := group_vals_shuffled) - + + # assign shuffled groups to group column + data_shuffled <- sub_data %>% + dplyr::mutate(!!group_name := group_vals_shuffled) + return(data_shuffled) } @@ -79,55 +83,57 @@ shuffle_group <- function(sub_data,group_name){ #' #' @examples #' \dontrun{ -#' df <- tibble(model=c("rf","rf","glmnet","glmnet","svmRadial","svmRadial"), -#' AUC=c(.2,0.3,0.8,0.9,0.85,0.95)) +#' df <- tibble( +#' model = c("rf", "rf", "glmnet", "glmnet", "svmRadial", "svmRadial"), +#' AUC = c(.2, 0.3, 0.8, 0.9, 0.85, 0.95) +#' ) #' set.seed(123) -#' permute_p_value(df,"AUC","model","rf","glmnet",nperm=100) +#' permute_p_value(df, "AUC", "model", "rf", "glmnet", nperm = 100) #' } -permute_p_value <- function(merged_data, metric, group_name, group_1, group_2, nperm=10000){ +permute_p_value <- function(merged_data, metric, group_name, group_1, group_2, nperm = 10000) { # check that the metric and group exist in data - if(!(metric %in% colnames(merged_data))){ + if (!(metric %in% colnames(merged_data))) { stop("The metric does not exist in the data.") } - if(!(group_name %in% colnames(merged_data))){ + if (!(group_name %in% colnames(merged_data))) { stop("The group_name does not exist in the data.") } # check that group_1 and group_2 exist in the data - if(!(group_1 %in% (merged_data %>% dplyr::pull(group_name)))){ + if (!(group_1 %in% (merged_data %>% dplyr::pull(group_name)))) { stop("group_1 does not exist in the data.") } - if(!(group_2 %in% (merged_data %>% dplyr::pull(group_name)))){ + if (!(group_2 %in% (merged_data %>% dplyr::pull(group_name)))) { stop("group_2 does not exist in the data.") } # subset results to select metric and group columns and # filter to only the two groups of interest sub_data <- merged_data %>% - dplyr::select({{ metric }},{{ group_name }}) %>% - dplyr::filter( .data[[group_name]] == {{group_1}} | .data[[group_name]] == {{group_2}}) - - # observed difference: quantify the absolute value of the difference - # in metric between the two groups - metric_obs <- get_difference(sub_data,{{group_name}},{{metric}}) - - # shuffled difference: quantify the absolute value of the difference + dplyr::select({{ metric }}, {{ group_name }}) %>% + dplyr::filter(.data[[group_name]] == {{ group_1 }} | .data[[group_name]] == {{ group_2 }}) + + # observed difference: quantify the absolute value of the difference + # in metric between the two groups + metric_obs <- get_difference(sub_data, {{ group_name }}, {{ metric }}) + + # shuffled difference: quantify the absolute value of the difference # in metric between the two groups after shuffling group labels rep_fn <- select_apply("replicate") - metric_null <- rep_fn(nperm,get_difference(shuffle_group(sub_data,group_name),group_name,metric)) - + metric_null <- rep_fn(nperm, get_difference(shuffle_group(sub_data, group_name), group_name, metric)) + # n: number of shuffled calculations n <- length(metric_null) # r: replications at least as extreme as observed effect r <- sum(abs(metric_null) >= metric_obs) - + # compute Monte Carlo p-value with correction (Davison & Hinkley, 1997) - p_value=(r+1)/(n+1) + p_value <- (r + 1) / (n + 1) return(p_value) } #' Compute all pairs of comparisons -#' calculate permuted p-value across all pairs of group variable. +#' calculate permuted p-value across all pairs of group variable. #' wrapper for `permute_p_value` #' #' @param merged_data the concatenated performance data from `run_ml` @@ -141,39 +147,46 @@ permute_p_value <- function(merged_data, metric, group_name, group_1, group_2, n #' #' @examples #' \dontrun{ -#' df <- tibble(model=c("rf","rf","glmnet","glmnet","svmRadial","svmRadial"), -#' AUC=c(.2,0.3,0.8,0.9,0.85,0.95)) +#' df <- tibble( +#' model = c("rf", "rf", "glmnet", "glmnet", "svmRadial", "svmRadial"), +#' AUC = c(.2, 0.3, 0.8, 0.9, 0.85, 0.95) +#' ) #' set.seed(123) -#' compare_models(df,"AUC","model",nperm=100) +#' compare_models(df, "AUC", "model", nperm = 100) #' } -compare_models <- function(merged_data,metric,group_name,nperm=10000){ +compare_models <- function(merged_data, metric, group_name, nperm = 10000) { # check that the metric and group exist in data - if(!(metric %in% colnames(merged_data))){ + if (!(metric %in% colnames(merged_data))) { stop("The metric does not exist in the data.") } - if(!(group_name %in% colnames(merged_data))){ + if (!(group_name %in% colnames(merged_data))) { stop("The group_name does not exist in the data.") } - + # identify all unique groups in group variable - groups <- merged_data %>% - dplyr::pull( {{group_name}} ) %>% + groups <- merged_data %>% + dplyr::pull({{ group_name }}) %>% unique() - + # create a table with all possible comparisons of groups # without repeating pairings - p_table <- tidyr::expand_grid(x=1:length(groups), - y=1:length(groups)) %>% - dplyr::filter(x < y) %>% - dplyr::mutate(group1 = groups[x], - group2 = groups[y]) %>% - dplyr::select(-x, -y) %>% - dplyr::group_by(group1,group2) %>% - dplyr::summarize(p_value = permute_p_value(merged_data,metric,group_name,group1,group2,nperm), - .groups = "drop") - - return(as.data.frame(p_table)) + p_table <- tidyr::expand_grid( + x = 1:length(groups), + y = 1:length(groups) + ) %>% + dplyr::filter(x < y) %>% + dplyr::mutate( + group1 = groups[x], + group2 = groups[y] + ) %>% + dplyr::select(-x, -y) %>% + dplyr::group_by(group1, group2) %>% + dplyr::summarize( + p_value = permute_p_value(merged_data, metric, group_name, group1, group2, nperm), + .groups = "drop" + ) + return(as.data.frame(p_table)) } #### TESTS #### @@ -182,4 +195,3 @@ compare_models <- function(merged_data,metric,group_name,nperm=10000){ # shuffle_group(df,"samp") # permute_p_value(df,"val","samp","a","b",nperm=10) # compare_models(df,"val","samp",nperm=10) - diff --git a/tests/testthat/test-compare_models.R b/tests/testthat/test-compare_models.R index e5c0a867..e5c30e7f 100644 --- a/tests/testthat/test-compare_models.R +++ b/tests/testthat/test-compare_models.R @@ -1,65 +1,105 @@ # get_difference -test_that("get_difference works",{ - expect_equal(get_difference(data.frame(AUC=c(0.5,0.8),type=c("a","b")), - "type","AUC"), - 0.3) - expect_error(get_difference(data.frame(AUC=c(0.5,0.8),type=c("a","b")), - "type","type"), - "The specified metric is not numeric, please check that you specified the right column.") +test_that("get_difference works", { + expect_equal( + get_difference( + data.frame(AUC = c(0.5, 0.8), type = c("a", "b")), + "type", "AUC" + ), + 0.3 + ) + expect_error( + get_difference( + data.frame(AUC = c(0.5, 0.8), type = c("a", "b")), + "type", "type" + ), + "The specified metric is not numeric, please check that you specified the right column." + ) }) # shuffle_group -test_that("shuffle_group works",{ +test_that("shuffle_group works", { set.seed(2022) - df <- structure(list(condition=c("a","a","b","b"), - AUC=c(0.2,0.3,0.8,0.9)), - .Names = c("condition","AUC"), - row.names = c(NA,-4L), - class = "data.frame") - expect_equal(shuffle_group(df,"condition"), - data.frame(condition=c("b","b","a","a"),AUC=c(0.2,0.3,0.8,0.9))) - expect_error(shuffle_group(df,"group"), - "The group_name does not exist in the data.") + df <- structure(list( + condition = c("a", "a", "b", "b"), + AUC = c(0.2, 0.3, 0.8, 0.9) + ), + .Names = c("condition", "AUC"), + row.names = c(NA, -4L), + class = "data.frame" + ) + expect_equal( + shuffle_group(df, "condition"), + data.frame(condition = c("b", "b", "a", "a"), AUC = c(0.2, 0.3, 0.8, 0.9)) + ) + expect_error( + shuffle_group(df, "group"), + "The group_name does not exist in the data." + ) }) # permute_p_value test_that("permute_p_value works", { set.seed(2022) - df <- structure(list(model = c("rf","rf","glmnet","glmnet","svmRadial","svmRadial"), - AUC = c(0.2,0.2,0.2,0.2,0.2,0.2)), - .Names = c("model","AUC"), - row.names = c(NA,-6L), - class = "data.frame") - expect_equal(permute_p_value(df,"AUC","model","rf","glmnet",nperm=10), - 1) - expect_error(permute_p_value(df,"auc","model","rf","glmnet",nperm=10), - "The metric does not exist in the data.") - expect_error(permute_p_value(df,"AUC","group","rf","glmnet",nperm=10), - "The group_name does not exist in the data.") - expect_error(permute_p_value(df,"AUC","model","RF","glmnet",nperm=10), - "group_1 does not exist in the data.") - expect_error(permute_p_value(df,"AUC","model","rf","logreg",nperm=10), - "group_2 does not exist in the data.") + df <- structure(list( + model = c("rf", "rf", "glmnet", "glmnet", "svmRadial", "svmRadial"), + AUC = c(0.2, 0.2, 0.2, 0.2, 0.2, 0.2) + ), + .Names = c("model", "AUC"), + row.names = c(NA, -6L), + class = "data.frame" + ) + expect_equal( + permute_p_value(df, "AUC", "model", "rf", "glmnet", nperm = 10), + 1 + ) + expect_error( + permute_p_value(df, "auc", "model", "rf", "glmnet", nperm = 10), + "The metric does not exist in the data." + ) + expect_error( + permute_p_value(df, "AUC", "group", "rf", "glmnet", nperm = 10), + "The group_name does not exist in the data." + ) + expect_error( + permute_p_value(df, "AUC", "model", "RF", "glmnet", nperm = 10), + "group_1 does not exist in the data." + ) + expect_error( + permute_p_value(df, "AUC", "model", "rf", "logreg", nperm = 10), + "group_2 does not exist in the data." + ) }) # compare_models -test_that("compare_models works",{ +test_that("compare_models works", { set.seed(2022) - df <- structure(list(model = c("rf","rf","glmnet","glmnet","svmRadial","svmRadial"), - AUC = c(0.2,0.2,0.2,0.2,0.2,0.2)), - .Names = c("model","AUC"), - row.names = c(NA,-6L), - class = "data.frame") - expect_equal(compare_models(df,"AUC","model",10), - structure(list(group1=c("glmnet","rf","rf"), - group2=c("svmRadial","glmnet","svmRadial"), - p_value=c(1,1,1)), - .Names = c("group1","group2","p_value"), - row.names = c(NA,-3L), - class = "data.frame")) - expect_error(compare_models(df,"auc","model",100), - "The metric does not exist in the data.") - expect_error(compare_models(df,"AUC","group",100), - "The group_name does not exist in the data.") -}) \ No newline at end of file + df <- structure(list( + model = c("rf", "rf", "glmnet", "glmnet", "svmRadial", "svmRadial"), + AUC = c(0.2, 0.2, 0.2, 0.2, 0.2, 0.2) + ), + .Names = c("model", "AUC"), + row.names = c(NA, -6L), + class = "data.frame" + ) + expect_equal( + compare_models(df, "AUC", "model", 10), + structure(list( + group1 = c("glmnet", "rf", "rf"), + group2 = c("svmRadial", "glmnet", "svmRadial"), + p_value = c(1, 1, 1) + ), + .Names = c("group1", "group2", "p_value"), + row.names = c(NA, -3L), + class = "data.frame" + ) + ) + expect_error( + compare_models(df, "auc", "model", 100), + "The metric does not exist in the data." + ) + expect_error( + compare_models(df, "AUC", "group", 100), + "The group_name does not exist in the data." + ) +}) From dfe817d11bd5c0087f91cd34dbe82a53fe5b6066 Mon Sep 17 00:00:00 2001 From: "github-actions[bot]" <41898282+github-actions[bot]@users.noreply.github.com> Date: Mon, 2 May 2022 16:45:11 +0000 Subject: [PATCH 06/26] =?UTF-8?q?=F0=9F=93=91=20Build=20docs=20site?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- docs/404.html | 4 +- docs/CODE_OF_CONDUCT.html | 5 +-- docs/CONTRIBUTING.html | 5 +-- docs/LICENSE-text.html | 4 +- docs/LICENSE.html | 5 +-- docs/SUPPORT.html | 5 +-- docs/articles/index.html | 4 +- docs/articles/introduction.html | 4 +- docs/articles/paper.html | 4 +- docs/articles/parallel.html | 6 +-- .../figure-html/customize_plot-1.png | Bin 26642 -> 26656 bytes .../figure-html/plot_perf-1.png | Bin 29976 -> 29919 bytes docs/articles/preprocess.html | 4 +- docs/articles/tuning.html | 4 +- docs/authors.html | 4 +- docs/index.html | 9 ++-- docs/news/index.html | 41 ++++++------------ docs/pkgdown.yml | 4 +- docs/pull_request_template.html | 28 +++++------- docs/reference/calc_perf_metrics.html | 4 +- docs/reference/combine_hp_performance.html | 4 +- docs/reference/define_cv.html | 4 +- docs/reference/get_caret_processed_df.html | 4 +- docs/reference/get_feature_importance.html | 6 +-- docs/reference/get_hp_performance.html | 4 +- docs/reference/get_hyperparams_list.html | 4 +- docs/reference/get_outcome_type.html | 4 +- docs/reference/get_partition_indices.html | 4 +- docs/reference/get_perf_metric_fn.html | 10 ++--- docs/reference/get_perf_metric_name.html | 4 +- docs/reference/get_performance_tbl.html | 4 +- docs/reference/get_tuning_grid.html | 4 +- docs/reference/group_correlated_features.html | 4 +- docs/reference/index.html | 4 +- docs/reference/mikropml.html | 4 +- docs/reference/otu_mini_bin.html | 4 +- .../otu_mini_bin_results_glmnet.html | 4 +- docs/reference/otu_mini_bin_results_rf.html | 4 +- .../otu_mini_bin_results_rpart2.html | 4 +- .../otu_mini_bin_results_svmRadial.html | 4 +- .../otu_mini_bin_results_xgbTree.html | 4 +- .../otu_mini_cont_results_glmnet.html | 4 +- .../reference/otu_mini_cont_results_nocv.html | 4 +- docs/reference/otu_mini_cv.html | 4 +- docs/reference/otu_mini_multi.html | 4 +- docs/reference/otu_mini_multi_group.html | 4 +- .../otu_mini_multi_results_glmnet.html | 4 +- docs/reference/otu_small.html | 4 +- docs/reference/plot_hp_performance.html | 4 +- docs/reference/plot_model_performance.html | 4 +- docs/reference/preprocess_data.html | 4 +- docs/reference/randomize_feature_order.html | 12 ++--- docs/reference/reexports.html | 4 +- docs/reference/remove_singleton_columns.html | 4 +- docs/reference/replace_spaces.html | 4 +- docs/reference/run_ml.html | 8 ++-- docs/reference/tidy_perf_data.html | 4 +- docs/reference/train_model.html | 4 +- 58 files changed, 143 insertions(+), 173 deletions(-) diff --git a/docs/404.html b/docs/404.html index 0f7568a8..986332d0 100644 --- a/docs/404.html +++ b/docs/404.html @@ -49,7 +49,7 @@ Reference
  • -

    pkgdown 2.0.2.

    +

    Site built with pkgdown 2.0.3.

    diff --git a/docs/CODE_OF_CONDUCT.html b/docs/CODE_OF_CONDUCT.html index 7733df56..ccf732ce 100644 --- a/docs/CODE_OF_CONDUCT.html +++ b/docs/CODE_OF_CONDUCT.html @@ -26,7 +26,7 @@ Reference
  • -

    Tidyverse Code of Conduct.

    Our Pledge

    @@ -156,7 +155,7 @@

    Attribution -

    Site built with pkgdown 2.0.2.

    +

    Site built with pkgdown 2.0.3.

    diff --git a/docs/CONTRIBUTING.html b/docs/CONTRIBUTING.html index f90440cc..632efc65 100644 --- a/docs/CONTRIBUTING.html +++ b/docs/CONTRIBUTING.html @@ -26,7 +26,7 @@ Reference
  • -

    Tidyverse Contributing guide.

    Fixing typos

    @@ -118,7 +117,7 @@

    Code of Conduct
    -

    Site built with pkgdown 2.0.2.

    +

    Site built with pkgdown 2.0.3.

    diff --git a/docs/LICENSE-text.html b/docs/LICENSE-text.html index fd319ad2..c35487fd 100644 --- a/docs/LICENSE-text.html +++ b/docs/LICENSE-text.html @@ -26,7 +26,7 @@ Reference
  • -

    pkgdown 2.0.2.

    +

    Site built with pkgdown 2.0.3.

    diff --git a/docs/LICENSE.html b/docs/LICENSE.html index 8cbd8d39..757d2260 100644 --- a/docs/LICENSE.html +++ b/docs/LICENSE.html @@ -26,7 +26,7 @@ Reference
  • -

    pkgdown 2.0.2.

    +

    Site built with pkgdown 2.0.3.

    diff --git a/docs/SUPPORT.html b/docs/SUPPORT.html index 7fbeee2d..a297054e 100644 --- a/docs/SUPPORT.html +++ b/docs/SUPPORT.html @@ -26,7 +26,7 @@ Reference
  • - diff --git a/docs/articles/index.html b/docs/articles/index.html index 43fde410..11263a96 100644 --- a/docs/articles/index.html +++ b/docs/articles/index.html @@ -26,7 +26,7 @@ Reference
  • -

    pkgdown 2.0.2.

    +

    Site built with pkgdown 2.0.3.

    diff --git a/docs/articles/introduction.html b/docs/articles/introduction.html index 1fb497ad..4213659f 100644 --- a/docs/articles/introduction.html +++ b/docs/articles/introduction.html @@ -50,7 +50,7 @@ Reference
  • -

    -

    Site built with pkgdown 2.0.2.

    +

    Site built with pkgdown 2.0.3.

    diff --git a/docs/articles/paper.html b/docs/articles/paper.html index 1b1d5df3..4f4af106 100644 --- a/docs/articles/paper.html +++ b/docs/articles/paper.html @@ -50,7 +50,7 @@ Reference

  • -

    -

    Site built with pkgdown 2.0.2.

    +

    Site built with pkgdown 2.0.3.

    diff --git a/docs/articles/parallel.html b/docs/articles/parallel.html index 4d9c6fe8..40b4bcfe 100644 --- a/docs/articles/parallel.html +++ b/docs/articles/parallel.html @@ -50,7 +50,7 @@ Reference

  • -

    preprocess_data() and run_ml() use only one process in series. If you’d like to parallelize various steps of the pipeline to make them run faster, install foreach, future, future.apply, and doFuture. Then, register a future plan prior to calling preprocess_data() and run_ml():

    -doFuture::registerDoFuture()
+doFuture::registerDoFuture()
 future::plan(future::multicore, workers = 2)

    Above, we used the multicore plan to split the work across 2 cores. See the future documentation for more about picking the best plan for your use case. Notably, multicore does not work inside RStudio or on Windows; you will need to use multisession instead in those cases.

    After registering a future plan, you can call preprocess_data() and run_ml() as usual, and they will run certain tasks in parallel.

    @@ -296,7 +296,7 @@

    Parallelizing with Snakemake

    -

    Site built with pkgdown 2.0.2.

    +

    Site built with pkgdown 2.0.3.

    diff --git a/docs/articles/parallel_files/figure-html/customize_plot-1.png b/docs/articles/parallel_files/figure-html/customize_plot-1.png index b136772558e22b3c384aa70b2378d1db79c3cfa3..558a7c40ee5914cda1479998b36af0f330dec318 100644 GIT binary patch literal 26656 zcmch=2V9hQwmv?_sELZi7)3!K#z7PWq6kPg8YKe)Dowfx0!o$M8Puq26wpCHIu<}g zK)Q5G1gS&sO<@>dfRQ@%`JFS!-t7JF-rc)5`=8JH!7$^z@B1z1Jm-0y^WB@LG*r0O z^R6cl2wbF}emp}UtZ^q0R{#9PTKJ9eXZ0rVW9=_$DnG(cc+DQmC&P~~Fa4zNL?H0r zz(1=3X=COD!nXv{k3VR;M-F#-73jD`vBy}s<;LG=e?{DK;P~Xau7mma59+)7Yx8E6 z49bODgf*V`=Z;9sgY=e?B#=70-D$w&Cl@y_pZG( zd@da0eNb?5KqZK18SN4Oas$3$bI|Y*{v_PHxBQD!XRL`<) zOZ7?Vxd}HM-S?#l!H+lHrZwHjsY)uhKivAD%TU9^Qg{r{`zzn@nFN005z?0h_>&+c z^zk$N`Cpg#ryuY?&JAC{x@c!t-QG2wO)R^Az-)aqjO)Nf__s&6S}!Iua)$BG9#^Y(9l1^!^P~BetWGzo^1>VCZxlBX2sLkbZwbI)jAmFhK2?}%&>~Tc z-8!>ceum!(&|!e!^=C}2)qY$*d7A8{MW@kR`X17WfSvjuVY-fJZh>`U_3q{6)lRS0 z@O<;955lMCtMBTH<}vj?K&tELh#bl3F8bpdIA~cNxshM)(6JkvsveVqeiK*+v-9&m zo!S5O2O7^#!YDY$yt=xM9w0yZ6J;asp(i~TaWU?Jb=7dEtI*5x^{-zZ4}>*fvkO-* zy#9o(mRRx-9w~NyzhUSEzkFx6dWei)zqB&c8)r z&Y!g3v~DafB~8{vy>LThJmYDS?&G#>@8n*W_DMCDw@JljaW1tX0viairTEQ9DCc|L zCz#&|DmI?nZhVj(l6ror%G880BTny%pcnbh@b2Ox#D>gJq;Nw1jSaCf0R`jbqO$CT zi6ElaaGFZB!bIg2VF)Om6o@ycj)m;kr>IF39wHs|U2w?yk{XO(0Ajf=k6--?-x-QPQGvujfVr;oOF_N-ENm}grjEH~kNhl1)kyHoc?p?M zI%Z{2m2q&l1V6j) zIN#sE8jAUu-f?x&dN8e1c>2+iRv&( zOAX57+}zwhriSa+uS1H;{Ev;pnho3fBDxnWIB{Y2or=<#Kcs6VT-oG~Tpaq91 z`FMti8HPmI7v&Dzg_NhJTOAelmV4>Rz=6E;Z=C{DCuY)16XNrp+HQc@_nh`cLBt@# zllP1RAWMLj=<044O%F22!$HI$TE4QrKy-{dSK zqFAa!LqS$?1E)PZr1Nj@!;jhte{>B(gfRenJGP#8o&P&=(>#fa=oi9?y0-)pN*h=L zcUIZUa+Y*n)XV(M+{z1=il(ztY4F??1GSHw-f+qEg%&TEZ;~5`UFer79y}}K@ZtA{ z#>UpGJOIWA3+H#?AB<8*c)lp$JDseH_$DMI#K?D+<^dB$AqJ}o1($a(?P#|)tQqrV9@THpBWS_%-ji?6yVf z(4jt2(^=K*R*Q%LGZz;Z>jk%C{gq)sRPTOCYkt%8^Bi>EnqleI{Tw&uOU9L#`8Q#a z6rboAc?QvZ>O=>`xrCykqKf9VgtH^F^cd=w!jX4~tz`D1q{n#QO8(2g*)>xlvrX|c z!pjrMM!5%z7f0-a&Sxyq0VJN<1ov)r;X=!^3Aa}BP=${#S6lNJ)1$o{8&0dl86?{^ zO)cNE;AOWBiK(fYPEHdBrF?vRZ;kP$V%2OGum~fZIui}snjkllE0o$t-C#0M>oqKQ z?vyLv_b*hKs|crDF-X+xD<+SndN4=2)y+59NVzu7kl$_IeS$UPLm-rwV*omm(4lNO zz3pZ`X*@zUBl{ia4Gg76Liy!vv`=t#>?1i=2@1Y@s%q}favPQZUS(7+UaQ&Lu1qwS z5hk2xdxGF+u;-)Q@0ZsJ+&8o@bSY%o55OLH;=<&U6)z3!kRXk}QC?&jTz1P@b#;Pv zHDI(siY#dv0!ixmO`BSLmQN4!WS7s}WSR0(uQFij!0ETz@CJhuIS*c_i8b<>YW&l4 z*KMG4FNk?e|KJ(DnEs|P)5!ygJMOzTPu z@45f{LTaT(`sM9!3X?U*NEF+gPU-pK49!weleUekr-~4J$+O?#`~+i5PYyaj-Mr?NvQ23Pp9!y&%SuveK?KgI^i`a23?>iT{LJB0;M;AYA;l0$Sq5V6UuPfEgf$02_=ln~ ze|vyWvhpLDTD6Zjg~Jgp51<|XJG_3r8M36Utu4Q1kLz(}SD{w2F8psInpKd%m+)vK z&flqHPaCbUF6(P7Y=L2D>;GF)U`;^0F&loCdtRPk#9M zv9%kw-+2@uCju;(afpOzQPZ4+&!8F-Nj$|fZ;551sReZhn8&yk8yu<{OD!tz-z9ns zrN+l0 z`muYGyP93Q3Or=z1|B&9DtI=w1os$WomwcZJXaf7m!y|^GV-v4KMNR>uvDKB>)C3z zq9Em^vA`YDHZ5{YNw?9CoxU^g!(^TMc}tludILD4(R*8R98i|r{78;zsn^KJi1kQ* zeu|H~o?w!0Qtm{xyi?y@71sh*+xPQrF6?%opn}Z;6PD3bh(@1dfIvE<*Hm} z(L9y}E=#^Poy(Qs5%q{e7tf65GmARBC#o*m+N#4aT2zGxL((RaG~@;!!yO-h5qA^t znXG+gkR<{$?$L939j*qXAQGZIhvZTf$x92vx%6nba|>74*BIz8T3f5Zm%;mqJ6#%1 z>le5?Om~ynCu5Wfdka#@i;P6wq}d_8V%Ua^-N3g6HPuhKuG$bBEI_DGPd(4ofwo9#U+3>|Ba^9XrJ z-AiG7*{-D_D|Bm{So!mx-;B4I1=9RhZ~Dp1%nTr222{8-xN_BIm7RBqvZF70&m?L$ z9WE|fYE;nE)603){XT|08zXt{r610mezuLiNU-X3hZB`6Is@`w0n>-6M#0g@eSe+r zAh0+9OU)x}v&#;Y?hIn27Auh)E5vg8!Yrd(d!}0CmN$7rCX{vJk52d-FlGSY%A3S^ zjphzECN&#aPq$je1T;hPe9gVTqqYk!^sF0f&tV$So12^6e(eKXs{rfKc#a=# zMZlx)?#cV#?}G5tb>S-wN`zhBK6q4B$^xCdM3Y*)M+~F-mYJN?=LTuNyyays_6aKwy88?~xRJtIf)Zr-MRVrGdTFJP za^4HQf>ge)kJmM#B+bk35zFiVesnEfjR+5SeEp>`k>op-Wa@HJ`Zt@_l7tVfCoGMC+pTU-WY-*AkLO(ksLn3QKLm2J=&l=0OH5mO>K6 z2WVOUuL@G;cX|#y2$OTM!spcJ7A9?iNT89N3fqLeG#^ClEe-7MX%k%fcw_cGH@PJ) zcW*Q#OMFeQ=(rPoLh*mp)g~M>$`4|A29xFzs82rVV#4uU=;{-d9Fy zVn1GQlfo+and!~G+1(#UQ}diuS{?31Z;=^of5d{iDb0I!JfN8|_gS`!exb0Prq_r9 zK-1RmSdX_u#6Q9{*j-Fkno88%s<>9lt}k1@?*S_&g2&KH`lDTUlBTEBp(%jC=bVP~ z7%8|10~DuQI$$Xs56rmUtTZMM+w(Kw2MHXuE~XLB+{h^yQ;gHM=TS|>7Z;C)qUnvMG;aP?-gVp z#$ZDA>Z%`+teHUW5%9s4vEEWk2%aY?5bI6PGV?ZwzxWnzM>FnB8SJT?5-9J?Gy!a!;Up~#yxu^eR6>CjL&Uv_;Y}*VVdnh}w}LK!%JnrhHMWLI5JJRfKkyXa zC(1IOWjGGTg<1^O#gQlwa*Q0yg!nUj7>3=d#lDMqv?3e*l1ucU2O3lOfxM>oYN47Daw zU8-p@hG32eHA&8^aGZPLuns& z)PwvX_XSk5@J*Xj_nU?&7mj;V`rMo>coeu6{4v^K@%hhgb(@4qq;*Q}9@0VYnU9cx z(^@VA8vI(bNI{0X_~Xe?s%cwxdhgVRx8mJ z#=QzsKH^rdk*sFVW@nmHYuj@ATvi6Y+_LBVa7$X4f`=orRr^hg6fnj?rgEGcOt1uO zMXi$QE-3N9D&)!OM~FFM*b8*N0u=x;kebpnv$AND^_t8EZ@|6Xz%13nM*u-D08s4g znQLT7#~UQaCnt*~=_VV{=ZzPQvN@Cpfza?5h!3~5^kxd;3HYt5AXkS&f$-1Kbjca!aDf{VHpPnFA=Z)3xR2sBi}uJpJeA zegJ&a=#Xla)s|@O+|=(YZYaqCK?JMc^(~`jdpDLHE=<r z*0%j;@4#=|cEVO0nFw$z*i9OLB)E7Xik=N43*4XvB@(tMJlosfw&o5!O{!~XxCcO1 zniJvsAhR(VT;v|fEB)8idg3GOR4Is1Vb3tG8RawGN}|}4 z=i3<_3ZI2SLBH$;kb4i@(NBhkh6oR@ZjyVVCD{dWEm2!Ub|kYNugmTG@hXYJ^rp+h zFQf9B;)hbG=GGXK8t%YVCI{< zXcGYUE=0=ibZxIHYQm@NaPRd;tnCPj^y3E)c99&reON-IwcG-cg{-c8t>%SUDmj`s z+d)?fg`o`Sg%sha(o@(+QG=1sY@I09b%TYx`xBaItAI}0NiCN99@IA>92u~rB(nJ{b+2)Hf zZ-<00YeGpQ6%fcPC35Wg#xg(vGJt-LGPyLoM=W*~oz z=3>m74H4$AlSBK*X2ld8M_}3P&FOGdi}vUbgAJ?{;gTF5AMZVECRjlWHI3e<72o%A zHFqT-9Y-jLggmAa+pA|_ms4O<3X@#n#u}PaXB*D}1cHU;%gVR@ysNpSZF+jz=G_o$ zp2aLK85>y~HQvfT(m5gR02tI3y!Of0~p6u|^Oi2MSTo)W@Z*t<4d@ zu&{@VM7vDEd#6Ttfl{{VW!~6A|4+(jJY-mc(2Es0Xl0?Je=IfO4)W7s=KbDD8uY-M}_B zyW3B>enpRRGchnUZ1z|JmxuiHI$uDf86{qwi@7$o%b@hhW={PMu%QEDmn<;oNJ&`C zj?+jcoxDFcJ=2$c|Ld0V5qq+{&-`?ix=d&0D@K%K`92awINJR?(m2QvvPeoKSmeRZ z3bI)!>&D1N;?(sUw(TD{>|1%D^n6aID>7h??{?kj<0j7>Q+aHh=X?=a4P0Rt?d>`2 zKVW;aWNSATJzZ){$eIHvc4tFSRhaDyl3n=tM$)Y0HaLFl_33%Fk5nv&8WQ>Zv!Hw| z|I6mx_jDUVOFp|7KSwcLI5+V*$L7;;=lz=-mY{DSC4p85)WiYi?l|zvk4hgebN3-C zg{KaIkJ;BvDb2T1T??YW|M_Twh5q7O8x;YM^MF2|*G?tnT%2K%*(|95?wyCfhb&W0 zB`-}xI@ghq9OLEXJ^0J-f5B8S3j-dJsFe_(n0T?Uc@&qw3U*TfSLzS}?DXZ7Ux%JMm zZe3I7@VsDKXp3&=Dx(H0Bq&TSu$^allaqA&DJtY(;*Ku+b4nlMW!z>Bs<`}R$GObT znCdmttY2C!<4GxIm3-j=`>sKMVgH-GV_kVm!0VAX zkVXL>0JWuGxYEjl`K@`xjb6OU0u#$mP`ppKS*K26@$NBMtKtMSlO@0e6iP6SyA?8w zb3gnhntq}7q4=UkC(0(k1yGW-MIZ|>8mItjJ-t())JH%L&{-{knyf7{3sJE308^7Z zSMML4;W<(QbT60hTD6e2`X5g#G9w zv$1nIBJa%wye89V<>DXHe7zcMG09WNwdNnUU|5}uK45kN(WXhErw~cJXyyZ4XjH&n z49I4zIDrQb2jJ5OJn->t;zsv@QPfG&rNQ>D0^OLFnUh2JVS1U40$LG>xp?eT3itsk zZvIq0Uf%VQkFEbM79N|Ac4^WNQ}l8z@#hN#VpIf`VdS9!<^b1te&*wQv@ba=PBjaY z0+g3#)IF_UBrF$AALX*4Huqj+m{P&wSeIkpjg0MVQ%+Iv+gm)TPCnMw*7zvHK-Jx0}jXD>ttjg0$1tKJ?5AP;9 zcDgj%me}Vxh>3}Xfuc?61K*BXjFL}N{&ZiEW;L(@8RuSIL7G;D-uBfp2rxa07xrkx;Uga@qpXs33{Vie#^?v?vK5?sT^wIae#%&$aX^6h5u>Kp0?M|yl{Gx z(oD~@vuO`|BR5DCd=DVSNQpg5wH*8K7O=lQsPrj?(`~pafRJaL^@5_vqBY%|MB#Vq zH~@GAwRV|{*|)^=_L9ca0)0Sog947Dv_3q33;bAt&dn;%Lx6M=B{_b$DLmbxm}mHj zA;HCgvIGXi=Q@eHOVi&BUhxvAez@VQ+p=GjKD8i_9x>$AJ#T{R88PKa=R@i61|| znZ#46wk_qiev60uK5#vH_xXccSrNv%&7~=%L7ANV#*~X8p)FRYMI}7c4jl-Vs%vzf zptxy2*7@P+n$Le-bv=yAdwONtfh#~0h;$Yt;+@m#pfzF_;r%l-~avvgq(GqKie z?@_0iinvgdaM&#QaXtLspF4t8s5j)NDp;{WcAa@y!(=duIRBL5(t8rNLE*Z$>d^|h zaXpJb48Q2gl`l^>ub>p1cI3Gx#4!;L)a8M28@D?%`Ke=L$53O^1keyfk|%34YP2&y zXAaCJS1Ck*au;MrVGaL4!JVQGtll2=@Ow!qc_I`2RgC90>?I~UnizKIVx4Cy6E+@7 zAG~~g5X@Maa-&t|A4{GlCh9Z0=ch*^hzv2cv^b}A9CT%Dpp%mq!ki}$bl15Jhl+;K zuGtGlHcH&H=W6wD1a@+Kv0G{5h<$Q?k`xJL0R;83%z_EDWpkFu+)z?b!RePvK-ZV8 z13>0V50`hdLq7@;6Mbg>ZN)pYZ!4e1oOvVxEiLdy z5Rp8qHs3@{NlB>&I){R32ttq7_#45c?@0z*U-KnF;N&s7^aJHQXn3|ERq#T#Gl&Qw z)<&x2)jGjQB5t(qaU-`b4@-{Mi&5yyQ8MDyBSPZTLW^kILFB7m!((&@aC23jwTF{F zph!QB+R3?x`wN|{cta~wo$tLKI?&fSIB}W!_-VFE+#_gh(9_WgCePd;U(mOBZ_1uB zH5@^wcR$*PF#32891|>0N-YWp^D}1RTv~{s!ALR08GGG$6cF``ygg9T0?d8>YA%~* zx+H|`)x=f7g>RwwF|?nD5bOk@pVqxJ(#;7|ieA(3RLBvIz2$*?q%=1$l|2G3Dp`nG zPJs96$Wg>=#&lUlbrJUkt0Lu{4o>%-P@n8j@^<$k4_C!Q>W|ghUo_aR?>HYQ!vE;w zQ$}`K_PQAD!R=tLx`fC93J}cVXj;bSP|dOHj(_3IXmpI}XD!K^F!%=Kl~Yp6k}7zK(v zByQ)^r%xS$#N5N-EgriuyAq#1RRvQDVlt3wgmGdsK+4UId<5GQhy>%hmX`aUs6ktc zix)1u!-9!e2C2OJAPXFK22It4v;7gx?Yf4B5kQT7kC+SvZu8&&vN;pfcj zga*rl>N4QGTG`9PYCe0hc{DFQt~er`#%`b66nHsi-oq+~ui#ZLuLWD_Ry;W=vKIKQ z2E-`XX3)Jj>V^Y)zV_kqRwx%H0E{05JY@yZwy?#8G*IG?j9@uGV!8L({;8GV~P`*^Yv3fLw-cEs^j=Sl*>|Wy3q- zRHbb`gFTJm6K&4wM!H)IO6Y* zZIUJ8DnjK^I@Er^29lPa`1z5xEDuOeK}6Vf;2q=>=Mqh%r;uapnC_NwX3lwA12avM z#An%L!PTyvJr3%r65@J9LhOd5#6(h}wrBeyem8KzFU*bGa&Q&p`x{^GL5eOO~pccc%h0*P1MycZ_gS|?lJT?XPMlEKc@$`*Zi!&;iLS- zljmbQMfo2Dw~gG0%k!eDmd0+lw34ej#n+wxsB9W~-Et|R)9hZ;p}wr{MN7(K*Zfz5 zK0Uv>2I}av?=I|?*_W8IusUzA((-kTyQFuQT1MO*A@}IkkLww(@3NC0rtt5HYCa9+ z>)B7eJ+a?#iXp<9xKHKd-sOLN? z2c72gE$oP39jB!`b3Uaya$K3GfNXN!h2)GcdmbcX3>NFb)jnX5H`OiMIma{_-2Pku zE^&nC3>ZFPejnEzfNDqM7cNc$^t*C~Lst>Lxp!r`sq3HJAYr{KhXf#;I<<^w{H`qD zpjjZ#M!o#74`HkajzVM-;fE^@{2217yXL=*{oi@Ce|Vjh>3hV5Mbx~Ve^R#lP-b%XX4wqR`0u3bkOO7~N~7$h{FPvi)aRO8ZPEOUt5_ZIh$Ow?T~ zwR~$9vZ2MOT~JMSWHax)D_*|!RNQD2v_)}|u6=-)@b6XL@jGkMN{JNr(z=D_AX@>PW&ma?t;F$|5TmejF*)&v&&&G)F9h z7<=|Tn2wL+;=FR;*ba;a#_v!>L6jOC08j4x zwHn+4y#6m+a%J6W+s2`o?SQZVM5;ZcJU3=EC}6sWK~g`ZF^W@h)CAgIHUe3R6%E+> zP7NN0AqZQJNB{1i!~L&>v_E{srx8a#$THDAH^tCf9&x-}-koSgFAESl|Ga^Cn*2LS z5Qwoz&?9GG^Az3L9jK80H9Y*2*t)jHeNrh;oo`n+7i3q3c2BcvGviNQh7Fnb+cyPq zXZ8VV0pJ*X*yqMY8=FeBazQ-`_&>SMfj~fmCCYWAgW#7!P`Op3A)7s4-W441{P=7! zTjpUMTxIVQ_$p%(c)ef-MEF+|z9k{=6nW|Omm~`E+fc!5)tJ3;w z!LJynuV!Diva4!%NO5?Un#ogy-Adjqo>Z9(Ij?5!1DsiC4d)o&7xooTY&>V2oyakr1~!|{pU zVq0Y`^fnpV<)XM+4!SmiW*i#4C;ax1MI?dHTn03!S_pb4uomJF`Zf075Mr$dmHIk& zIL_JQP!aO(ZQAFKMnWU^_t(VOOCN4YSO6Tyx+lF}L*BOG>$Qv~8W0G2XwU*=oMYbZ zFVl4&k)G+o2OcNX#sRx9Y`wtqhbFnWygbjsG~l*r3&WPm6`&azOq0DvspZy1qiI z65cGJ(kiH8qrenjtOoED|Ghn|c=TvWUo41O=o3cgP!NhaJD|Qtzq15DdPG@H$R(ir z4_FY1^C;~D5#!h&W@!P19g5-Oqi)4sg=uL0MK2B(P?GK+djQYVK*{A891(c_KAu7ci=GdWpJPz`v61G>v0*Qe!EN2<6&4gG9+b@e$$B@@Wu4RgN+3_br;4-xNgr zg1Ro$0d@V^(A{VbEDDlp4nHSej@3SRC zq=2Wcey+cU3B_g*5s>1WwJa!>Y^5(&y3ni*t*m9l?4X!y_g=K$&>>~rlNurqE!v6mLBfOji-{NVj`zn95x-CaU*c3LQI&VgDzGb_$S*hnl zIyWpIX3a)MLvu47I7FK%Pbd~?J2-da>M@^iweT9UfTVHodI_+?4!vnx54)M#{!={`Gq7mBR3#%YiyAAV)sq^qN2jvfFCFqBDE zK3L7&7F1SZb(6`KCsq?$h$xiT#i?aD57r?ojG^1_VEZ#6Uf3H4LFYr4nCX05^LUlL zuOte}pdSc8K=Yn&?;d>;X)XiSu?|m0@@y)5E)^RzLE9t8Ks0{_ol3~0pt}YOI53ae z$Xaxs*;ElCRsj@SJ&4h&Z`ToiqG*G;c0o)u+uJ%K!f8CJ_#M;KGrR7z^ab`Rd0P5z z4b{}P>==7!=rV^9jUa-<$Qfh~W$;A{lqoOPD<|iSfEa(08m%Y^d$~$cy%pS}5*(@+ zZyH&-E-No%V`K1YFq2uaIiRD1yv%$DA=n8Bp7T&ABh@V@DGg-Jbnymuq`ZlZS8j&b z+rkae6;Tt>)eiE3Pn^FY5Za!%uCa#@Q8Zg+-w&o*@*qggiily|S-}4%e$yr3<&LM1H2Um@7 z53urQFa?5Bi`;DTOJ?WTOM8^PK9CIxeVHo7_xQc+yHgHG_&Vzr&G*;`5$S^7NB0a% zkJ|$E5&Ly{+Y8uHNRhq+LRW((j!XH-y7u9wDWthjGXXKf0t`(g3J{3MH6r6gA!00* zVJ#!N1^+6O?W~NLMH>xhP*@mHQ7#YkoqVL*9te~L+N;|;u#<;C_$VWt-_wbwLjXZS z!%BUT)g1$*5EUFy6-eNq0MCRRaz;m|akD-LFuMm;iMt0q1|bEhC$tc3e#3d=`DTS# z=ts2hVt1IshNPbOavGyHbmnJS(o+s{{%p13>f>uSPxy zUTapWo`0iaS*d$h3N&D`&A~tf&9{O|1+Oo{#T}Q+3tVXc(#`+{2L(#nj{pSzE|y%W zK3Aab|E_?oxdt9_av6F<9sAy)S@((aP8gxe6|@ed6o8+-91Q|1YC?Gx|NmUN{{iRz z(|cPP1>5Y_JqeMHs+9BKT7B||n&7AIA|EF}EIJ7GYDoKQaTNYgbN_o!&H0K?ARL=X zKr-M~;-{`{tP|{zKU~<_yWc2xKTBevxQbibVt!IE1orAu z-Fo<2(<)dw?vhw2+B|F<3O5?RTo$2NO)fqx&ypyxnOGg%D$H7#KiOh?KF$mVa_v8_ zt+@49LrnyQNhl&d(l~hi9{q8BqUreen_EN?p$-uL15^Ttt$pC~R;Abge<|DJw^}6f z3#cT7Ii)OCas7ZUG$^+P6y<>1f3`ete;vAiy7d0_J-ou}K)0|Y09zsnaDpfh0ffCW zoO?N=r4g48|iIHIJ>(dq}t!1Y8x6D$t+;R-nlL5To=d5J2e` z1fCyZ5|CQje=Jq;Bo`<}lAIy5#s>Xke$N_0>dS3CH{VLofcUmvjy*k zKNVw(&2}x`(3XRyH@+T>=uhtaFy-UeenW%tQQl3yZfUh^d9=?uU7Q)dSlmqO+cKC( z)`$=YBiS^z58y|5wLOM+;Kd=Jj_{0+S4hyIlcEZsSa`jr@+Yj6v6e%jfY}tyFC+>C zZVoa#Wrw?T<)6dPzf(0f`7HsrU|$@Ch+Q72YOlw$Y062Q>tl`c$qAoN#aStGg}kp6ir#kl;xTNs~qo#dsSmzB|q@+53K{Nq?4DS_wbU^FL0{r(>S zCg2UBflNlC41=%*TKAc2*y&Bfy)tk#MBF$$qgqasbkKA10zRFy!=I1=5DhFeBE(CD zrZObIzG(pk-=HW)-Ytc-jd0|~c1eqqC;WMJ2DcyK`=(%PQ?g+WRC1fJEpZ8q+Wmp# z1+%!_w&_rpfVL$>f}dwZf+2BS2!w^Ktyet8BM*c{eMg7U`yY>ZarCcg{~UMhTrZkR zw%0klxyFRi??`!OoS%J#vf<+W~@e5zy^qU@N&#`v}pCPdfwQe|CL|gi+8u89NmkKa&Qix#YmKdJj z@GWZ7SRsMZj_sDg0jo3XYaSe{2bU^QNpdyqw2sISS#IbTyTYj^Xm@~Z-^S@(L<=Ty z79KeoGY{Og09vl>K;R^nLF2*i9zv? z_95WmM<(05m!t!g<}OgqL$})wIp+)4wq#M!Z3@k_Np%POq+eWh+u#5WR}e$8!=<-@^GhU-n7cKpTw_U+<6 zcW$V{(Fp~~%vq`tNod?8^Lsf0(BDCkPr;TKr+Oo{rN1g76nHZt$rZG42$e~ki-E@n z-{|vSC&{(_`&iEJGAz{5T|*>+FbmxhCOGMeAY<#p8x$D*O#<=B>{%8#T!7l>CfSaZu-GFzQXfgaYT2JV@OL`@$N>~RBj9C)5h9B6Gh zU7ps^M5muUCMUPr#$V1;KoQ6?T!8_HKl~*UxIkzF)`O|zw3b8TJCn?YHdj#bOFc^g z^*NzA!22)1x{8!aIcU-27zcr>4_CwXK9tA7Zs}3zDB}RPiv;rdy2m7}xPh*l9uM0A z5S6KW0s4$c6d2RYHeFCZvFU*{4mQCM+k${t@Tv(3RPPl6IIjPa)N@%{0 zCW36<+LQMKL;PDHNB38_^GB`9K0&*$LXROY8QK+(mGA`-*9o|vtFVUUr9nYzPNsNP z-2&SA@nkk=ukrv5IR^G&9N0GaTzp%2gU!V4KdqG=PJ?CY+T=!T#5DM`NDO?#bgeB3&%0(hJ2cRCjlui!u%$h0(77P1th3`fSR_g z)35-AJ$mgP$|z?z0m02_?7`+i>~KKWWmP%BSVy*hJbroW)1uWtnc*0n`T#6BN6fzv ze-ly)+gx(5d(J^d*Fa-KL%~Q%|Iljm6?rAA#1%R9cqX%O}-6?+~aJJ@^_ zH)=L-8K0@#=)9J#`8;?K^c~81h<1D(mYrIMq2*ie(_oY<24at-AbX}1Te}f%k{3$J zm1rN;wO;g@c_Ua26!m-EQts{Fe~bJ8ggZPDl?M8L1r(n$vg@2(X!k238PdXg znYbJ1P1G~-FjtyPAE#Bo*RdW0dxZg{TzYx$sZbxGE(Et*2}*DT#ydP{z;Lt}+*3#y$ISdL42_!| zR$htPR%Oz_3i+2~jEJlli3){_9VkBl$2Xi=wx|LHm+@-RDD>wFnxy0@J6QWQU~A9< z83W=Q4P=ayz{ta=(fBGso~?j>4LCN7FDupyn&6fbhezgURWBF?(oR<#fW2r5(^~~Z zH7X(&AP|Ug|B4s}?cUghgSZ^p{#vSIzRoN9n|neNI5@!LOpAT_yk8i6fN*pW!%>n* za{f?;jt9gal*goV{6YOexd$Syfr)6aDxUO4A?6M|ymj#WyE|39sfDEk2IEO?2Xqhu z%_|Hv{Zs(_8)&KYq(Fif!qZGLpw(ZRl?*H-cAJCC!mf>btzp=-(H5CQG41lpA#!ql zxZS>35qkxdW?pgkL&){5aLN}yHj90va6pz+Gi14xMpw8`RV;r*0M<>vLpKs`eb6dR z(L@gxc6qbXfU2c=W_SPbhlz>%NOeu*Ef=WnBeHAHnyOUMWRxGcbxw16OiW`ysZ*1v zdZwRo(5WA#Az|RP(C7!SIT#}rFwq?yf?M(mrcJiAhf_aIs@$B`WvOVsCMZQa>3;Rb5YByMo&+3_1t!C>U%Hz;j4Is%rW2 z%l0OqGNIW}EVXFPs=^vX?X49+UD@#CBif`{P@|*HBsL6zS8HEt!MG|qn1x9F)R>sN zyHzY2pXzL_n1lJtzLzi4ap@S;5KiU$614YoqjrVvBg1DyL)qO=LE}guvzKN|A&@P% zWx+#8g&Mo+3&$IFLu;C0k)WcdGps$45MT{jJOSUp5_w+xx}%>0PmENGkrdjoc6WB_8zQD9>Xf%=OEeFLh1{?UsC2#C`s z0baZcv-Sx()U(+?>cr)EvA|7tGX+ed*tS0AP{W8=yArLIKvTa ztl%8*15U{Zv&qp;d6fA9y61E;U{#?LlcE4L0onra2#=5v(44i==YcrdHoyz$Y*44b zOE>B!3yl^sXTL>#Hy6B}PEaSSV-vudBVkau0F;BK%8wf9OIj!3KAEd zFZimvOiNP{+kZ227n4jEt+0a4%C(Kc1P<iWm<)FbTb z0LyA-+cXw^;IvhPFdb2fr3kj&b~~oXSTke5(ADxaI)L_O!Vxoxaeea8A7W^03f@vo zVaQ9oxxr1v^bJkX(7(42+b&81_#*%jTf*782)1S1IR|lJ6HQVx^>FWPsC3cEE|w|< zE^%#5f9+T3BY0;HbXjW1O>=1v*0?vk8A{TaX}9Xru4nKBT#PW32rV-6X~F#f}1%3$_zYCisLj# zZ?qvl3krp9EL70Ti?aaz0%{)E3v1=!!)9>FPcYjUAB zRwJpBRN1CpZri?YI70hTJ_55`rbfOZm>g`1n2f;w`LvN7`?lUBzVb#7KM2CaJi%I9GM>K-`$FtUS59 z4D{smV-aj)IP>0?7zTFyZ0q^7KydH47Lqn*x7L_3f-m9RUadb?V zw(HnSQdLz2{V=%j_GMFB+hh?@Q6W--M${QN4ozO+4b8Dt@pC9qjBL5J`kpz`}ouA!(JUncp9(M3; znk@hh%H#fi3pj|hg)U*7cb;uxU@$bZ)Y{tWnBPfx%xF$+ryrADe$F4TiSTs@yL)Lh zSwSF>ST^D+Dk-T2x84K<;yntb?)@B8MJ=WE@#-7qA~Q1&$qy7b^j{lz< zzQauJ+18zVTlQFAzLLxqC4qLC8Ms=;P~LO$32`56^ntmTr%#_QFDtv-XGm|BxYrAz zm+%FTBH0+|)sOZKO-)uH&Xc_h$#1g2-{O3~lkZ*ne)R}x_2A%OXr^&^205ifaO&0e zNAhl?Kh@bMN69<&pM*CDp>fA}=!RC0SYqex<`*wsbmuiwx0^N8)WmWNDh*Uc$apR- z0V^y&P=gOX4x+-XDzv@;ZlqQmXg=>vuar4qego{VSrU4{sja~@Z$!}U(TB=MvgFr=*2~@Fkba|zly;VIM#jn} z9pr|*WUwfyQefO!)VUtl{NVfd_mZX!3SPn4Dv5b{26a@KRr8@5(F*&@WUfwwXcq83 zh1l5Kd}E=twzhV9k23@v_QesXPiU_BSr%2i^G1hnL#`ow+rSz5oGhExgQOxCZ4d(Z zGE*R(-n@Aembx5F_`CVL5fP#!`0H(G(Z1S*=A}}7ef?cbmAwkS3uh_z*Bt@{6n9!H z2vF^9)2D?Oy3-2^jG#+0Kng57a^$6gA~HKA*NNUne`5~umq2*%Bkq~JWIz-9_Cc?I zCXfKJF7mF!HQ@Z}oAP9dIl(;RX3vzo_`iTWXN0|OWQ z1(i*wl1+QI`kFOsDBip5jf^7qSDb;{;e_&XDmBOY zvuhRZIJHOL9weLH0)@oy`w3bFPt;+DqjYLaGeFX(YigujFDxt!g4SsbjR4U7riv#YCE0ZqBm#zD0A-TCUk%@{u5N&HK%dV<&^#{_dMkv@@{bzkat|cfm7!aeTQ)9zL=8j~`ye5trb%Kkmf^#d-bZ x)aT144TxtCa943&-yYozl?vze|N9pYDO>ZMQQw*)fkGRBbX?=dl%waa{6E3;3wQtk literal 26642 zcmch=2UyhSwlzM+s4~Bm!@$gc??KP!m-C%_a?j2A&yze7gqipEmc7?rYwb5zk7}s$e7ose z0)fCoI`r#t0%4g4fw1({H_PEW#$PpizY@nkl}RiA*FPY+LE)KU=R9-Lxs>0XNO}{%A8~1o?4z!$_yOlL z{>x+4%xrfceO=7#uaust22Ea%CHNQ8RU3HlCkX#%Y@(Y81e>=XUSD^9!)HY#xzT(X zk*-bp%QU|x5T>?i!Xv!9)zsAF!(q{C7$uyAYH_D!@Ef6j-80?9iBqnzwy`NMpe>9N z%buR={x%wRLBBmrm@*Hatjn3v5)S)8DGXd{o#9^Y6a2-qs?6j0{xa_#uc`hU0t&Li z(H^Jky5Vz*Ql9j~%=OFg^YeQSCMHmbTU7&gdBB|A%k{j-izB%g`!#M=!gZ1O30i@-h z#t~O>(ctxOe9|Asn)Jo==nN$?4q)Duu~;_F7ukaooRQB8Ofz7 zstWqLEKbdl|M6q~@x%XNqCU+-=5&)GJ*VA4jk20=PwbcTQ0&XHtdAuf2!zquZOb{s zoreFxME|o9B))mD0+y1Vm6OxQTQYlqV&3xjl+0|`rR)6P!1C*GH==%_$GLcosBmJ~ zzH_HXix||?E8u{81^;!c@p(c%563RI(@j6ePM;Ec|Ni|h>+P3qB+11SJt-bjGup|G z4-y6>r$v`axRmGdd31d|5cn0L_8~ms`Ax@7O=FfUUA|snxlsMgl#0F+N9RCb-H>bu zTSk0gn6u7f$@%cDu(6&Qo`|#eqz~m^iGL@_5IGw@nw(}5Y~$T7TROy=(zIy2MkK}A z5A+GFBHU}nsHIFXAAcV6Hb;2Dw$99@qM~GPd)^H75^2X z-?!MD8*2WI&- zkD`1t^T1Nv4u4U2g?Ca&c=BX-cWIy|1!6+}m)r2v%PST$!>xb$HvBSToZmF@;kvlx zi{Dm;{e0>d#gA7t*DI_c5Nazv9}=(W!OAcRVbUIt@ve&ra3H)Qg#Yab^v8qoc?}lk zrYi;#VxqZg@Vo(LrAQsl{i7c*tyo(v6l!#QY9oR0%<}Yg%-a8}82EVvHoLswoxC*S zj+en22N8LAc&x3hcX<2+=TjPE?!7DX%o=j_vkOZl&UQUJ623RF{~Ua@N#9C{!~Fp^ zLL=e|T2t#I?4~J~4+K7DYRlHMKQu&@q>jn`z_Zl<`+fgKcoq*wfN*c~&Sc%5ff7OE zsf{lCORwoQ%}H#Y(j!N_7c6=33{t67)aRpJj)Ab2$90-#KxuU0K;V-y*^rXRoCLw0 zUN@>9q&<;Yn30d2WcVHo5IMUEBKeQFBctde>$0Io7KVYu6537vq_%s5mq8mR&SO{`EpN&goGT4lFjbl4RcZT>L2p_ zzn!lePfWpjOhxA8rFgg>-RHh&$$n;I=sh< zUavs_vGY$q;GYLcpYTGfTf6<_{A7(VDLf)#)fafd`pg;3R;K@{ompKUr{>IN(uGNG zR8M-O&qUcLumtJ;-j$RSPky3DxtmbTKfJn%fbZSDzitjb?}AO6Hf0Q@6>K9Nys`1v zH~=DLN8HdhZy{+$2*>zj{Igvkx=E^zBBbW@`g=X|rWAesJSPjw%Dcy=4#D0MO&6O_ z-{n273aLUgIoj8}s8o;rzFNCY&%O_!h?L+3_jOTZHq&$XN$E8K*o{)H<)x$5@UNKH z4T0(U8_w4S;85wUEQ~m#-cdA|l$=o|)xKUH(twNlD_a>rS7@Rjm`y3bx!kR zq*j*xrf~DIRNaQUy4%c2Iz6L8tYVdT;VojK_D=XGieAK)#31&p?}zbXQbU4f&1k_` z^klW1xQhQ8bz@`WYWc~^Bhm6Q)6>&=AKz@gF7O@xxX1PHJZQ03u3Yiy6)`2nDEi6< zHa0eaGYCuW5N=HrW%TIn(i&`zeMOpRu+TF^a~FMMm9pyuDeQ z4|c#Ae%IXW&Y}_swuMvmYGHdU4pYqMXFulkP}Dx3x3P#U7!*$6vxC{Wx%3Mw1P)EK z(Ubv;*hIj4+7FU{7N7%P_Z7xeQ;Y8>3wDaal1(03lvL349+d!?n* zoeQ5lIo81itog6sq~qTcgnBWQ-m8cPEw| z5b{UT-8M#b<_bg;na?{`G*F*-r84{(f{|OPQfX=oqsOO74hv&6M#~KXA^S*I=kE&) znvyilXL56K$$bWltVx|};lsA9^dnJXD!o0To)t7a`%#8*ervu<%?e)jG58RVAiFy5 z!yG-1pZVxVQBhH5d&xpOu?&{gWcX(oVEW>r(KWOWPEN~D`e%pc-wvH$7Jx@pHtafa zD?dN~U1MW!*QK>$UR&TnLjU)r;2+=NbBtKNRv=O&?deaXJwCJJpYQ=Z$x)TL>7md- z-(j=A9KC!-t%Y-g$;?2`Y#_0$>-i-KhmNNgg%KzRaX)^8sKK3*xnxKrMCEy zpDe{%gS%&N^L;sM@ZbN7vib8=T;bUq!kJ6w9HwMtWpM*iZ95ky*SECX-EH;;a591L zIIQIXi~|egLDCEosZGUsxa=KC&~v=7%_0yJjl$$gFtALfLSQ<)Hna_r&3-dF21 zcX|9&KKgs^O+$K1W`$|1-39lA3o1@M)hxS^Y&f5XxF@0_M6?2mo!7*&S6A09gA@Pg z97aue{e=}3#51dck(_@l+V4{(^0=0mBu z)zi(Ue-&6coRXa_2_?L>gF`hG2iFDss}>my0EOxeRBALaQ8#&+=8HZqScZju3-fWS zudfd)%>#kIkQM>(KSl*<3`R^lGe7ihV*%H-vKpz_{DuKl3rLZGp4du;hkrVWT|CMbQIBZs!nyX zm{BC|J2zEEX4A=OlbO|#GLt-3Mj7TLvBL2eg_BWZM>y<-%!QeDtHF1PVKUC~q5_92 z7a{e#SV>l!7|f6AmUF*4z|PL5mO#97o^I5mt18VMo0}Pl*r64lGR^MDbDqpj_Z6iv ztKBEyOq##?{ktJ!tLGP&$xPL%NVt!7WCzdy#BwL2%{rl%>OEFf%o=THny9Ll%Cbh& z90jS8rAdu84kZgSKI4}JhkeGTkIhWeQ)~F$J8nM=khMvbgl!u(lwP9PY)bY8)LYWw zPV*gTPN#XLTUBt=IG}>IcZpigK_%asw!qsy~Pc(wtEjhQSGaaiZFI-6b+oO$ z9pO|$wAW6=7$KsDA-N?BBVm%Zx^!214A~DxUp&{CzKw*13+xm4XBfD=p%mvD1xC;B zTLb!=YRHUEcPn6ZtQ{R|X!BLHo;-TA8f-N~LqmsRX5U^($zH^LD}^I#Q(0QO5PcBN zhd04)Zq2l!s|d~=#YqP|gVbH9iI+89rrj=6UBGYl{cvWDBGdzi%*@Qz$oAj!nT((v znJ^$4)7Ir6#}^`I9RA(bgFSVsf>nO3k* zc3OHmS+oYAo&C7gtUr0yfAyxl+u_t0;tCv&qbR`neyGI4cmi8Yn8e+e?wziDyHEc* zI5=3Jpc%6T<~MJq%{GWA-RYtg?b#Q0lNuy?or^D&)g$CRonV17t3N(5D|IVYWDRFF z#%ScIA$ZBDOaK!+`F3ld%aGM0?VVu~ zRvO#X!|wo^yt(|nq{GwqS9Tfa?k6y`~PHx}BeOyAQKZneDneN;mNGC$X z^#is(#>3)NK#WWg;J%EcY*ovp=Ox`b&XrCJkQW+KmM&l2KjAwD8-iGd6QcW#2~8G8 zQ3=6K!m2t_EIo+mzs2cLXgWp}4x1H39P=IXmi8E3OP+l>KklzI++??24IeN+_WmfD z{Zm)rYjJbx2VUxUM_jUjGiOrM%{w)I?uvuyDuQiFtKUf#>Tt-4bVmQfKsUCt8GNh6 z9geKAiEx=$ZeYt@Nz?72M5u4wbDJzGgdtbIyM^Q|P&FD9>3I>4$BBxM8(NIm{;Y}9 zLlY*(RwuP$w&YDJ7UeG|*jtABcZYs$V)3 zGiRK2=GcXx+6d%P?p1^jZq88QJV%UoS$6!{QVcip5Oxj(W|&~)lWMmMI&ggh9xQ5C zUa272$8RPxrgy6#^AH5bG5h$H11qODD{;}ltQvWtnH<%V?duu?^`Er7CKMGZ2+lrR;bpPLD{v7^ zcaBpT#nSB#1@@Gll+0?cmy^Zp>2#~Q2M3@M3zPSpFc{+cW)&*RY^AY1jWpoZxA!O zYm*%g|14Xzp{F5+?NU77D=tjJ(sA{ceP>g>w|Mp6k4krMIpx21@7}5YkhOB7PtA&F zUyE-e0qYo|CYDDO(M>r-L2|gA#&)Co@EW@uTfTZ@t(k46=c3wEt3KvKm&2eIYg*#R zgr^Hdwj4&7n4VseEHA9#$<$_>PAvp|IqBHzy^l@A6A7=MYc z)WnPKtX#K^%hTS4bCcV00m(Xxzm72cEW3QBETLMjO3}}x5FvPM{{m~pLPTPX7C8R* z8VdEdrp6Iw=XLVoI0I3y_5DPI<4|d9YgO&e<#-3oy|$2XY#KQbxHM{U%eDL-JTl@5 z(4Rsq@nk1nOt`>7_77hFT%!66O}`h>?I{;YKSH^&araNCtf1^-$H$%wR-$t_lMS84 z?8bn`sC3@!)VGb!g2OXKaIPmx+&d&*8Z38V2|T{SEvYb#;imb_g%g<#8KILUj)nUg zwY_`%CiQkz^Kxo#Mz{n z4+QS#8ULCqS32^IwVpacE$9}woT{|&WW~hrgxR@YtBQ!huO&4*%=Dt z$90m$;|7iQ@#UW zGCm2*BY~N?bV*;~ zO~^R-jlss`W|p73QmAOuVr-Lq02qATwj*ybM?vk~;n96lAF$=l;KRD1;F+7b>0N1BBE?{ zl8abE9uaeSc4dnj@X>b-@#+;pnu=j?&Uz9PqRV^Af+Wt=-UFr~?*9EBL9wh1Et&5% z0*FVVm<7mKrY4vvP2FEBM2hyC8<1a^YH)m7{k@*iZKBM_SL@_dMj~99y`c~xi)PvY zWGGG2YibHTC-Yqb)7swESg4^S3G4~aT_v(#ye4P^dA%Wq_35T1((;Vo)@i(*{n(ue zal0~HGI??q=)H+(Mxiii;-TQenWE+#V62jo(+*Q2oyvC@ho5Uq(#e}B7ioC+?hXLj zoe5EH29%y~+q8jt-~0!{ilL%mpJuh|%4UZ{6ojf<58i0$%7f4trr_-c0CSD}__J)P zc=X)0&t3qz?i&&mB5kfIx2eGOg(I2Oo|T)6(!I#ipB!VJ=pdLUDWwlJk#&! z8F7PcF)*EmS?^!^TfM&aJ*Mo(2L?PkwI1myvLohQs4p*9Wjsn{u#HFDOIXGf^Qr#2 z2$vYYg}Gi}`6{4#j1ipgT9Mavag{91*4DPdwb?k7$m~~DYKvnP&reh^8`J%Q@!iLY z7#b68OTWeaqO{PUWHC22kT;xBA!R)?J&*`1yGXqpZD4y3s70lR=BAfV%;+rBm{ihb zjE8)!7AmGrSuG&%5i~ZPsOh^COAfK3sd}Gpwkv3YLS`1MJ(VPkOQ~ao|Rs?NkpvLufIDJhJ9- z<*Yld$uGUzw1x5bg(b5#l@sD`>JhIu>icZTik|oghjKWmLFc;P*w+_@hle{&lnHZr z`WB$jN^>pb&V{%>s+i|DjiuQ7zLK@X(k~C)8v!Gc(zKsUHFG>=n^V8=eQ~;d^DRU^LMa zbbjflZzgM$Vu-+@4;z$v&q4TH#4$&g1_1gYmIdpXgbAsI*o}7p{^z1wV5ki=D~qXk zx*K&3%Spz?@nR%CQILtq@5PtxAm3ZQ^OM3iIDTmnRb2Bibm zX|oZAJfId9MjTCtJT%E-hMN`l-Pxg)zqc7`{W}&FAr>H;LgDqws|Ge+g_RPkl~|Tr zJe`_NHO%CHg`zTeTBc1(PCma_y>LNk{KDFlX_GI9@(N)Rz{8(&sb01-G%^R73!oU( zCfb_5ne3~ocv315;pjVLI6x~QV}->-9fKPG2t^*sKoZ5hu$!Mx%I-v{Wj>rQirst} z4Wk>$367xU#}BWgN(1F5-ojsr;jW}Y=;?!kMa3%C_{r%54n@;VH#SR)f#(ECqfCIw zsm^23Y@;0FCsEzz6^2!G_|&Sbtr~94bgtdMB9&1Hg`;I(W!UDM*}&eEJv#PqB0*_k z2g#`~%*xF=8Xnomce;^8dAU}}=Mo)R1=LUk$nzgNoGL^Y!{E)EH?ex3fe;D#i_TNB zeFTCW#*|zt!wWVF#beR;V)ORx?X6f=bU@`CChKa0n+uEZO_Dp2Zo`eqMVRnow08@N z9A=Gsar9p68(A`|a9+1E=~;&>u4fuRKSUp`42 zZ4H^;I}qoW0vlrqCyzQ#RanTg+c607KfC`&YWZm4WHoLbC=x0_ED_@h#)R{2PD5ml zBAJma2w;98U1jpJ?^+SNkhPt2j-IoZttiAm0kX!GsS(qbUGJ77z{FioOdT_?N&#I9!{ z^farOx5hP!$EKh{7=DiCFJ<+1=eIMUMhEcEt*!1-6#W8VN!o|*S8?uspV0UGGI@AL z!DOWW?E&Xe5L(Bix04I#`#X#Ew0zCBjb!O<8kF}g<7(GH)_St+3$$k1-4Yy3OCBsV zRzEASuXq+&T=S!XbOzVuZVWiEUlCvM`tLgJ>r_M&>}V4JElIx&q#XJg*WB+jlzmLFxHyj3kog zAnfHqs^lt#Ek_oo;h&QE&mxWu9Ad4~{9a*M$^kANabEg;9$99WCizruq0eOVHGyTa zivaKOcfmR$8538%gNPMYG^634$7NAw1$y#Gc0RlL=%;e!A0Pa4eL!YU#z>rc@f)#h z{|f#)F$=?1sNX)eII&PZpQXqo=ll9*7o|?RmX_;f4By|J@0*@~rEhAp_nAS~n*JGu zlRVrx07?kp>JOxhGj$J8U1`~q;%DI(&7Mz7bjxDAU1&3%HQQh`*9@RraVw)*!!#`D zUE}oIBZaQ5T$up~tht^`T+E$BXWt6wW&6uVRHMy+Kq$1xy5Xs3Fmv}r%bc3v?0(&n zJtdE)6BQPp(~(cq`$%Q2S@H3fxm3*#k^h6Zh1{Pn;|1)Tr&gjk^Ca+lg_(n3B=d}# ztd1#p|1+Blp9>H^N6UH7U_!Bide!HpGn^r${vyx4;cJBwt)af2a$m*24~P^~F`!+6 zJjyau($Yk=40cox09i;p1yi;_0~9GX1LIW>>@EPJCl$@QJA-|?H`Dqf84!5}4yNln zkG=uUkH>F0N}#c%pgn%`vqnB2U!#g#-_#U>Bm~r*<$znme^j|*4h)zvJI^b85%meV z&$L=47^J364^7{9e*fa~P$3zZR>zRm{LMjMz+w^wl-!>o6eN@yssaYucI56YpusF5 z?W0ti2@eemhTJBkQoCveMhD`2;CS%Mh9L?2-#?+uM`>wkldb?}E(E#M^oE)OxCO!v zjd_(ZN=gv;ho)Tastt!iOjp?70aLEBMy5x-D$AxZ$p{P>XrMf%${!3|oG6$6uzR7Ie-6ProYx>? zZikZ^kWGv$^ysrMR`MzN`D_q3iG*5XdeCU|Ph3d<%}B9-#{nS17}&w;XW1 zh*h+wB|aDuMX8f|0J}Qf+C^hH^Ylp;jiZR!d2S%V8L)uWyGO@hFR9AbjP&;(0;ZWh zMjK{7B{>1fn{Yg_=Z=;J36PtH>_9BB%)NLr3f4tnwyb&VzlHZE5!Plo*Y;4%jb z`yf?R096wz?(=>X+UsCLW6_wLUMvea7w9J^dY+!qxdX`T647U(m8v`WZ0Xv<#MWYf zJljYbip(BjnThvEFo+SdfQzdlWt_bRx1PYV0LrM*r`=xmH=~%)IA*Ee*V?aSxSYP2D~E6!XYhm;UfJz$Gb zh#;aE`rn{^x1JmPT&#IkalFjZ-dL<^cxrcwuVG#%@tnPWAQ@{?Pg~K}CLdrbf!#9- zVvUuQSH*F=5u-(X5UP8uvxv#!*O<3)n;*YyQL%?ZF*~j__B?~1 z90jPVZ;Ufzd4_D;sgp`_!de7kQ?*5?aR_KlJ#RKUR^F7R-t9^KYQ|q_{uGtNX5`8= z<5_keHLN(XT2Qg@b^V$lhx(aIg7f!YpQD;IISkW1Su@lk?{vm2unZL7?&&ajl>p|N z<74h;=5bNpd3#kI*PZ!{UF~{OW}Z=NErvanP(_oLl8V27|5Q=BD>B!8o1dm?>Fb99 zfi6UX)0BNvfZnKQs!kaOx(JjAl<7!&fUZ_A_XaObf;VFnfac(~5!!b-0ZPX(RVpVW zrlzVhMjee?=YiyiQ#aGr)wM(cM_7qjhJcYd>Ru8-6#e$SdLwWQi;f5~dtSGq52)hd z>}C@Z1u{j>#_sW#Yc&@u6on@E{0Jhz&UapcFFp}3_fCjCMPjAa%M zz2ceUapJy-uNARcW%ru0g}~V2tO#^WFI4h1fa{&$Q1n3+cIF9Glw&0a!jQp5*9b(Q z%G&!X(sx*LWd~Bz*nHT94u$IF_)o_8XRi{sPKWrQxRqr9Y&O4&vr)yJ{f&ub)8qd z8Z4j<$$IHLeR&|V1`)AzJm?$<#Sv8b^gXssDTU0wh*sFN5xG$=^&~zSCv%iGQMprs zienNMUrt9yM<2>6%c}5TiRyh@p~68$69v16uV3u+9 z7$K?EXXY|SRVWzFPy@r?aYMrhI8I58yFgT6w$}Io-wxILLFcss@=rZSIzapNEwTrF z5*6-D1P)kHq9m-XrxyFjK&(~u8i5bdiVY)e2mW|p3r1x#$6Z6JWun^=Q z^_pk}tz*Y-1N#M=Q6LZe_?h6;u>#T+IR0mHMeq2_zlUh;lD|JDI((2{^^V-Z1Vh;& z)kN=pUQq#ZLi0nv>ZSTVL(Q^#HW41*l!an`5-4A~A#F|<%UjHJpugBEvOfmWK>#{D zMlEV?=O~PMYMOYwzib3mejaf35)e*MyFhKZT6QR93+T_H0p}`2bFrzV2Dq3^>dkuU zDC`RBYRH^1?v)F##r<~2Fe|N`2I4~d%mi62UhO%cXF@S6jA|`TdrJg!ZAMQ$Cp=lDzKUctU?D4vdov!843W^0-FIQ-fqPfnVv@v18iKGSE}Qo zOzHonDa|;7*VJb>>A03w=JV`|@>I)al?LB&liq;30bNB;z2v}>hM_ub)8=EhVSA(` zCGCtmKETh|4CFvDXJJe_U5%vy$SSXlwXF=Ce0LLK{Jb8rp!|0T%1CpD!d*&yWnDOA zjT?eWilE`pdUkVkEi6#VZ#W%AO}&YZ=vAH|!)3rp5+Z^Af&_A~G}NrR@u%KLrg)q_ z*cw24dYaT4pxzwIEv`;R9Aj`gBl9_9eNLt4-MY8gVk~vB?p!f;Z1fvZ?djGX!9{Z+ za?6d+i*P}If4$f0c5xN9wZWiOjm_l7xNIKJNixmAJtx2B*_7W{k5r(}u}+1ewNZDk zB{|p&ENwc*MU=I@_U>Jz*3oA{ZS(W-Iu#s)@zI~?I{%h{f>#d+=eyj& zg>*0I#3ioO^)ClR`1UY2Z4-_j#Q+I>-NobF=OwJav&fFzyO&hK58t231O$zeD2W1D z&H`YXg$5X-U5N06%Dfpa1NA-=K`|fz2`CugTN+0SE%x^OiBI?!v-ZUx8+)O1K$(Sr z47ihW2;GE!B8>dJSnPub4(%i%b5+;6K0VKg{qS$MtB? z37U9XHWYgG?y&Y4sp0pFJW#PYyu@^xt~&YQ@xPVx#?W8vby2{>ey7%9MCRJGq2E6=R81M7A z8`-;`&-S1Huz$IEb&N-0LbAAWPIC#{m0}LqdAl=q7(T=;lz$6O>Bd< zr?jzmU;oxXdCGP1;(b`cI|KF!-U<(Ip>0qg8+Szqau+kaDQ9k6u1Fy6jU<2YHxS@N z)Sese3`R%*FCI=eB>(59@^g< z@L1;~wDr{LTI}#Bpn%P4uap#|7w9MJ0i#tVl0V!IBfw}nRAomkodPq2RXXjS{7VS4 zBA5=M9?-7R%||jW>lgF{JE`tS8C97%Oxvn>Y!r^0-GR2$t_9FK?w}Wr47>pHBrt7K z$XGalCev~DB@eRH=(l-Jr-2ZW2hx}`Aiv&wm)4X6MS_GikRgpSbNvq`tm|W^hFc{N zf})(ZLeS^u&V1JjcqR$(gS~tAE^Ffucz&t@d4z$0+=CR>i_5Dr!SaL@2T+UIY}>O# zeyQiM%>p=nWXECB_uzQE07n(lu$vZ<2#F$1_1KE`48(u1C{LqIpkzpkV|5knnNrIQ z#UtRdW90Za6@D7U(g@B1mVcmlOh8T0V?5^7yiN4fV+Ug@NOB4&E&~Q`E`;jL+H?M6 z>11%a4;P_lx{b6==O_S@$%F2U31)n0|tb{2rOFopq58k3X=xvRL)2n z+)`On28sZ47=#Gy!Gs#75ZO4S^5E}drr+d4LAD3VP|y^Vu%QHZDV8L8?71PXGZSk& zK;0Dcs&L7w-cfz4sNk&u`+ky?lth&soI26aP9p}4iw2uj!jZ$JhBuC?@f~&M)(U_i z`fvy##do;R7hwUBEZQPLU2^PQWjg!9@@6xXUh~jvV3_YZ1ds(Nnm2)h zF;GIr=j2FnU6Dyi;kh8#S#ym-FdsSFJ(6LT0J2Jnryeynjsh{k5|$6^7RZuV0ujs5 zy?~ApLC!}(XTarBfpzeGbkT#48Xa+yQ2hTm22~6tHgAiGd++!gtd8;q;v)>?6#+_P z=N7DYTT1gve2SCTzoo)aoM$siaB|it&US5R_lJFomdLnJleV7Sfsh>foZO?h6?yz!+&VJ^*3Jz-}|xDw9i3xU1y2+V)HB-P;3|ycz`p(tXd$-22*JRe$yZ80L;` zb}alDpacmAK#*0-gNyFqt{_pMLDueU3^XSl#}z!IMY&n zlHpi#&cL>5p)w+w%waS8q)S>qe8e+?^j1S{ZEgHHM4{*Hn>T!4IFy#sc|wLcHlK7U6kunhI3eE~mb^KxrRY88%B&KpiR%@; zl)GldnvK!6fy`fXhWd)c2+x|%!0sfXH^&boZTeN==(ijend4Yw?=RqW1|nz}R5|to zr(`>TD!GfCIhyi{K~&CdG11Zr0@t2-(dHr@Jv}S%lwfRjhW>p^&^#-E=IRIor~3BC zsvC;#ay$nHoJw0B`LibyDt(x(J{~=h@6A?JeBGIrrS3{;c_-#-@BhVpc1S*$_g z4%-rJo6|&9YzTyFHjQm1)Rd<=hK#A9W9{+hLekPSjbA2h2wa;q5;1r!C$&7st}1Dl zp;io0KhI4);tjxqZk;=ag2T z>`=C1j)>IIAG&L*Q+y^m{dmoosHe3?(F;V#$`o}l<{+DeJr)p(GNC(} zh(x~$Xg#QRkSKvZJzEXEAX&qU{bqys4Hh_Ew}BnX4IQVw-ZYlU^$|QFR2G%tFmQ_d zgR`3rjmC???op1$s}3m0J^Szbb7gG^-bnd7f%C=^%IDgS56oFogEIOmnT1~;HsZe5fD}~5FAy9jp3q!HZJi^gO z&THxnVi1>eA!qQ{|Di(tCmpz1rm*uS6Lfen9r-qPnI4L0+ ze3;L?PdP5 zzEH4p^41Y*_bsPV2pr%4>nU`|SWc+DYzNU2<pS%2oVd?Jsu^z=f}Ek~kY!%fWGyM*#Z{)Brl%+YNI4%Z+eq*q-i&Xsws z{(|!G^XC3U;%EYWj?ym?nfoxj0fJv0n5x>#=#COE9RDTH489lj>TlfZKK>sZo<5Rl zIbYQje+-?so4I7~HQJ!IA!}|n%aeSF6F$j*CyW zxmH#P#&y%i&b~I@5XX5DxJ167CN@RLG(2@U|<_!tWhDYRDP8gaD3EyYtJ4mVXP}wpL)X z#EVDR12OExMW|%SgYgJ!h`h047KI43{eO-cL5PEe8HoZm+~i8U3@Bicb4EHF#g)JQ z^!=T@Tugo!m#u;#F$`qgEw1m5^#;HKxIE^PZ1;SjU16FHWrT|{hA4Rbk+l6Y+x(xU zM8vmBpiWVUP$P(a8F);Jyn3;R<6mvM|2d2JnTV0JZR&xp7!3-#oS-w*Gz_mo;wN$U zJn9xGMsQ8I=#%rr-evysf4ceq=i~Uyfq#YvtDy}9AU});l!-650hHC50P~PW`Oh$9 z*uRsOKbS-;`~ADE*ar8BAzLIo!B@$o$AE=Y!7B@SvHX310YJf+!ydfvFHjD5PMrW| zNBM%NX>lQwg<>An5OBf=&|lEzqBFWtS!8{o1(0{SI7N6_5cnD+VRfE%+?Fzgs0HF1 zg$O+6j^M+9ezOjMYO+&=HupLJgkO*xgJr&}-GItln%|Hp6+7sr)5%4ZSI_4dNQVZQ zkF;f59Qn%bYT86tUPi|8=c^2NEOQJi0?nr8!GY@;F!O!z{NnDBQb>Ye%Nnr^_?cX4 zu!{0I%$(2%XG|fc7th&%Z~187ugZphM|z;j5GM?dLCT+02x7Vbfdi!nk&nMK{$v63qLn9uUlgOu~MR@ zhf5;%0O3({%Aa2sKNfp)!ExbA$z}h|4az*6j7Dte!`)1B0*9lN>5^~OBWzy;dgz|Q z8^{VTw4I6Zox#pKEu3zn=LNz0-v73)od13mG}Cq(jSKir)m7>7EP0IslHx7I{u=mg z?$Z_m`2voM^GlXiT5Nv|Qb1l_7j#r;MT8#z>RN9NneFWS9DsE{wqL=JTqLPKVjTJJ z5c4&l9(THbbIl%_Mo_@xQU$h9YUto|BH7^yE|DOj+(?$z@-?%N+PN5(U)T z8(U;DaDlzRO~9>C96Jfwpn%}_km!nVy!#X;Il+C5${o??RLI47e+us+emSZLe8WK` z9p^!Qu9)bp$b@;J1BR|bX}!F>YK+~!$Hd#}y!|onnul>}p{T-RU6XP4{R?O{5MROX zauj6!P;lYpv~YkNt3am|Pk=OB^@JAQh^VDGk|Q#q?Xu!gq4$0528@z`#)7sFX{HAn zmT~9r6IW5qgBE5~SdjdLeP-RTQBh(kf7}D`T$S@*td<5&$Y(&!^sp_5$p9A*w>{Q% zI-bF?=Swe=c$V|1h+YR9(WzR7o+WESK`V{Shu^n0d1I+kA@29_I!}NMjA_0Zy4$ub zK2L8>r5fGhjNr)bP}FzeIv{mf6KK}J`RKdY?8(<=3f;IU{3fKOiD4sb#J46fVHW~o zyE>Ok0+EB47J~2$?I8)~VB>~ymjO#O5yN4|a7|{=M9_o1I{r7CseC65-%%&eKobdS zxS2p7;{`2HgQ}r|4Y#17t%d6>n;*7PssJ^dHecun4n-8nn+v?YX>s8{r_s^jIZ(KI zz;9$2IZLx((QrgI|A03X{|iW3Y4LW)t|9(0aQqKOH|5frPwEWM@@iXjD#R6jU1 zG~tREFy!4r8|46WCc`3R02$tAYsxZ!uknOf0`bHz_4k!m74Rg2g`o`F^3b(cHbaG50OvvChK7)RjsPxkd8uc-!{$YLX*4* zSaxp&g~(uI-Cfn|V92ao?Xq{-nA92I;z$(e63S}P1&V~*-aG*@!==SzBx*^{ff0qk zXpqB+W!P#xI0LOh+ej183@EA81WXWOqGsSAr@EebwRwx>TA0X+tFZcSV>`C`vFPCMo>MPYi4#e1 zRG-!@BILNc+oj294m~`P;3#UAH(?iSL=3D-(bKa9oNyxe(rjyDM}eSKwEz7&URnkvcdJ0Jlq#lA)W7w z{3Hds1+th=pwEFwLL~}rSd@jR znW^J*9)Rdy1D?WLI8D%VX#b?slTY?d$S%(h5)!q1>kHfOLhA&$yrd7!8RE)Pb{XcR ze17ywPtXwVAnB|r|M6EWa_SzcN;?`t6E7t3{Bq`nGu4Fl2LaKln7@%tM!GT)Shobl>BLBI^oTaiu|Oe+ z1T)J^(P$-D7BFp^9)e~Tq&Yi4oB@94gC+fZTH60LEopgu;l#2X36UAkeIr9p4Rzz9 zbeP+3&Kzytht(zEysZ41eAAIlkiPwIEXI;!lhDOt0ege_N?7qWFpb;ccsZq&u;0Wp;fFG;&GKQT>IB2eb@fUD33J= zqHXfaT`YR{3A0BQH{osP0SHm*86T6~r7utJaZ!XYT73_~=uHs5NVs0PExJ%bt`P9} zLBVV4tysZu{i(%+kj;?y0%=IXxvz2ym=p1SDeQB*Ia+f1^)jJv#w*Anspwp>YjIA_1|dCAVEeXprGZFmi){OgCkgP>(mNYe+I5S7w`;BvbRj9>9(1i{sz!oiX*xv0- zuT}^Ai-kNsHuBd{VCKasZFBLoR}jvTiur^kvKXvkxEkXpQU+9+Mcq#z{T0F%E`N=F zFL*7pfFc-P{}0E@AFH{)Q+)rUBIucJFW$fk*?=8jy1-g|V!(Rf8@=E0B#7u(%(TQS zMJ&Ojr3n(r9*gojPO88IrXB?V@6~tO@q&y9@&Ps!VAru9+#Fy5St#n(V^?S? zLXvS3q#9|Zrk0iyIr6{_=RAi;e5Pv)U=b&pzv(P4h6v9c=rfu;aEgTKXr2i`~}k+J=$47 zhYm?i3b#=a8`{tfg+@$le89g2!N}In1RfE89Gu2k8=Zl3LYXK)rtdBm$f*7UFX{yO zA-=e{xJrf+3x9Z)pkNNkA8b8uAqJxZD;^Xy8R*NMfc8tNzy%0(J?N{1+9C+p#J4Q#+3aE!dpa_5t)_piJD5oAz%&Md3^wz-i+TGlF zcx-AE+;ri-(R6*{IduVq#WGO1o<9HT ziDrx^>+}&Sc zYT!ExmoY`6d9JF|t00H%3FkfaHV7u?e*whR98m`+P%S3|!y{rwrNU2tiwjWQp1a=^aHtt6@{T z6$Efx07NI?6tRR_9L-4B=l~#U7to#6V5Kc-m0>siwsPGBT*@JZod|gCF$mgd^@JXA zXciCOy90d{xW_Q2f!wAJZhQ;{aR4^4H5Be9s2gm-?h5BT7B=Q$r#=L8uy@}=>#rdD zyM1`RW;UwAC#!JUkEuJ=F>dpV>hlt{szkJqDXQqYuuwgW?X=#FEMouN{Ji<^JM zo^}L95SLO6vUj2j6^#La{$rWq-%`_x*cg6vA3_t__vlzqBON9kM%bLpLYe~L>j^YC zgfsZTG@1xu@O3e|NOq8{!E7AYTM?SuLWLkE3!dI^29wiXkb7%8b>EHd8oNA!;}U-p z-T&>X`u=Z>@0=FCzm)&ijGx}Fdic|-1Nr5|!r#{%)uw#Y7Pe>SrCZzgytuwy{BT+R zo(Jp9Lw>p%oZvzyKN)`9=U|b>aqDF*ITgu}mKeJ+*7$OWHN#*S+5-k_TCkUj-~$>I zyw1(T!^3;|*OZ=Tbv!-yGHD?%H6fw&#LvxoA;jPQCa#v!KBH z?Lza({XKYw}So#0@7$)cj7OKqUVsFN&pb(`)(;F$V=bMX_6H9h45w5Au=AJ?fm>5@M>!D0+KTB_ z1j32{{C@;2v46?87L%9K>E)ncjT>IQdbP>TXx ze_Tx~^lX?bf{f73vNAV+>yWEP8KsptMlpIlFiFoZTf%+JG&0MbGcYz5N^Bi*i;Is> zDl9Y_qm}6S_{`rM0@&xkFRq{QYRDNBWXokyU??9yk?Bm=fU8t{)2R;UF z+*fC7Y-AMK=m+WR)U+AcWCUsoR^-FoG2)*cB7A(>{T>&H*yeceI}#c^+j@EHzB z+thTd?W{fA&vwa@kq_?-Rl+e>*KCMU@=Xv4hSTRxhQ|vi`RyW+v)o33Y0=TwPZZXS zUY#r${goqWsb%36e}7J0k>BgR>aH#>kK*DEK*;o(?7In$>cVPeH{Q*ghq9}D&7>Gs zd6sJv;^G!2s8H`DJ$kh9Y;`CI53oykzXs^@%+d2(=9L$c*QX%pHI)o|ceg!C&C6>w zD@cf2&HmNODq+hSew7-V z*GQ_oZ`-9ofxIOdYb!(xKy~_=>3@NSbN*OzK|#AL+qk9wLoqd3{OMPuftp{Mfhr+Q zJXK67*}i>yEtHC>G-$%R-_v6*BqU@F`5;`;H-BPc!Uc#iQ>xXYC7CSR&%1Wnorww| z-{8J^^EV(qK6@nTI;brvC1nG78)jfm*!OJAITqZ8OKVc{o;gC<3)RbaLk(vEfE&CY zO*1ZV8GN^6*@_iwlO)s|=rwMf5Xi(T#Z*YMah^dF9uz-meljExywec}m{0a$}(6!omVMtE`Q7*0%yw zyXmPZj}#^#3V6hLK_x$AkeWQni@`zsJMKD&h>EK3-o2X=>&u$)OqKJQeGK5C=j^_N z#1oxG-nqoY++4lVKtaZd1{y8B$eVGP0#}>H#>Nt37%zF@UKFviygR?;KF)vi=ulKt z)Ygf41@D)?=f?HXLm4|2KAufP6Q!}?B8&{5&x_8sEDOaHZs#(~m` ziY6=uM2X(s-T^>&bL@ozB7M2*>({TFauv^g`0d5Zm#MZ37K@AXD|d0SVGlxM)%IPx zj=-fRb#TrtGF#0vTi5gQCQvAUYqgV#l5^LHkh0(O#BV*2?J!V^%T)id1FEVOKd-*Y zCvkA)%9RK&f7-egMx6fJZ@+yvRC+cpHg>zPuqs7@aP#Wb+ov{uyJk%j&@!&j57PpW8VW!t z(R3^=;~{2!ziHE4$&}esa6P&Vx9nE%nLP?0Dk-VM$H&)}VR8MZpMEkAk6*K8-@PDV z&2ooaAYl%Jm2}2h(4^4g{)Z2z!IqSqm!}UGu{3}^DmW_+gAx+}D5I!}HF4{h*Bdnfp>c|I?ya2Pu$n kV6o*H{^I}dH-5w+_^j6xc=Pov*VIlrpz&+kFDEble}!x&TL1t6 diff --git a/docs/articles/parallel_files/figure-html/plot_perf-1.png b/docs/articles/parallel_files/figure-html/plot_perf-1.png index 4d9633fe8c0562f9e6fac092969bd3738b862ac6..baf411f2e221e546c6b5f0ff0e9707e9d79625ca 100644 GIT binary patch literal 29919 zcmeFa2UL`2+bujs#l%JuTNIHb4mJc71f=V09S~5ebPY|q(uNMvS52aT4g%6`fOG@~ zgrV6$I)e17(%T@tbM`$Z=Y9Wm%J==>`M&j^f324l7X)YKd7ism*WUZu*X{3Tl~1i( zy=^sxLRm*UeO#47S?NrnEWh|Q7k;DvRgD3@as8rr>NvjRYvvfs3*WxEd|KO{LgBqh z{w(wB`fNg>{79i4KdSB=HrVQNKiuS^+~Qmru5Tvht!=$=&VhsPT?9>3%L z;L&dzdsR$gH$Ls!=gwUD%x2}LJFEFGtfVemq4w^Q@i!}SW7_%}Zy zF1Bsid3t~XzrMcygzvJoN1{~su3Kk2e<6 zxRR2RxrOS^7twE*%D(Cw9BU&@mizpBI+JNq{e+e);ad7M|whL zFRrfiklJ#=BRxIcxFO{tjq&EzmYCe!+=|~R7dkH4*i`)X-A)FTmzS5^K=_7b%g(gWdH zjTfwkEgn;?G{y!MnJZC7f`WpQ9p1|sR2A8Y_y2fVtx}y(cJHHzBHTXTe8(y`*;7oN za?wf|e`@2EB3>;?)0XE~Tt;538XK=sF>Cjj(oY_18+><)ue>Z+RLjLGJ~5+YagN;{ zFV9_YqpGE)W$bxuY^;hz(hACK^s{GIt{2I1FQ4d${+!a+ueFjrst~0bS)uFn>Z+K> zhxCL3kBM5X`nep5P;qOWT}mP8B}Q{YnTG2nJMS>(zoA%Ij&!`(D&u@P*N9T!{l29o z=OwqnGVwsWEIszu1s2nvOG6zT9H_LDzT2~!4j#L?UWu_^@A>82%H?DbdgDD=S2L62 zYAqO>+PRogN=i#pqZ1Q364F>kX*J5yiK7jgy4C@h=wwGv?{$f|akE+*Ev=x)w&evU z)8|<$f`s%#Dk>_Zc7Mh7NX&8T5s#VnnO&u;Z?9alIl-xn!LP`m>e^&=xkgTz&JCui zOg5TDYU~gQDzyybd!UIu?kZOiP}-HAnOV;X4&~+L0vmJ4fK)X#>5xZgc=SOj(E4#z zV`FeqQj&&mu&9|b!}unz(|Pt7c?Qa?!i%|Gc%MtDZX3gRLgJX)_`6g7RHGta{eQfI zwfYpVqdLQQw6j=%7t6Br8n@j27uQyq^@P*ge(d2y!13!>>6_Nej#nKL6YI*G8!p9P z1>0qXSujHy)69D8Bj@xOotR_8z8e-EvVX23g<9+<_P0#CvV%ob8Ybwyz1r0-MBwf8 zmMb)dKyr@ATu!A&2&I%gQK~dLozBTZh}I8|Ui)jyJfLsPIU<`gW8~*|0*fD^ zhEl3FviUS#_44|ffq#Y`v{R=}H8nR6a6Y)ZyZiPJ2Y;qj*VYD{xV5=f^1B1yGBYyj zq`ykpHu+zOss1mUg#S1={MUp0f4n)LB4IRCb(e<*HQVOPa?6gN-7Khc@1Rj(=g2GW z%IVLaYN}Eut!+YPv-1OY9%ivCOoH?SqobKicUde}eV=Q5$Usz(fzGFy z0VxTRxeJ=cL$7A)3H5xuC9I77_qm8K5>V_QfLO3J-6n66nQQ_en>E>FIo7A*{fbvI92lV>kV2`b=B33~xrZO?J4cK?) zRz(5UOZO!bw+5UKs z>_UP?SMj+YAfV`^&o*YsLHnKxB}RHiMz{RZn7pWUQ+m$)XlbKqsI^_kudBN68~GOF z=ZwOE`O%6Av=n6b zZdu9hcyZ02x@PnK^F#Of6$7cXj2636gGzaTDrzy7W`5z7q8)BlzOs`I0TJzFEHR1E z*ex|NDwXeJ)y6(>dpQT#Z@=>TyRFhrgZuof^Jcnbo6k|K&H%~D@>xntOD~M?pol*z zEG&$$-dfUSCnYV76&TTUl^vMNoBIZy=c1QK6ZcgQMp6!dVf3b2)!f3)CA@1LF%2pg-3Ts)*@;9}w{H`^O_rEsfd?qZJ1m|dN=8Mv!; zK#{JF&S>(|Z1PrF*YWJdyP6lc4IUW$rCZqM7(M)-kz<>jweL@WfmC{S6hucEwisIn z>(g7sHtbd|!^RjzD2_=?6vQN$`AYRa9b~cd1n#j)9q9DkG|gIjdBhzW153~-JSl&s zJJjss>)*53d2=Gz<7A1fTeq$eQP-~O=&G)iSL?@Kd`q6a;lga~>bAv-)wsWej+NfC z&#rE1*a!xIRhPDkqXQKjdHnwq5Ti7)o;*|NRo3QolZ#Ods-(-1R3c5!tAOJ~R>r}eg3|NSS^ zlYC7KHv5p&F`fj0CO8S=(jyhW>SON=H7gqY)QLBlRMykKKC(Y z+>HRD2e#QPwb`0i7RQiX60aO)*5v7JZPc@PTXiyiSFE3}0)Mzy_PJJu0zJHjvUV-! zEh;+0d7|a{p?9g;qrZz*=MGIeDj#Ruix1d;-r_MA<%d1<=l?C)Evp!^*AaQUmO9HTYB6#u$UlTYcZ zx#DgVHmux6InVcRKgnN+)gq2g@t$}$A6itM;(U(r#C~Zxqp%Rd1qPzQ8KmH$nEd4F zZ9O>*YF`suSj`8^#~@DAKPkyI28k~_z3m%3q8jHmHF|PFDsELK)7L^o+2HnX3Z)X- z%PAKwSenybHM99Y@AuO09FWF96eO{-?MFNJ6RAb*u(=G-C2D? z%10g7rOfZoW!nWQDa5AjRH|HRYYtd;-c0`!w~`B{hZ-Q58E3x@Gf#W5WNg-^e!p>#r8l~Zj$i1>&b>D=H&A_m7 zGhJn7AwO^B! zOxdWnx7blS2YYfgTfcJgCsAn@gPNS{Q7;v<%-hlE-z~lqy_Z6ppo>;J7R%RRTWDx$ zgahuV?r(_Z-u=~4#g8&D0^-Y-a-Yc3Ml)Vq>knIjXa4!W z2g3hB@1KN%{^HKVz1B>}Ts5q|=FPjtdI<2sRIXax_SN11g{_iy#<@l*4;>tz%&Bnm zw2AL3-yNx0e|W84D|bt1)Y8l~YH`)WV`@5Yih^(4SaN$+1fVVAJksG4X#2q8(xs7u zeBM-BcfOP|FmZR!rw_aGyFW|4`eR>!LhepV{KF2x`w+62+>d^k*H8T>PJLheLT1(= zPlQ6$N27NqKRR;#=<><_vXPMy17(ES`INiP;TKPoE~R|akiN%x4Sp95F}dwv!;x*9 zg_SLHj};Kr^GVJhexl$KTg9!;GoAof4Gj$m_{8p-S)9K<)Af;`=_~nTPE&0_jpvkA zuF*!#V~m|QZ3ACPjNRS5C+{!1uPPLjcIn+Csonkgfrbxz-GtAdvy4xUzo3RX%7tsw zwG7{|vC%=-qF#y3PFi*~IV&f8&t1waGrYgssg=L^(2N~;Qg8mYI=K~;y5Skbl2hKB z_9tqN+0|E*g-Ut4S5b%3(4pM!*&LSIDn8-+g5p@bMz|}+zK64$3Qj&G7ix|*xx8rc zdc?Kpf_GDf%UO&Xg<^H+e~5my?ze6L$q*(Zzcf=hHvP%3CqT=f`0C0v1X}E-<$=-k zEa#g(a$_Bj=D;3k&Vy$rn35C9FDM%N=}+6f!{>U-#hJgDd%Z!WlQ4x_!g!`ic=f%{|WWX<_Ro33Po z(A|ZAAKx=|jcZ-Wud8<4cVp+$uV2B01(UR!sOh7^hbvwHlg zO~yr=+v1MH^Jl`Py{?rJQkP($w7Lyy^KCwPIhtMX!<*f4@ku534;ochRowwwh|&|9 zx*ZOISSYKLp=VdjFFEmEP4wc+zgqz|Gk;qddawV^!@k27e+!9-xT_vF;j4DuM>zM6 zs@W-qabI&Gm0$PPl!nRNIqliH%s)548nM03wydEHsOAUQUQ2VLCMEX{9ZJZ%Y>t8l zVOKrOdV^B2dOd$0o9927MS3_1{_u$sYwU0$ib!If%l-hcZg%0LUwGB8_5UNInP;4vs^RkrhhTe-D_7eNW znY`b+#!A?Rf*Q}$;?3*6tOLHqk(>1{Cx*A=ja-jF;S*oySs@#9PVI0Z*xMf>y=i8v zQz@z|LD<+Dz8*l;qxD`3^;z8^merumKo9$v^YhG zpFb5<$o$#bWLMvI6lr9;hEXaD?yHIE0V+)i*T2Z*mOny##7IJsXm6^%;Pvx>bBnMMoRioxHu;+;U z`-L7>RO*5*gIcRvp0No?jPl&Rhqs@18O1fxbfvAEo0~GqTwt?VJy{Z4a_aTz`gsdO zwlUA19n{jha-u?JtPIr!ofBWpKCGbgsskS2%Tdk0%LYX(NClTwvYK03!Z>xWE2sE8 z92ienENQZOE2P;uI`UmZ?v=?&)%ekpE-@_k&bCl~*i=0(wbV5W*`o~IN$%aDfZ|oyLZ@1!T4!Gfr ze=MSnIc|f`u#+bT8wcDgIHjj%8Cp{_+B3n!UaAQ!pyEJ6t;N zwqRNI_4g|h(=Sow09o>FJv3iLBJh8x1&g==~uMcRG@< znzkvY6eXulB8u&e31c8>pvq^~R~=<;Zcb{-iNc%;SCwsTPKUlhta?ex#aH6p{G73z zP4rP6$zU?0pnlX^I3^lluQ&VL;jFR;(-xW|LWq`;b1glit!z&T#lZdAJ!5rID>$_O z!AZ{g1%vg=nf`c-d7o3*^?v`fW1)hcSEDxUIsTZh{P*1v_o42|fzL9Zb9O5bDwY%} zcr)AwblXd`&B8TqDUt!8taTRCR-J?BxQnJNzc8+IhI8jfEYwHV^U2B32oCwD-|et_ z`xA|UV(~xyE&>~!#t5~|%i3H6kWHc7*|sT%>=UvY!=To=On*b$P+H;$UqPs*y|S;2{v0E(V4gQb zYjm&@33bJOICKBNPZ$E1?uu|-_F?kE5wAcYrB6fcDgfu&IeS_+$my|<-+6LlrEQdI zF50~8Pu_!N}{pVp!o9F zj%!;Orq};-xQPz6QDC@Z&y-En8WQK`3uo!f*D9{-NK6> zh1la;m%glRq3zNjo@UVET{&fB6DPfzTpSo9_m@$;*M2iSSNN5xpJ_?Q!L(R^Ms zH+Sy4o>%#9!7EckpxlMOygfIAZ7(JkHOD&_mD=1qC%0;zrj$2;3HX$Fmnh;g@~-2( z)M7OiS!T85r$Qe^MhoP+)dt>QHPJ6A=83t^$>ccGavyJOQX1P#rQMJa@9g~$S}Un^ zMT{zGH*MM&4f;hnQa$q0-rRj;ah=3PGcyHYxqJh)Llv%)0rKsFE^1y}xf=d#*Fs<3 z&>}uB8Il+fe%v@hDTA6n{YF?*$_zpwjR7l!_fjLJ;gSMPPQ=cc{2`Z4y-FX5j@CP9 z!2&}1x|!f2rZtB9A5LjU|&rOH&~q!6|%DQ@^vN!k8OE;*ubjVtDmRa*8!D*tjG z9o$#AQ`6q15$>fs)U3tX0egt-liPJC=YhO~%O%=n*q&3YG^lkw@{6D389C?XIq1$q z3p{|J>=ygV^5*u%NmAV)V)c6Y`4K(j9PQ_qKPdE+g@_3d=nUp00$r48@QJ;&$QHd+ z^e3sVHHO*dCoU({i1m=a6h8t`H6TQy7ugiHeTUY|S*99nH4tM?T+WJI}u` za#Oxcx_zc%Q$#y37g0m0G)Si+9<$^74RQ|C>otKB&)`D3K-5PKm*5rBoG#UE-g${V(KVm=~bY<#$3Pj0qa=?-D zL^YN~02fprMz0II|J2&rDv2759S~o_CZjC4;!`WRK z8Hw6Ugk|k)p5DM*wpEFR&WMoOz5q>E89p_rBxUl#s&RXpB#O@}AdgpOIn%qm$cf`WV2r ze8Qv@W_>AwI@3@jSEu$qDc&+EYZq-4$WfS3DKSgU2#?7|gYtK}oEsDWv)uZ=EeqzKEzK9nz}py^?eqzQq(JbgyJ%!iZ#_gM2tJ&Vl5;xqx1RG!byG z+uOt}e2f&MRCgddBF)Tq0lcg(s*s&(p%!-q8RnHFZXlxuCm8X-EGl zy(kz>;Fz~#edD{al)1ZxPw?mc!81cLv%QaL4Eebg^xtr zePOSeQrRLYDJeT#Wj9S3&NG8@Y5H(W3pQq>4Qpr=r7hWF9s$!ztfBcD`2{tA%xpcl z!Uh67G{oNzq{hYpI!`2f$*$QVMBqRk(7G8`^q^xX2M6)IX1tkDPO;#b4j(Mttn2&3 z!C!KFW(n;m_TSw@2sA0vq9#SC~Id>6{6d5?i&y`GTMj2O}ZlUQofL&{kD{ zKIWxM7Q_BngZo%HSuhSZ-qt^^ml-J_zRu4d-?+fJmDcV#$4?^xo?6_#(Au7OwmRvS ziNLS&F#IKr%!BNwBBFs#8Wa-`*HliAtZb31n-5jiWL$)2)Uk16?UN3V#hD&|D%-tB z+VY?*W`&Z#>QZ=l_SE1sazaNEHFizTNVO(>Y^Ke#=h<$e%wRpYs;2n5SHu(mXc2SD~VTLxXZtE&- z0^=1kFlziX~V*^bG9u$+e;*Lm8N`KHT7O-2r1ujGj>Ph-EO%f0qer*RuJx zpvs8zK~$F^W__Kr&(cEsk`g1pW_9L6Opy{+s*$fG!UgKowZFKM>vt%T4M8m?Sc#bR zu2_NZlGu)YxJk22$QHP+1SuR9xiJ*^y8jSJIz3vrBRlt8th#H32kZbS)M?FX#Hxz` zwNZ=H&1*CdFC{PbCo`z1x)*FqCeRqAKsf%8XQ~h?qIa!ivF(>m4GUXZ~Z?Ldd3IQ;cTmc!FY0h70rWOPI2Y}4) zra_FVir4d#89hykk-qetop9UfblnH~tMr^qqcKnwcE4^^nsMzp$Oe4Gc_m{A8m(#a zBj26c{_N$mbqm>67RZafJ7lH&cPW){6p`njd}q2P@`F*^TPh&}R{MOngylTPS1?{K znAyU>)b1xMG8L<}>%&bxWmtt65*a6*@wj6Txjps-8G5?t=<3=-u}Qp==jSnKL}P3f zHKn?IiqO@y&2>x`(Z?R9)#Z)S8^6-wJTtOKRW>Ttyj5z5h$xUz2Q2(-;n$wY^qR{IF?CL}%gU+q zd9dU)IV@XNlHJC$i)6&CTCW9W``Qz%UJ7bhrWwQ$}Uab%**mgm(WyDp# zK3hn8itaI7+48bQls%!}E_{UleSLk*DZd@GK=*kQ6U!aXCNrEDqo~DB`U9iWlgp*D zcqnzNm`r9LT(AsdxWGxVeK&oD)$MKDwpAzUW>(ldw!&)AM!CN^GCtBvN`H83IJM7i zdgT3D*+*$QrH}XwBdSAfskD%MM{@rZnJjcxwD&sU(5{{E=d(L!Uvjk z#~yu8@b+GT`$@++ z881OsP-Ss%qF3m9^B&>cDdm{(&~<7jzG{PQkDrG4;-~S*owdVb{2dJ1yNTpURD+Bj ziDs)b!o|q)wg(L9fh$L4YPeNPmtCu6Snj=SdC7XZz&nmUFmR-{J*aHIl9xnOAKU2{ z(bC={!z%>T?ykIe6B*vP_;wvPM_i?v(}&t|_e4abF}4`GUtuy47DR!xtst6&*M0f4 zlja?hd##(&(wfgJ4-8*Z?f*rRtIptr=|}`@2ZS4n+O#~!^49#gu@vPP7errwDhVSj zBe?W~I2#+FHg&4aEjbmtY)n&o;8ib4#NLjN-zVyZRLaV0*qbK1+>MTY-{vsHaMhhG z{kTC{f6UXug0Uks!9_Ub_8%*9utzEp7w=Yj4BNZFC?R^eqd*}ny1ZJ6%K}l-(6D{@ ze$>MOq{Q4NS5ecl)2YVGJ{vC0)^C;fbSL;y&SRE{PiMf$#2bxgy#78Acaw;!E(qpL zvYKjY{xts*!=2rja93^ChF!&k8WDtK;xmNhMJ!(dtsr16ck2z@5)+;gIhZ(ZQZ1J{ zR>_|RywNuDBff~No(L14a1j#)8<>|;=MKYMP8e2@*d#&B1W%?GIUP>lR&L)G&eC2 zB<%iSjmgwOyUw>cm{Sv<1zaEkw;CYkv^YPL)gKpMGW_X@zp~E7^4b3Wey5-@LTX4r z#LWFY-_$d zH;hXJ!}A|CC|4mhkYM%f&1T#`hWI`NwUmTst=W#H@(VTc0j;O_*(~C(Y2?va&ntBZ z!|a7wK+V8OVzogq8G^q0WJ7HuX}flG9uz}pR(ImVZL36hEzYDC0v%_}wDibnISu)? z(23Nx-J$9z1|D39NT>H&i&Cdrx#N-+=T-@`9QqhQOI@%@?PX1pd7M01nLkr*3kTeM zRKhddut~+?Pq$O&df$4`5t&P|R5a@veB0JbIGO|^`qx4D=I7fE=| zPs^Y(@nsYsP8Q?gwd)#4XTkKy2NNG_bMs;>$XBUz)8jGn;qVUTVrY?NNGij}z~|Ff z8?U8p#D{Q1a-4vr)hRYhexXlA6m_7~LZTQFj#`q`ilIeD3&xU)Pk6?g%Z>mM0|=_L zYrnNRK3gg(qH7`7yj~>qVN`T!Q|(p9SoLMNNTdQfAp~pd>H-<}h(=j$o!P3n4coPk z^wyX|;}-Bp_#KZEzb=YYMN$XIB$?C;1}H=wTQB)!u!*R` zQtB*Nz6P6?$hA@j79pPXRxv$4zu8in-eh>-t(xj8;j~LjKa!k`QYcen@@JyJkCg%WFFgeYf+Z!_&#?$e|LK+uJ2m z2eXOVbu(!K4B@QP5WsBCV0NzC)Ns`VnFE#+;_OWe-7hq!Q~7# z)f;+wG&&+^N;}A&y#2#Dry(Mn@Wv6pXOKQCterp56}KbhkXYo$XXrG9+E3as<`%?7 zr=he8srbQc(zDa(2=Ze7LZ1N7%mIoZz|Mq{6pgDjS^Z>pvIbjLN`I9am7ExmfA+xN0;mKGOElXPJS zK$DUnYZ5hy>W58^j*hqj_CQGzsYg5X>d0gNF)H4Tjb1bB(Zr~E1i<{h&2l18T1ZI-?4k52emRN_0Ta~#W z3`a+hq~q`e15TMbu&Wq@8CjzukAOik{Y{G{=oMy-wyQb0uYR%jlzh`e zzl8C@#x!iJ`Oe*?2oXy+X?6g=ozNh8Xq>fsJH>t-`q2Ek$jFcm5 zv|!n)d-|;G?CfLppGZ?30pm!UPkM}t?SAO;+)qdL6k<7L+O%ScWTLW>(F7c5U0O-1 z;iv@a*bE|`;u2&c=5`wFJb8Q1+~Xo5+2akwJRP*~tc9TY!6uCQv3et-tv+xvTh^T8%k&bw(r zA8KMAn`#a4*f@7Bo6c@~Xz4+!TT z46~rnAq+?lrK>eErCr+*4o&Liz{b)Ly8^{B9#6Frm&MOht-b~WzAbBr$rtgfE>X9U zDL)abVgj?ac8{Y!|5U48&~+lvV8_7#pYYaWif{1)9|jvS0f3|XNk)b^lPSL9+Qc~x zbEW9Aq{3>YlDgx^jvc#cSVJnl(4J_cnCqXhwFGXvK&eKioTr*axR?@~tJttGjbv9n zu)#)bgE1`sxnlzlv$9+QmWdpR`Uh3;6PECf{`AjGp=+f##j3%=(z3a3+92h^(dI=y z3STr?%u`P1`Os8n0*PlJ-L&@Q%Glg_2%Xg#%~3PM2d76YI0=q-3&uLmK0^e@Xu{npyRI?uNGM=36gW#_%|^+7y=A(2Q^*fe z4wtflFyMcn5`Y2<165eWvHv}(RiY}@r)a1%wCfVF=OQDDz|gJdInxz<(CB1y{Q!>gF5jWXsP2Mp9kCW9$hHb@+7@G(^v_R->;ye@SPm>bts* zl~T9FoaJTkL&vp*7i$u@6JE@HNZ}H;eDt&Bh{2UNn@D3R!x#`j^5bti_gx9uL|_o+ zMOOqMh~OU5j7ciFYf)!KvZS+KSfF&w*e_A@92=9oa`f#^KVxX)E|?2def{uXm#;Df zJccK$K_Ndjk@khKNo#0m#3m)ZT~&WER}G7R#;N|n&4652f%b-);bs+} z^~#7#n8IhwC7dF60RMN9bHk#}k&wTw5Qeip{vj}0ZtE#Zq9#$9iIpTaIoYtw(hn_=v0Ng`i1DktvLz5|54V#sk*K!epIKb?A~P?oCmYJpx*s~Oy>5z12?fs5EPS- zYZC>dccZIe(%V(Tvj^Ln#P} zomOuKXYGP%>MeB7QDi`4lqCD!Jw8PSouFNFE2}cdkZ1eE$#p2=hc~TrUP8G5j$ERW z7!J=6&kM0AAwh?(5qkb3Iys67zp~C~jAjgR-eq-m-{uh#j%Hr(mU}nZgE)MMv$l&JiE!OVycSV9q>y!Cr@F!@pKX4Sd_{O>q1fgC1?!~+ zyPB7`ohO)LNwv)EeKFQ`(l>3gi?`K#NA3H5L>&FKy^sBr8D!ZIjFQo05x=QW4E+em zQKd_>hD!qI6R;wh4w027CME(dk+OofPnD5|8It>Jo`z|Q#1_)O~)5y&oW?@%Onq}tdkRGY#Sr60{v?QINv{@Tz z1s#-w87?+HzLZ#est@9524LSad7pi3jgiX~k>pbeO+-j5h3re@d9m4$Iwlq}kw_Sw ziJ2vaIj(3rp)p8C|6Y*9-U6<8CQ}RzCcLiu4fJN1zi$!JyN_KJF#L|W4_S|+Z)JF% zLRV5#{eE7S6p&;<{e|I{EJ;KLQp6;MPNEK^UU+f!46JTM?j>dw_s*Ss#o$Y=ACfKg z<4+g30VM3E3FS?HyjP;PAB|7R$_zqZ11@n~8RF-+=sTQKh^n^pGr3k_UK5>onKY{{ zopsf%IWPmgt^{I_nxSDNHEXm)ptNFJ{YfNiF76#+M7y(X&jY1!BXJpBQ>LU8@ySPG zG@R|eo16+jyl2oXHF9D!VsR(PK~YIbLa1$?B9xI75d?UNpQ==dwV!ClAyke!^?M+h zp9_yu2S;PYR_e{sAvMw%7=6;F;eyE_4cmW2P7yK000ENt_>%KxE#$!d%xaV*ln@mm zyHw|UEpa>o=#TNI?hd>BH|?|=0#QFNWA6&JVi4VcW=GCV*vU6@n)p@bi5Nk^4tm(W z+$3oDcz>@(O7Ai=^>ND9C@?oshm z-J!PK9ES#}Z1uHybn%RayY|O!J#^_T*{HzBsgFDo6@nBt?SNmG2sp$|6D;fIs{D6d3}R%s|XFjCCSI>&<;1b;?Ps5>!{yhO+P>A}Vq-#(8tC>~2!xBjQ+w5sV(C z%F;tPG$T4)Rsy7htht$D6;B3v!_p_Oc|R(#JVC)|Dhy!B*)U|lPt4J3k7w`)79rSx zvywSO3*OwSuBBz%X5snymLVK#Ko8>s^|k><@o{lPBiDH1=lirQVA_iuCIEIQodv#? zhElG>A9}aPE1UUd7UNE z=4@n&GADE+QbJxD6<%UMDeEDc)n*~RhRG0O#teoMmIpNhLn%payd}%t9;$r|L}^lS zZ!RRoPBT)!rXnGzkiN&Ix%Qyi=1p39cd0WFLBgnwI6|O?MQ4{q$V#g~5zWqCAuvsN zDr{8{4*m!?(vNRg+ZL7WG!z0}B1^1t_#oJYLQxcV7^o8j2TvUDBXObOZMnYb=03le zoB%!*7t2kLC3Z-(Ts}r-nh4;A^B=b~N3ii6L?s|yd8j-qf~Ebs9Xbe&0avATF(XRT zZM^p`mDoUuQ-k~xkd)pA-bqv?TD5ph#!gWiv$FelZ}NE__L!S22e=9}F1BhURc%fQ zgIo&?ZzmK+z2F`jG~$aty`XSDerpb)3O3I_+(?Z~j*BZH+sPo;^&3vJ=k(b# zmIf`2g6Z*TyT=TGn`nUv{LK-d#R9ZZWUma_`WTf+s-o$r=jE*H=%R&Qjp?$qfKW|i zAk$vRQkfSX8pE28Ylt>YGpkfirFIJa!3=pZSDWlfVkQ~(oR&8Rv-81b5m|NXNHQ&G zYDijSZ&*hkZWEMtVAq-F%81x&nhF|-Jbs?NK}$d%m1d5?r5JZxJ-oI#YM@6sT#7lA zMGBVJR&iH=t3(iQ00|V^cz8r{6v{Ze-79>YYzBTBF^+acfS3=}fbX1vy^}Z_NO`Es zH@SE;8LO^zAq(g&Sjw^Au2eOFUAM2J6_i87y!r{p7V*=~?F?W^Y=IcWej55|6XHpB zJ`$rN9xgS(m+>|;voWs5wQ-=>Y!u8^dT3setmTENRT9VM9z><-_#49xN+K+$*$z=L z3p~gU)5Nd+eL))}Sb-JEya#c;9K>o`ps^p{+K8ZvmRxviv<_FMUL1JuOi_X-g-=qpPJUNEa#1&)e;e7BR2mVYAzHq#t&LI)60+X%G&@@gVU?Ft1EmDi zMwPbrc)$!8t-ddD{0b_T`=^=~7v_49Un-!ZlZS4BDnJf~Ayi^J;?fM_b#|dgD}2M& zORY%CFer5GQ1pSuY`r`)DlR^rj4ZVn>y=e9mNQYuaJ)~)*mp(9Ly_g2Od=THeoH8O z4>(vBa&GB1!*47%jVz*`XOEI`nqw`;LoxExrAwDiBd;?E>a$dTj9XvMwe~GBT@%HB zX3A@6imLt7w?bMV)Jd?@Qi~Den%0HFFG@~t;inb)@#X)`N5%c)23nw@#B)y5H^Uf} zU7g_>(rHa2wgWR*M4%{M{_v8-Mflnc;6NME}HHEZ#3fyjo zR8kigLp&|SdQLo=B!)(LED>8*h1}dA@i+k9>J3kmY8#0d2o;J%Z+95T_~=?n`gpzO zi@c~bIZ$Llx)n;T<&!Y_|b>g8l z4Y=p$ec=ExUqwk+U{A(-^iUprkMdLvz#X+1OCV4;{u|2E1f+qC@EmAkce3oIy;@IR zK)ZZZmCf0VJfvug(BJM;?*i`*(CXyy%L)3dtr8hg>E10$ z+Fn;rkm!q>&JeUBI-Lj*9nlDocgUweX9sa9eDd%i$MFJ-*tzbd07u(I(~+dIsl2J& z44*ng4)o=zY;LW%VShZ{Lg+@b>Uu0MT*`r@dBUShg7U_rHW}E1kz`(H z6*GQ+KsH5Zu|MYH4n$@u%wi*}lOfHN0H;P!x5~Ls@0%lXSqLetlm3BW;2e_Et&0e0 zCm?|I+Y*a)fXzK8SbcklNuO0DH+t8v^UW5+vfc7C#WH};jb2L&&AgewdB!;ODSX@T zF3T(~c1hUfm*30~N6qgWc56P8Ay1UDo`6#hiX_n2!a?bd0t29yE3w?Nk5E*T*+ml9ptL8GRMxW}y2^84UC(`<&Lz6Pt zWCxr{HQxF=cR%U~yOhJocynCc09+9eAgv(wN4a;#>%~=2^Q;xJr9P4u)q% zPQmPjV_!Sh*NwlqCE5|Z#h=Q@es`yR_<5^TV&vD(+M?aRa^Uzzbr9E3sLA{#9DgXM zBCDUR{$xKH5=v>x7u23`GAJ(J_IFL)i5Elf9*n8G$63Q8OzMf`3hiIkHKqp-(XB~ry7^JtOMTGx6FGYVomW6UNM@w-M(c-q z>_eo)?OyYriQ7Y?o)B0R-5<3h15R3eP0ya45 zEEvjbfkBs#$>YL7@d7Fkx9oD$sJO7!oagh z&}M8h?F%$QjdviMm;8nII#oCqNLZtdqJkT^5JCosMk@VUUvvxab;@`}rPtyZje%-) z*5KAGSbOPB2AoAz(62%5kOkLH!&Gou0YU%6NGJ!f9r6vwsd6ND6a5h~3pu^y{C^*B z{e7zGia|2-Bd^Hu4rB0u_J9Q|VWKBdJBDxu!y-In^D~uR6$8onZIjucb`n6`=KLIC z2z`y}#$2`xBz9rKYl-=hO4tc;cV>KoXTy9AowzPY=fSHR>*Ccp%Zv<0S%<#GEy0Yp zaM6x|uyQ(UAmiUd{*i*uu?#TDpPalBxiDyE4<#0SJc*G=28?2LIS#BOHZAml4DvZK z#+qd=v-&m)Ko?#Gh0RR3_r?0oqIS|`y$AP*=Ys%2m281~B zr5@=0XFGl!t$+or#POg&FS~2gnjHoA43S7l{T&%T;TGE5@inr{GM9@DNR|pUi(t5+ z9pv1QQD}EjhWVbB_S=3z%o+Pf%=uN!`QidoQ`DK*z2I`r7i>vPN~%CEPK6uz34AWJ zQ8ZPVz=c|Yj?TI~s73f{O3)~fHZ8kmQm1j;}0*dZxyq^*>|{s z862aNJ=w|$FIA?AU$v1YE*ht02&lp)lo?6|QYVMo*aK4RqLJz?Q*smMDtsxLmLsHn zBUsME6$O`Y!oh}oT|hGl-UU!G!KEx)_7GtiA#@q%$KYn?nGT$_OBpDX`o>wyjiIrkfO888TZ;wA z4@ZuE^VgxJ1ste+_N;#0iMzy05HK*9k)D1}Oo&Esp3qm<`vi%oDTDLz zDd+x-s7H<_)~;;>KIMmEH4w*{XNuGX3=hh&3Gs5AYxn+$j7PpU-ur7MAiR^WmHZqj z7^f*L6DI%vKe;;%QnGARIvar=E4tH~f|GH;^0l>!Z5yntt(Ab@VW*<-_{GJ=VP#UU z@Nr40;FmxGQml0kc=e@N%cM@eUE^>A`Mm2ZgZ5jLj#p~>`mLKs0u}hDEb`V#6cFa8rdt3#DHM*4sP}i z7@6b@Y1}{uNsh1!{MnM|0StkBK6C)Z=6H_-QLF3CY`^!Yhh+4O%5P>eJBYN zQvn>;^XR7563U++9mMgT@DSD8C5Own*nx|RAQd2rYV?pCE+of1WvksK!^ ziE>U2bM0znqWuNArcR+K5*gNJLr6{JIe!#g`*>!MZ+LSjwVBm_}?5sjZ)o z?-dy`7{O9m9NQEEzH5eLBaqKgL%qe)kFLl;i9$Gk-i~1mb=!ts28RuUN^EYL97CoO z19oAvfKP~$fpcTT^Vs+W@nokTO?W|iD5x|bOXU4#-n2sw9ti(%a-!csr|UDO6c`F3 znBderOgs!Kv6d4v^XE@RW^gzY_RZ-_Ky!}2lDM|X5Cq2O)?A-gO|aWJPGajN%0N>p zk2)}aRJQ9iXRmK5t1mXr+2GVI21PtMB8sKYg zVGqGU!~_Qak!T-u7D7n6eDk3e(-_coyi{Vx^ z%ve*}Ny+Pn5UiIZNNhRK)le1o@*KIwm+#QF4HD?)HQ2JjV-&FM?W(W+P20Q@~eFrcXG=B=P{Tp&U$f$3Zd2-Uxdh zB-ED0r*s$^mOZ>(Nh9vCqY&H56Pz@+Vb__r@ZH#AN~?YyHk{u z&rcH*l`11-;gcA2AW*2~yOe`Cg61@FF%dMy^Lks3;?V!z$S`I1VOSLe2_=SRr$&ln z-?M<3j?ZnmZsf%EVgxV-0tI1HG>{=u5r+kXoQa;dmGk%mWn93!$IQ5(0l+8VT>-}e zF)IKF^OSAlphiKiT~ol4y5!#opz!egFOlQF0K@+YVE7+s!oP;|KQNq6iDJ?}xNA=( zP9!B87|fXGAuxlwmavOT3KTUPUdqS^?MXSkeF2cAn>gAB3YHMPot$`%NM}1ZEq5p* z(VDgL!cLNyG7R_Q^u$!>&mWbCzov-qfb`M_k&(ea^#Rfme%J7W;6HO3^fdg%X>8px zJrT_RAB<-6h>$@}!@n2}NgWR!{QG(7zqZ4FU_1P;Y5tetTwIzQmM}ectmzND?Q<=%JG^)!r+UVkHY`5pp3145 zT3Dv=*8$tAf1Y^A`12odx%bxr`cMA+2VD3y66xL_fLsHJVhrIhG7ScpFl2ku!bODs z8#mMj(p2bmmrXff60agUThsqJ_Ll#aW^tu?ZPMfDz;UuU5Zam z7DiKjjH6%y&pT8`(VexM=IZKN-Q288{s9+H;sOPxr7@PTT5k-08U78J+owRc6abrW zKGn4!ME(3HCnmaZ(v!s>kKU5rI?`T5jr--#}jd0w_9Y_>b6IFMsXy>ZW?hE9<3y&_%Mu{Gpu)tSRNMT zyejPWe{1es!=M#RV$J)f!9K#sEbQQleG?!-Sw!CJ{|G5d;MkR)Zh} zWGj^-)Sw(D0fZ_dr@ah=_>D{&&=M`>+rDva8?2`;zy4 zazFQT-~a#bRz`MpaFh$uSe!GF(YzJmA$dqbrKF~!-Wz{`&geOg7$#HCfI=8q7+h+o zhZCKy5Ex!ljXxvu-C*;4rYaMnn^;!1PA1P)VyWYjX1(*#YWrg#Ov8_QvMG!jJZ#7& zZhzkLUddbJ7g`N<FJ}f&8Z?;_#H_yr9-hEtMdfmaBWq}IsXx4UB)7D*RQ*oO z`-x|S^#;!8*k;qFvK1HJ&3F9qzqf9)v89qOX!`LV`9qGch2*@GvBAeE)nZ>Nxme;# z+%xSb7h0DZQkqU~ENRE4vSYw{0HYZ*nI|aryZaT7Cxv+Ma%E*@>zSCzt4J?#3ZrR2 zzu#d{Gr!N&_@i(u2<|DxPIC0`DA~`GqON|Ho{<$qXHcI+-jUbUaTi;9vRow_FL@Qt zhX0&VHG}(dKR)(vhDAe+OT2ZN^H%7$s1@lu%t{aHEMLo8hXsj$H`7S_Iy@TEIpHb& zWLuWvL|X3<95kV>?fg366Y>|LetGsk*Fi1(!EiKdAnmvUR$x?AlzGAjI7PM~dxyK* zpR378iO~CAyUt&(ie2UjxmU%X_m&5BRTK&2fp_FMXAGk4=5cwKk-e9JGxx=0vKfyY zJjj^)zV@$e`@!Qgr|{SaAeNfZuuNe)IXN8|*J1oU>AS!_*FT*m8}F(tr6+e7k5bq2 zz!I#1J>V|m=%_*CuOGyK-#i(Aa8V541xZ~=H`kYNFPB4FiVu11rNV5ea#C{@X)S%&%sHC7R4Gn`%)R2x>kB2)5S zVIu#Br~)D)Js6JpPxBJjb{RPDoOnAk6ORf_W-!?rF42*S;2}~_*Vj_;1DsFsbP_*w zm{MjjX9kHz6<#1_OBg0YOryNh?6v7-isr)Izd#NGq>{*Z1hbS-nMq-zadl?G znK`z^1j^|i?kQjhsxe@q@lm0+s2ua(L-Ol~R}cu^r`5AtaHif&v*b(RL@sR&U__@f zY=gd6M1>;&66-k8_ChWvAtGM85_?pgxzyWZa{d+JUtjLu-$lb!>U6`~-TBS&gJ2gk z-xdM-)(wnO7^DRtnb)e1Gcg;lwWABOCXbV#3Y60mCQUQ=sA_LDyI-Wn13oHQLQCl;j-NTk*L~Au>OWkwOXFT~@P9>)RwTfyGCSx{& zqIIKYGX2CxR-D zcCPbHRjbvgZ~d<4IVr?sGMTQ?vw7;wYq@bKF$9E!+(t&qpUz&!xaCPq=PmYC^ebX@ zaoZ*m1LH@p9TG&qg4nRGuC6gKlKL_+iOlUEBt0C5L?(6R<39YA4gkYgA0Zthg$^4V8)Lf=gh>*pmP1Dn zD@XCp4_!rXdEg!>0K<1A(H8Dq6?zCe?UJOR?m8#57jKiGfFN^pYL|9Y`q{%wsyHbZ zjHMrCtlA%U3x$x46bb9OFwVVa9}N}O!SOF#CHhPvA84p|!x2tGq5ilcC>|tKk9xX9 zj4p=n+>7=l2pmD<%{rLSM31F7|LKDqV=MCq{Ao7U23fP<9Hi} zY?1d3JHZ;DDS=_6WsfmuN~JQarQ%KOuBLaD$`ngzN~^GFUZpP(l=yX)A0p)weP0Og z1AG4bBJ}BSe*7X)(GjTs?VxHPDxvPo-}~!lzROEq#Pp?QpI_usB`8ic%L@MhcSEhr zJ~#?8F9$NG1PtOBr&Ou_&{9SfAD$E*gUe-Qlj(9i674jdrLq|soPf$!qcOtV%VY=H znr`DMwU*#lh8PeZl1(O$hcH%`0cHLPh0xqzqP>y96)ndLz}PzvDQWiNvNW)_*-{Wj#X2LA^4H5>(O*N2hG~3!HV@O% zQfXWV(ho#K(sPiiPUX;BfQJ$wg->*rhrQrDUT^=jlUM@Slr|6{q0k@=6k~l98za=D zKn#{Odj~hLa_fv^<-9=<73g?Zkg;C_{8ewddhr>>nFJHYsA;}M`aXWA+R7oWEC0SUY&8JTdA+`sgdvW~!1}O2Vm@H+Z z2+CES!}ccHeGS`@%JB^*qy_S<&e8i->k+pD4nL|U*?^%XP>E|haHx?viU(m^s}7|- zx53Dr*{lanhArvbHFN`!<7aDJlYASKOptfFOZ$V@u-+tNn2Z+Y5^B+uDnTgk_w)Bgl76nWa=sN$iNkI5tpF5fBj2kOT(=lqwxVQ-%_0Lx*UBNsOR_fKoLG zNSOgaKsss!=}MWQgVLl9(mUUJ#_aR%@4e2s-t(PnzyF+XbHzY#<~P6JQ|^1Mb+7x$ zuje#W1lR6ZOQBE%X=lFCqEH0fDU?+g|F#;xF
    if^mGS5x^0-|;o8oacvcpZ;*h zz?ni3zE1wE2Q?go{{2n)-tCMY0Q^5w%YX9#<^;I zLQz)H*4A!rZe~)WqN05FQ3k&M^Ow)onr^#q_lDfyfAhor3&!VPa0v|Y{m^*zh|pwq zt|-@pVPa@x#GPH39q;Qa5pt|lm~IUY3=I5TL`iXtu*fv_+0vZ@yiwYab2$`GQ zNy0?5@3wuAoh=hHSD#SZ-X48WJ3htv*?gsBO?OI0b$OC$^rv6!>wdmY;;eRE`J%gD zjge4~lkB>!;zQgV%3y3}rc~^AH&vO3O!H5ug^9DAzoKlD3l9&k<#0l&{X;`%Ra8`> zzF49B=9ka*9K#fOmQwUCo;$psrI&1w?|x;M*U;lgM!UylM@Js_Bre<+#_S7`rCVa`r{7D{mUt!GBx9t=@!@*b}mkCX5@7^ihmX?eCL!?KJi=(LMm z>6G94q^k_&oJo#RFtt>e(W+XCk<6HG_u5aht_sgQUA>CD?0wK_*Wk?YYWY;&l}y7b z6~(dJuFdBnQ-BEYIK z?c!wM3W~Cpv-9}cB6oqt3*|jd1$!vTZ@Chwbxlq8s6LEDlaL<9P2(q*-gw4*980rn zZUT;ey2%Yxv8CaZf0nVm-;qn(~Pj3#%~5b-9dXQ~df(ef^WK zpA8NkIk&N5S#BB)K)5%!$mSp- zQGh49^M~RUmu+p!u_$gwI2Ouci8k0uIrmnUZ7rI4Ni92=5Sd;p*f8hy*FyTomg*UO zu~CzG_TI6*EEY>lI8`rMcHmJglNx61L@NmtZea|4C>5!Z?TgGP_rPa&MEXoMVUdan z5ArDY5e7Xv8`G!39fz=@44ntHbN$~qP(5keDS(4 z*)`f^`pl;qO|R~c!KuWo&`Rm^?GeczN$fz>O; z+%I;fulfHR@&5%8?cbO9m*(LAYlrg@!8IKDPv{8=3Hz6vS4_Ts{rVQ-Cza;t=!l@n zj~(ZMH~=ZAi8`YF!^22T<_D7Wk}qGroXA*jzM0??*Vs;hfwz+}Uf$v>{pUWoZhN#Q z#>>1@Yjn#Q1o%I`y)huXYTH%h0!iEYM^u`PjZLVyiDBP2lp`+3rKI9mtiv?GnvvheGlDdn9wjUR8%rUN>?Td*4Er<PpgqNCCxNFfv$@T-`` zAsgRpdGQK@sTtN)1JhD+^77*$ zsor~A9P)-mQ?(Oy#s-p8dxstUyoyr!={GHF^|#NASbF1sexXUFWwp7r=mr`THctT? z@TOf07l$2f%9wizwAJ@`kp=dXdX$euivx}ynCcvU|Nac~xU_V9W~OV1MQ9jXqG-XX zi2VT<>mJGDRsXnV*WJe!rEI6le2;M^b&pD5rQcGYpEbav$A_0&Gk~qFV|=E{-d-1p zR+nzFyvd!#z2!B3KBw;JsSUbNpSnO9?8=l0v2$=>TYOI0=4qPu^kHJ+!57yzrdsSU ze`MJ3wT!v>woL+#2#w5#ES7zy#dp7NyG=C!w3MqcW9d2LgZ9%j$rrsZ3;?2A1 z#y`q?O$|}YR;<~1E68WaqUrk^hqXjQ#>&M~EeGlo?l=?`yhOR*TJwDj5Dh}7|g;Qg8!%$+_HZ|FN2e}1OY{mY)f3;nZ)vs;fP zM7bFvJA~hpw6UI>=%0PBqLf=MyFqcL^X%Ot7eh)h>UG(|0>vfut*x!oIa$fc6^^_W z=6h`I?1oxG9~g@#Bqt|V|LXtXQDS15oBfefg1=p5%ucN{F)`^M88PjN*&xxOcrn60 z%OD}grY1TnA=*>4#j(gItJSGuvfN|z*&3VP%6lGTW&8Ezt;0;*%Ea@A(#tJ+oIGys zl|FU&u!Ospr{`Q&$qI@sFLQomYHDf@RO98?h`=`Y)h{glq+X`Bu~<= zF)e@C-qC1K786a_DYeO!Yr;L@_Tk-2S;PY8=b!F>sFj%F#O`S#6d<#jY<^0b+_m^l zg--ufBxlQyJl!MH;q46S0yc=}3YPHwaD&Fw){wPgIsBQ`S|L#sSzTRiS2){C&(ASr z{=QT8n4{l(OU4gxpL=%+b-coopL1B^fYIWmCxaBGE=aez)aB2&@kW-{5JiVFs-)V` zbGZA$N*Z=f?YLYWfpmqA{MjyBvYpixyjotXiL5q0jGT74OPFRRB`MS1&>rUp}vm zQyYx@8zmqRtnW!?=3W{?M4S8eM>hZ+8n16%{pj?S;LILe{rcJHld{+dWd+mi5xDkH zdWE>`#A{_fr#v?z&TVOlpc?4bEMfxo(%jlT`Z9Bu?jE%uu)p<6sqnp{7T-mC4kR8w ze%x|?IJ*WkD`M5glWDx(zmaFB4ARfBI)#S-_Xl1=1$rbR=KkZ4#n4ZluCChkQT$Q9 z`7cz1{yTHIRD(Qur1hUH=RDdo?&SpzpM6uqB2foGZM1~{$+x_JP-Z%X{TTnIh(@_Dg2hQOKO<&;Oio?J z=Slm~ZGR;Bc^*|Ln)4{IfHoOV{`!&-Cp_R+d9J*GvQAu1T>EFZP{u#z^}iG=b+ijs zQBs0>uZx^{@h*@`3$i-<=b4!HmLHQjfVaoO8LgZa=ah>?)gDPV4*n@nRc z3bc>tOit9xufVz?Dr&$Jr1pn{3mF(-D6B94+C%?q0kvG6L!oSI4#hu$8~KQ2$J zEhzQUYJ0+}{`I%yW7ZiH`L6Sj<|u2=U@H{AKJNWZd0u{U`@yxW4{=RC0*)Tne@)D{ zvIDK6tXEiT`Y1tl@6FUgrjF>OhSr5RrySmuuGTOudt0_e`1i?nSsqDMhxxZgNm(Tz zQLmO6{@L=CZ+G++8uO^vN=E6aY{pMtCvOl6-5axXRy0wA8mlNrj^ov)q78-DiJ`7ceKQ#19#(8;dl{c+LyZC_4pJ9<*2)lvtmbp5A#WUp;9K3Hbel~s*DY{s3h zy*jl0DpRW2)Uu(d_S;*2mZb=o|8EQI{54CP+?Y}8^Gt~&zgk&0C$Ife_nR+bd=v{B zEC+kzjdQAu2802~{a0u_Tes>|X2SmdOR3xYxjG|OFY9pCWFYSA)QZMG_>kv+ zqWF7_K0kHD`x{lG{KWQy*m}b^m;&Sc)lGT7M!@mMJF0QWU=8xCg;W{(p38j~U$iWL zRkc4-i+f_dNkLK09)&swRp#&8j(!!V>m7ah(oBca%r-@D_W(hmtT$tVx`Kj&ihZkY zSvJ+IsK|WsWM=N1^39t!eS&0=rq|O)RZ`Q8X_+s%c3}1L;qCsS zA?3Eq_TmyAVj;iPd@J-Tce}r36SuSHrA4OUaqyD4v* z-IszWA6OPB7Z19)xy9%IRwD_#zFo>zhxsr$Sq$p6$>p-L7CXCcP&oEA+a&GC#=@@Qrrf4%WU%5+ zE9Ko3>0kPCD<%UeVvlr!w^k_dchNcR&?l6%XN!{@$7-69y#a6A`tw^Ar@!}Abo3>m zkYcw9p+WlTaDZTimy9sVE$h!aI$UNm5Xby)mTVEkOa7^t^C;Z$%{h<`Ld&>CORl@T zN1x?S)T#mPUK|15pq3Fq-FxPRNHqjVTVa9KkaTs-?iukmtv}xoIEClJ*IZ(}08e?N z!krfzl}^?uEic%XdJkyFs4|IePjDW6`yvs1RThBKsXs2v zdST|nSpTrI;BA0V`lUC2c)wQ` zDSvXgt~M>;d0bA8?DCkzGE?8wV{e$5uLrwuZXohX$@a`&1XgJvo*v{KK^$6Z|K~>b znXU?F)UmFMMAiQ32Vvxg=8qok@xuU zC|PG~rX`9GMuW~-&9=$ZvTGX^FWfjoUO>5~p`LYVo3z~pNWj`G_I%+VYN1cZrB1zx z%AbrF?`LHp-2^;Z7vptQDMqFUX_!Kh{w;h{KF@D?-uZ=~OgFe;HP~yGUSDODybLWJ zC5`&8KBox1$$AmcQxFk|i4AxplYJaS3@i9M?cl>Z6eklk_tF&S-UP8He{1!dOYj?z z=c@z$GJ3&Rr*A25D4Y4vHD0f+X$5v)NtAU6*Sdi|n5HqV-y%0>VXUIAxj8&oQ!!`D zZT@qtjtaMX^BTkej7I_Fz#`gjX?7^P!xwY<3ilW4loLsnoqHcUp4YJ`j@9Vh;ZYZ_ zUh)3T(nCI5F}JU`6mVqzE-WvtskG>_;Y6oDl@_N}#`B%dE9^qd(KgG;1H~g8N_Mn( z6@4@Rmd#%$maT2?X`hIfu_Cm2=yB;-m&o#o>^3(evTa;kU9B)W742TUD8epf-`~GR zh-$?F{%vFU*R}V!HD69J{^6NGPUqEi(j`6ozX|y1Cco%fd}_M4H~!EN3x!TfN^tnr ze2M(guYD(z&3ZsfE2X%`8P6Zr_>4HucCJyn3j}`zG(N?3llk8Vd`bfo?J_6R89QvV zSSJ9Tith}UG+-W5GQv0KBb-&CE_F9iJ7G#;dhY5p2yf+d;dB86gWh*HB&shbM1^W4 z>Li~j{<3~jpT{D>g5sZ>Smu-s?PYPQB{j9x-oc?7WFouv>5C2WhiJ7R9g@{2_L$8) z-)j5W#{oK*M0+N0QS$r3RCbZNg@uLhM65_$Z0sJcnct#nmqY%Cl0(K%kMu<`Oqtz( z+z>DSLi6PL|5%)B+liP{mcDWtBurCHt-`i6aF>GIR^uN|7f!Mq_tM0R7RC&}A={d= z&Ba!0v>jBE#>5JDXVuJB+9(E7N!5=2*{9?cm@V3~JZwa&L05SjSLo>X-!Qw&_Q#}r zOC^8pe>!2G01v|&Nd-x-HjE)9Kx$b;$3pKvf5a~5%A+)g=+FW#>1+@0roc7$qfCh0Gjs z`{P|03dg%1sZN+7gjsmpVznZE&! z@*{=ijes7-nP(fVM_O~pGiB5$&ZoHKL&GN1IaaE)bkuBlj$8NoTpbn@zOJcO>5YnW zu@Nz6NB2j&e-A2-0$^lxbTl|ERHsZ%7=ICLdtz%kTBw((Q;vKg-Mg0$DOR(v!n3lo zNx?o*FWK?LRF$Le_;+Xck?Wi{qW;#yhIa_9pMAE`_ZBLHDU%*_?UnrdJr{eGf%i3e z^hJ#?jMs3#_vGJ_R_xX3`Hp2@vz+-4L4GHRIG6?f4J5bzofH2R%37G8L3W-vZLE#w zE0Yl8>v>q5e?LdM8!TzVdxB0fsj@wHv~7=ql;A7z{4;saiTBj9w2LqP>32i{*-t~% zr2oh3yr8Y@r?CsBnhnqLE9j?ebw)=Jgk)zruQ_9r_aF?gUhd^zHy}UW@a!BJ)xKyi z2NFW_`xJO(*iB)30#AHpBrlyY3sDXC8|%Vey3@QhzVEJPz5Hj?#EZ|25_O!&0*{Yy zK*i2Qctn7{c+#K|Y6Q>E_R_o{dpdXlrY` zOs7|9`i=$X^L%*8|1oF+ydcCTYFUEPV!yF#gD$GAG$z3=Bu|jQNF*eF0UJuBgxP#E z_ZWt@@G)f**V)8mw(9w}gcVFWF>*jMktV~DCMB&a&oCpAio`Z$iOjxIIcDGd)S&V+e>+&y`C!mn~&dNUa6AGYr|!} z)ndh8>9xP?K5}pJZfxyz+Qs&mx`k_Eb4vOa7M_f@Jt0wU{mVN?m2AhfHHn<}@bTj- zKlhd16bhr>lcIA}`_4w5I7)LGuaZGAc&t8nTJvQoT->Wt#9LLA24YtRU9_$ur zatZbn0h8Ln zF`oO*-o3R(#QO-$0B4Pij5@~2HehxaEe#gw$&vayjR`F_14;kUV=ZRT&8FibLK&AK zF*FBlJ*4WG)bvN<(XBykN$pnIeSfW>C z`Gd%Ow1|SI%MKwSNuRk1Wb4V0TpCf~_tCJJ4}V1>!>Jt|^GnA^80Xs9X3-0N+CCp< zb4_BNB|O*?9PK?;Mq`rous_DzX#`#iYQ8;4pZkkhruD%r@Zohoi!mOHj4i(9>Y`%TEkZdd5 z@$zj;gAe@uJYj|+Bth62)pBD0NDX$6Uv42N!)dlB%;wD>--<;^Lf?M!-qGjX`QsWu6(;c4l}wAd z@!a(;QU~3Gt(I^5&eg^AB6?1*t>Hobg-HSE6zc9hAWahgsYLp8#@jif{3N&D6AVbU z`<8M#0xut1kSYz-*?goKB19_AH`Q2YR7vx<<>qQ3c{jU~Adwhgwg`vZ%Qw02hli`; z!c4>QiyGgljKMppT8Z6*X3L&{>i8%A?7j4o4;7^aw*gG>8ss0{T|D04vDXbLo$8^( zFHv%tj)$T=x&S&e*@~nwJCUFiHI0ZX`}U_VoPa)hN!mEvh3Q64&T{Fm%^x)3l~457 zWi;p+^^oBA{O5J3&=3R5R#W%g%M+T*cK}Owy4Ii18P00aJF&dZZ02ibw^7F&XGiKd zE~ls#dWfd1tULuw>RLWiylSH*xC(3=!+ClvUPWeMYQ-tNx~q&Ft;8p_&cXZC%e%{h zoof`PyVoeqoCZ62a`can_u~q{&lDJ(oydw>!eI0NC1yf^Y{@g$JTXUu>*2jLDFG!qq z_p96Y6ujKXVs;I;FkxZjZrT53^?Sr>%CPA}d_K!O8zFso&v(CH7;fweQc{G8ioWI0 zgJoi$F(f(Ep+rh93Vu>+D=}u$2-{PQI0v3s6%U0o}71CWOp^6?re3IQGuGM`2<2;g;?jG5FW6R;BmlI^N)24EB=XU9h z6iI~4^O7y@Hs_F13)Sz#R=>xl<3ue1!?KIlhzh5(bM}!eG#BXV=GNOxPgQ3|y42DJ zkMmI4a6*2tNoHl0%XiFHQ!Df;qdj`>&Gw0*_nlC27UL(~mSUsN51HBJ^v6fAr;TIgTlRoCMB4IF<_`nFIaSb0wYFA~VbRvsn>Qf^E=;Ysf}^Sqj9uaSGy90(Q|w`qo7_SdvL>Cn_qmS#$o3ZIX6l?K)5 z8QF(~nR*^%boemZv4E$Ajkt7bzgTQo%f~yqIt_`#g4vBI$2c5MK&-#-<-2Lx;fgbF zb4|cGzFjsNbucGoVh4B-Poe@#uqc<#?(OZ3L|ugpu~ScozBLv#@i>w?mt|8@i|YsC zoWx#MW!fR)U3NFgF%b_7rgk90>m6`6Gvc6_!~U?NBFHhlY@gW4qTOa;KVEvV6yE8V z55l8fTNEHn#^(Z~Vwt%FZeWo7fdEJn&XG7+V>940;4ddVNMs~Daj!vE*f4SEX*3au zp|G@|pa6SCEdQ{KRrwtNtcScs*z@c%8}w*IjIg@Nok7X`4pppelY?@R_$qY(M1q58 z$LO4tzmRQy_2bfA5!9T4q~vr|)T9E32L{StZZ(T^%E^EJTNAe#G^`F5PekK9UZMO} zGw&-*E3fzL8>(@&jFB4>g?DW;b8C%2u;II7Qh}2$yj>t5i^e2@=&{ouiFzxiyx2&ydNrz8>~wd}Uat{jaVRKg3NWM}u8@I3Y5yvr*;ih_U4 z@Nn(sZCMoWwhtyYEubB+l-dy{^8xikd0CWA+{uAfxdc_%C|qRZ55hx-G0W_DL9iGj?%kX zU6C|~POy%QY;k8MOp{ie6a(TO9Y^x&R^WNmh?c4Y0b;BAy-0V1WJb^^Az=Gyg`+=9 zq-aOSsfG2D^u;8VUz4szR(87_Vq|5CK?VW8kv}~Dl*WWIy0A1|R6~Rrz`Y+bWmM;j z*C_I#eMqVF1o972%`o{+s8u$6KOl1)0hG8mfUvx3-@Li~p{um-{nLxqQi&GEuw)2;V$xR^{aX;j^u?n-LW2ntnBhN@OMp}Qz?@7ib3J9bO z2`I~Ssg5R_m?+7blae$T1FolNhxW)=&%eH|cn6ub=`wo#PeV($7)akhe?mC*>%gk4Lz4Kdu8;Ii(iZ9k3K&x#WG1)j1^ftz+Lv6 z8c}F1&#e@iVTd%@ojZN!^yy8NU+h|4sch^1xB5+cHzsbE)8IUco74!DHk{RyRJYnC zpRYRL&R)NEucX=MRY!!Pj-Ax}?Q7AsyG~ULH4YWDE4erQ7|B^38nH$(X0VwnKwso6 zw$6X%-Pkc{G(DaX!Ze`IRGj@LJfVwDNM zL?4U<0@QbL998HCI3iMuB$ zmv*lxF^L_Rcm8pq=&`@My}GrV7CLgtr)Xi0Tcp9%tzNQ+Zb^tN)dQi2YnTilJ%Z#b z?BUjnf8FgJw`SQmhc=Aj<|i8(<>LYZ0v)cYpZeBqlmI@y)dGPD4yOFiuU|L~@r8t4 zYJW>MPn5fU+mSE%HW)B&B9qcnWU#}eIG##;A;tW~0oQ>$y)`CxlX9FnIp67yzpD>i zFl9b;Z7EW#*uQkKZ9b8$k#%8Su=M`U&7bnkFvN{z6M5x30wX7QW=#=)9?B~_1t0hh zz47YO^w*PBG9AYB0w`Q_OWbXC#wr zN|1dydyzmJlyR(BfNvHV2q~eNZrH1YE?10n<&RfaNVnK&bDEVF-fr}G|KNJLVGE$l zI*I(WA;s@*?r50$?nN_E){J)=)Fh%jG7$u^mVA&nj!0aj^pD}Cq0r(Ym%hMeXp zQwFf0V?tb9TuSC}&Pd_X+#u=bz+eHO1{r*smH8*B$ml- z1E)wD5;EMd{TP+BuV5yS>2=!=xVw@HLmY$GZpE^%aYY?&AAhub2umQYzY4=V=&(J% zOI9DqbMs>7K`gqoL1E%`I;T2XL0$uVWlCT!(;X)k zaQ||6x)@vv>7#EbqeeV0Xv&ze5hR6>7JM?#b^4CBMj3LP9;X=q=qUPNMDwPIqjBm) zAxU6>P-1j@uhhKkRwp%6RK?Oc8+IH&@FGAEzeXX7`1>$nrHWIJhmg;$SW(BURDBn^ zM8ZtwQJcDyH{00^xoyuZP^vzEG*Ollv4J3H#cIQ|I-|KI)Fe-V+h!3cyH8}Hb4`U| zOZ(sm2zxMgT<9Wi`lg|)tEy4Rbf0xd_}1lNu9VcE{yOQ0S^^8xe#?x@NMv!jxw%@* zTJ$k!dZnj%J%UsIIwRDH!|C_QY-U9le&kAP2!fjaI(oiPjg+T!FZAlT}!V z2?Oj&G|vIZPS|(mtrb@hnqRYS9?-*3Gf76}?D^|T3lB{C=4Sh1B=V+6^NU;CE(Omg zgWfxFM|cO8CEj6_?O??%1Xjg60KcCM8ZVQ_Ysw^iHGPo81CzckcsFs8Tf3F<4$T-v zEAA44F)j#^74J|2eZxgC+bz0YaH`?r&^^l8UeGy0O|sFwhWHYcPB(LT^Q{%v2m7zt zRgrXZOf`pJ7dhOMT_iEQ&={j<01F(?L4O!=Ftlx&tOmEX6{}X?`FM#AzvXm#jMqqx z71spS*1YP{C`P-1+c37|eUnTx65Qy^yyeBB3U*7jV@LWRcyTRhG)kujZ9DphNoLq$ z3p+9udTyx6G}afLo8yZpTF&-)xmh2d;Wa%nEGGhcCY}_*wnaA&zrka@`%MgEgJqnB zC&h!{SJ;acY$JVrE_<9rqI9}n_a2eU`oo1j^H@}oF?&OCH~P)4?9w0+Gq+))%b8@N ziRXgbFxFo7P%h%1Uz4j^oYv)H{}s3(SYid6l!Dx%f`Ga6hMLKq%Cjv3Ys6o_|KmnE zI}hyk<4|nCPQp-tJ7Or%35KBn(nZh&U`#W~OgMWF5hfKp--_I<|dSwb*jJyP4~^5ZSu7ZXX0wYw)pms#|h~B zd0peiNZM$;t(FwJ{@YNe0+cJgyG5i5kj@Z-Htr_k*hDY)kV%YWu)_-KWG!Fa@_zS4 z%qt|4;H0Fae5{e}5EV-xn-Xn}P?_fTqph#`d7F|_weR+EEV+YawLzy@)wmBPRl#%O z0=?i1jT09`MD&NCxKyKFjt&B==RbYAUBM}X)$JbHpZ&xU>Zq*CSIsF z(77slq#*opW)C71E(z@`!&Ztx!8IY?q-a9hxus6^jG$FAq_b?cA1OQWrgLgS7SFd}HrZ|5w&|rJH`gifw{$vVki#z@D!w_zkr?5Uh9TXx>C0(UrP%h1!Vz~Q7e zco;CZsJB*O@Si>#+^3eB*Aw`26S}{eb?4Cmn)1qZnzuBRB=d6l0 zxG(F3aI7kZ3=ieensVR6n3`u+2a$zfYuea&%<{>}$w@Jm2g2XjHIaoc5ezlR$GDD2 z&81%=W-&$W6cDr!7DqTo`v=k2dk+&(Ya|}%(H82Y~H_qkSxgJ#_^|7IuZd`M_<4D z+F;t7#8(^IsX1tHBqehfUb}*(!eqUuquh7%jnCG(Y?62HP;whxIckc339Z6^iMWlX zixXZ6zjig&@?BIVkD1NYhSeg~5`%{9P}+;y*RRt`xVykLLN0wY6$EQz$zd~KzDyIi zg=R8VVqeL-{}}lF8s2>A22)CV#RgMz3-dKRElu%D<&M@hW%m@as@*1qd=ps0WQd0N z?O^`uG;9NHvpIUaj7*T;+_8G zg zvQ0>xpWpdO$QKGUdbxZN++SVe+g0s^6>SH2v#z}R_<0v$pcoS->M1fr(p$Z8S0uLl z9%8uJ+rs~-#)mqGI??{1GX8d>-_zk~QosvXzny7X;Aw?TLfQ|;0krnh2zx_>RwCjM zk$n?GiDJ{ZK7_gOp0(Y>awOlC{rI7d?fm4OZ*NGx>(S@r==r$nemC(x`hk(d<7z4M zd%t~>4{j(s^0bRd1!%6A`MvX^S^$4hO_D>t*vhkr&wyJ(R(wq++2k0R+43f z%JOAqH&Ne5`3Je4hN0_Peb`cVT@Hi;`zl z6C?yD#WB)S1Z9rX@74TT^z+eOXqKiiW!l|ihq#q@kM5zNX(3PN5B`TYTH+Lk_?(;^ z6ElSPM0YdPBUHh9ZFz4p_kqCAf^)QRcrcJ~=EAM$Tp&5thn8rytDS`+rrAl%X z_;?RW>D2&$`ZOM%Gfw`O!NEai^iADq zr~^nOMuP+(67ZE8f=j971~^^vCV6m=DG*a0vzsXB7F`35ucje799y+!1Cg4?(Bh`^ zBvmSrrSq03HE8B9C*BcR=iY9>N_A-S9Vu>$XxHf>U4tn){05O?FC_VcYj*x{3Nzd= z)ks7f*qXlTm?@DNGh3cC+kV9SREsm|=s_-wM(WQ}kVXx~7-b;7M~S@H57aV-in*qN z!MEf(^>ytdhReGg4u=>)64BOm&nEgF@fFTO1CMcpnw-dHzbb54fQqHbjC6nZYkynb z^bW3`uCCR)KY!iaAFCkhyl<(9*n%7jCiNT&COM=9Q%s?QWt44Tdyio zokPEt6v=xeDUDP1M1)gASOn8`trEavgEoght4o3hMFW%FDY8ac@P7`{5tVM zozkK<(~=avrbpUfq>Ch_4WWTeDc#FU9m}Y|z+NTl_PZeKdwKV;{`y31qIMHC6;;A} zGLELin*=QC0$N^souqYh$?W@+jxxP-Z-nZ`V`btQ<}yM-LJm)Nqm8nhbe(GY&DGfl zIRf^Ox*HC@xPzv*!vvd>0>WMzaZpj!R9J4q%~CY1w6I7BX+$9<=Q{MkOI!_N5ka+= zw{OYla)ZT_G&)9*oJH1hx=Z%ZR@686ePiU~kR=@!HH$?TW^4R3m?%DpJUX+edLdOf zpR7rTr*8w&MN*TJ%C{KV-F9fp{ggpcqq|vt(~mO6hQXm)mxphNzQ@kcNOuVxN{R$x zqcLGHemMLXvdlh%+~Dbgt}P@@qnmci2s;_oPAZKcw9Seiic)G|ib(ukMk)dvl^M62 zaXy8+p`BPj0YIvi7VDH)Kct5#_*TpIr7*7HgDGi>ZUZ_UQ-dZ#Lg}SD93piu2USJ~ z->bs@l@dO|c7aR81KJ?Ja{E%0(I_4fBad25Rw)UM>!C>L8(;aB*xDi??GiLV@;>5- z-K}&59$U(ckGKB01I$qYDcI=Zp_e5N z1uNT$iH%fb2tOd%l(??}>kOBd58c*Ev}B8uTNk#Sbs4y0pIs(1Kiqh+VCS(9{R6#Bu{qT3Cw?#|wSQvWA~<*inr$mFgNCk7lYu^~`i$;HbUcZI z2`y$^QmtAN6Y)l6=ap~J-&&O7Y?j^bNpDFt;#>3bpL35vF|^J5vE^7pLqmT< zvOcpLE7i2?N;?W+)H44StMuF#pqkT&YdI^$7f)7O&*k=k?1Davv7@_P_tFzm7W`?@TTXHJa$9J!HyA1(s@X@bBXK2oPk(sb$!n`rDGmIPdHIqu7&AKMKpkxPe6hMfNKt7vn z&de|5(7#beL>LG2>_!YMgWUxV+uGUP6{~l3K(QXqRpwVp&Us)sWx=3ZN)CC$dtrRJ z3Jm&os&|hV(N!Ra2K7J|D+5o6At-~KkD@v_jl?oJ&`;|2apv-Irjtwe*F2+Xnzrfs=2OPSU#k0@w-A2mmzkRDYEr&e$>Rkfl8O z33$PU1Rt^jHd{-rpoHie8ATvC?4=RB;xXRq`h!Yw?<2T*ZVJ0wq)*x6tL2xw)aiVA z`Q;1{q|N7V}bDWE-KEqZkew(*FB)7TK|2xa3mB z*R4RR#92e^=%hgV5RJhe&{(z1%wkX{^Ie>SglAetVj6KK6JCyqH$_p1G`^FqvV^2$ z11P~iTPp$L1xaAYu$-B`_vHL1tjOq?h3@@gZRYlTM?IMoDmk*Ig&j;whv}1OHpGrV zAQ~YHK#eMRHoheWb|7*Rz%ht|-jJ?q-oJ^}Pf|jHBIh~TUqt!OBn*OAJTwEaL2?z+ zT}zB`1oYfFsGSahC3M!z>V*-4Z}K<~f{DG1A)ob1D;9W7SEFM$45uDt-rRS=6qp2R zdC-%aPLB|r@mbErE7tHWLhvcoq)dc^RHq(pRS$LeWEi`G)CjR(x^YrZ4S~TSrqFBn z4F8I-%MdulEjx>aK8Dxif{-SrDSi1 zK~{G4c#Tp_kazbUQvBT@_x`>vTL{PkdLj40TZi;$Nq7Ce?qJPon1erjlpRR^@E7ca zz+6SpBnkE;AfSemjwIFx&zT-En3yoyTz{SCtr1RqX-%qzPz( zw?&Wr@c;{n9Ph=9dvQBCDM@-_x3t=K$0@=u(GL3l1bz})wP%NPaTbj^r;J5kSGx;Anix&`n&oU!@=laR!5WT{u&li%P z8M#z(GNpvuO4e>@NvG#JS}L4Bi?_@Z0&5_*b5_l4P7B3*=Vt<2yVG850K0s#!^{ha z!~VUS^!AUch$$WjKM7s?qXuSOfU$)d&)=;t+1ghfmB^2zHf280c6iTaM9~}eHf1>X z=@i+lO63bDpAbzUxnNIwv(Eh}^#K2^?+>gb7h4H2A?@HN`Ga&{Iz205#*S%eZqawP z(IQOHQ0cSLbO($td1R!bs*3lRUERh++k7eyf7dE@RUm>@z zPtvC6_g$z*3t!UK#Th#xqD!QQ2L&I3F7!Penl1v}2lLG@$mBb({P~#}YY_O4SORNl`+E?q zQa@tt?QCWnu+Ke1HKx%|>jNRlg4`a6`D;iVNXF!`c z0rnd5n@<|Mbu3r&NX3)HTcQ?_PC1>_w^26{gBMzg7LD3bo+KV}lI0a~z?}Xcd=HcK zD$MiGJ6+S_*9+<;d?o1~gechx!b;d%9oYTm=wv_*a$=cRnQER(#be0tf^GcvmEwJgAjssr530vpe?oKNMLgWmx+xlicQUn?$B`1@rJ(K}iniCddAF#f*Da6MtZX2*! z6L7xRf83fpTsg@4WauPOmZ1`&vxM42rh*bAtgIihRf2ZigQ+H9gRo(u8-E)}%WiIZ zmRCs#08tmAs8=A4hG4=E5OdEKG@)P^j}o5}tXP`e$RKBZe0=sA@i!Uf9Tg!h6`9i> zSX9Jn9q6#CalO=E))*%2KnaMM*W_3FSE&Fl=DH5@j}5ulTNS~5CcGjd79|L@TtOr% z7S*x)hQCoMA$$3fAmzX#!Ous<4WAriue!fElTOY*{2h%@$cjSm{?31n z{8G=FPEy;LLJ<+I4n83@Jk$aBDmB#oN!j&BuWjYWqEn^#sc{?Dj8N)i^Rql8X~HM% z-#P$)lQa$aem{&kd^_M?7#hp~QJyZf@fZ4QYhZU`zXUtuoAtzS{#-Rqzu`CRflBZj zW-ZAHCuqUR|L1p+-aMKU=^8|Kr?+HAkfT}%86?idy)@DTGby`?|1iAArjqW&eL!R) zLVEdU->^L(2$0kQm`l9vgRKWhkj1+HE6A>rbide@X)5pDhZ8}l{cmESPZ5SnBju)l zy7mf3quT=wVbH}5FS>jja-ovdpSp8BZ%KVp%)f&ap)xV@!>QAs{^Ll~V!r5l?p&tO zAFtrCdr~5shw-;!sWF!p=N*6Hi+>bp4^3 zLHK@%LwS@Z`9pe29#~x-HCDj}zKBg3w*0sb{wJS}6o4=PtDpblL{@)(`{MokpO8j% zH1X<^&@0^bNsmFgB~2xpVi%7cq-|0#q3{fx1Z>{z2Ng=%ED(??7}V$7i}$-ovlgnQ zu`Lw62U)Q~_kDiKKc#|GW(;^@Ba>LF5HGwbyl{o{EM%v zWUsIbfY%CvH;kO@a~@iQ+|xajMsH}GS;OIU2Zt%i6|Z;mEnfoWQk8!#?gJ}u{>A@v zg#XK@{Kx+I_a!I-{~n9~SpEOPv3S%R=wPnw1tv3o!))Oz;vzkUQ^ue#iSth>BE&B@ z*T(}@xD4XvJPCNsk58Z-L#S&Lp~Q1~q@bKj9R6G!AZ1H;01?Z9GSU2TqZcRHX|3Ra z29Kk90NZim@MDyr_YngqNveEYZ%eP358aPeb<*nav!P?*tTj2XlmP9VnRhoNQLzg# zokmdsCMv5;SMCJA>F$)5-LVypzzpP&CkPQlx#8n-9D@kwNiS*kg)$LF^g#YmX@+zl zEaI7m6LpLzsnA?C`>LbrfPKJvdr9qr{q)^$#E1k$*0MI0n2`vNLtfQEOc-bpAe*_- z#d=inKba)^Apk>tu8ma*@*bv}so~Y+5e{8QGiPOz>}&-;&2~ALOC-OdD5@}3`A8#+HM0`_hgy%-v^FG?#)No$8 z)T-6%&oJS)3G$O78!RoRN zw&dc}sR%lY0oa@RVHWbwggfpr;l+i z%tM5=Z`tr}dH^*#avla}X@&HY9>;SA;Ozqg>hBg$Qx5#plbVNoO@#9`ZmaCr3+pLD zsp?;K65kLr#pOjtkCe3ZJ{rW#jOtR-8AA?jAeEZ__5v?76sK*D_5&GZ@>>sZ;#u>q z7=Q<~SH)PNCOg|JQ*;k=?Uh+NgNF9I=VQNPjpdLYWRc~WQf`i)zu<$9UIF2@)fBr> ztOO#Wd;~u-G7Oh>#kO{zui&L0A?hEC9@g=F^V- zZ@6eU+rHx`>pK>~e{+|ILzRlYr72r3R??R=vb@|HqS1p!j*TO`5wdzUftbV#mb`EY zJwTDSh{q>`hK+M;jdfmW4tK`sjs7^wHn`Zi%l-6PE*>rH@rj z+!>_lh14tnTaLHoG0b?1VT;OkfFZ^_e%mi4qVn+H`A6D8VVh8`sfv=9A^u>Te@s*& zRwf4sPZDQ85;+(xMU3*bj6M)GqKFdb3|O$fXCXW~lT&Ec{(&_~VgZF?9SmMZ( zhT7#ABpfR^HK0}c5D|%Zuf=}KtHAA0X)O*dJ*(=iYFO1?F+6NierFMeZM)s)yDG`) z`{e%-0LQc`JSjw6f$T^;!{9Q_Zd>`jZU4zk+HhxEebYyO!;1S|a5~tnAL{)1CU>9Y zwT~xVIPdv4{`wzq`CkYa{*6WdC0O+TMPRUb<#GkhC4n$P1_t7IpJrucmBg*0v@2DZ zU58ACxFi2`kGXR_N|R7{kbW-w4BX~FMM5=UKLOaEExu`XsJ%1ys(oZ7WbYvF(n1U0 z4A`u^zvs}r2c>c5Z|>_J>WC7P3Nl*Sc3YIA5L=mXj__ehL7XR~3BI&-?hz3Yi)Se= zzks;IV{t*yv4|h+Hr@HpRP5ht>%Rq<|Nm>B6))3y4irb>ovfh2g&OB>81$-&r>aH;!Y?q6k*> z|H=Q;>glyOsMP)iIgZY{XisX znp{hEOWYw7hQdZAMd{Zylj$N-H*Tt#^z*9ecTG1nNxJs^8N276-9Pr+KX%VLj&nMu zXJ($~`+T48_x*XlUmvM<-2J}3KFV&7P^uK`wO082D$s-^6xoxT*M5BWzQ^c_y-*X_ zV@`er0(*5t4C@ODt`~h!+U!mZ*@G?C2z0?K2SyYQy_y&QESJfgA|fJOIy|H0*WqW< zaCZmQ>YG_vS=2h_jg_d3HazFpqw3x=l-0Y$6B18h9fHOqo2cDP$@eXP)7I7=c`Yxx z0BitZh6t52fE`2|R!RsLhYnyFZjJOyduoMA=_JifzM2!P-?-YGZ-*sB1+l^4B>?Eg z9u^3~+aj__>|hXsEKqPU%iS{TqGp@W98j{pcoc;sOD-0Mk&iv$4FS<%_;c=66dg8P z{^!?^_un=}048Wg#b6t9Zymyv2uN_ZXcxB~+ z*2TwDx@mDYJLiwHA0c=4mQb{c?ksHCK3+EWv4g}|QxqH=ysYO1BkR~35I$5WhRXgr z)>X>|XaslF1etBgs^SMI*s#;2Tf(f4heXup^5m)ICTWwXoJA#~nexWBSBG2EZ)cjQ zJ)_SNlX%+<<9tKLN{nV64+C_%e0_Y98-EkMBkr=biO}?w;9&eFXK}|bpSxt$D+LyS z?9R^4X=*Wc$`-`XfrD<1p9l1+WrNoYCl^AeQVZI(<1WC_l!B96Iw#3P@lYTjX#e9M z6bcXEkRLu=4iOe4&UGuXaCSzoAAs&wu#9z7%{98C!zx#SKmdfo!N-e%4_G547n-~O zY4?*`a{S4zw<2@$q)Fg+e!FP!m0p7LlczaMGu!0DeIKFe((spjR_a9Myg08G-`JBl zNmLgu?YFV5yT1p%77WH43wUgf=1Q2x?%&XFQA%(M_K4C(aOCl8*i~~-aYVws9R*U8 ze;ory5$*++x*gGDF3J?+$BzeoFo!>?Ms3@hD~!bM%kict`N}V+%r2$t+H{Ah^%tzS zX8S(*>r6R>V-f6L%ODqoWuUnXc4y(jR#Y-xKwWnYui;7iJQT-`Q$@%Kye{(sf<#)c z!?;{W6#FYzN1RNXCGhXN;$#|a0Ar3&l9k2$m*8dnoO_DC<%)DGrlS43z!0eG3c1dQBx!`kL2bP_jvVFB`p`JVF-|w zRcF3NYZst3!%kmz{G_8=nEYy#<$tr?)W1(n|BHSD>F#J-G-ObG515L@`1RNFm6CMB*AWz?~M^s_P2Bx6mMT29!xp|DYFQ5fbG>`?)E zT;~LEQ?2AksIV1nj_wL#F4y=w%ajIj@9U-spe}PBVS{ywS2Bg5+ z?9VACMW&bP3K0G8qVLYSeR%s8#F1#`gWyx<1Ut#ue| zUxqZ&0d~Ov7^8d#cus<{Omr(hKR+xGnU~sc5AifE8tUCZ%9bSC{thwc$YeVU6E%-OI=(>FUSguUpoT>S+gSsL^_iUwR6QhRi%2PCMTGoHLPipl zPMSb(Q2@rvYbaiisk(?eO23&btip_Kg$q;>@&}bjALC*WG=&Zje{?%am_*qsD2K!2LZJa%_wl$uk(`=hzVD(#*^wD?1=sHJ{BP&m%jl$jp?n zNQeGy1;LGgw{iP1{#h-!a0%=q@b50|G0LFEaHM?Acw3HVyxIQ1yH#d$BMj}|{n;(o z)9gUwKAlipPT}TZT~ik88KztnyY$VI*qF$Hfd2X$QdyZNg_`$3Erw*!@=I)5i{r;#fBy1UPm*Mv4`F%rK0=|$8anlY+ zJ@5F#BWcO0)Pww&={WNPKqR*+_F&pgeuw*I=Qa!m8g5wWw!sm-+;YO<}7fP&- z_K%w)BM1zQ2&$bIVL?43PmPuCLR@A9K2~=b^)1Q;Y=z)mScaX0FO6eGP6)rU#8$oS(c=Q(9Jpl9qty`IC1fN(MvBtyyOqxq za&LJWTCaMV4YBZMP7a6qsnG)k>tE_?LVDg?0<=QZTmI$}ZXiF5SgsJhadFQxd<$EP z1U!w<^pFV;T86}*ALioDt}f>l$D11PaFMONro&hjU`G?!ROA?fh8X7`&~Ft7jAQhD zhPtY$t;eM%zzcGFSq3wfCI;2z`BkTH)I9uaIG^FQZn`viN7H}QX8TR+OX>l=r4`;p zZkoG8X(?V29Rc*~l$8MR)ns`m0b$aHQ%{poBZSvB>NbJN@D>(zOtc>H*y*5d|CgRp bRRS%&BUM_1{s(gDQ*c+gtjzp_&)0tkR6>;f diff --git a/docs/articles/preprocess.html b/docs/articles/preprocess.html index 63134069..5ccf1133 100644 --- a/docs/articles/preprocess.html +++ b/docs/articles/preprocess.html @@ -50,7 +50,7 @@ Reference

  • -

    pkgdown 2.0.2.

    +

    Site built with pkgdown 2.0.3.

    diff --git a/docs/articles/tuning.html b/docs/articles/tuning.html index 04ba5940..2092dc92 100644 --- a/docs/articles/tuning.html +++ b/docs/articles/tuning.html @@ -50,7 +50,7 @@ Reference

  • -

    -

    Site built with pkgdown 2.0.2.

    +

    Site built with pkgdown 2.0.3.

    diff --git a/docs/authors.html b/docs/authors.html index 16080f18..b1dd7d92 100644 --- a/docs/authors.html +++ b/docs/authors.html @@ -26,7 +26,7 @@ Reference

  • -

    pkgdown 2.0.2.

    +

    Site built with pkgdown 2.0.3.

    diff --git a/docs/index.html b/docs/index.html index a8639e39..eb215a3e 100644 --- a/docs/index.html +++ b/docs/index.html @@ -57,7 +57,7 @@ Reference
  • -

    -
    -

    meek-ROPE em el

    -
    +

    meek-ROPE em el

    User-Friendly R Package for Supervised Machine Learning Pipelines

    -

    An interface to build machine learning models for classification and regression problems. mikropml implements the ML pipeline described by Topçuoğlu et al. (2020) with reasonable default options for data preprocessing, hyperparameter tuning, cross-validation, testing, model evaluation, and interpretation steps. See the website for more information, documentation, and examples.

    Installation @@ -270,7 +267,7 @@

    Dev status

    -

    Site built with pkgdown 2.0.2.

    +

    Site built with pkgdown 2.0.3.

    diff --git a/docs/news/index.html b/docs/news/index.html index 3497c02d..35695dcb 100644 --- a/docs/news/index.html +++ b/docs/news/index.html @@ -26,7 +26,7 @@ Reference
  • -

    -
    • Allow kfold >= length(groups) (#285, @kelly-sovacool). -
      • When using the groups parameter, groups are kept together in cross-validation partitions when kfold <= the number of groups in the training set. Previously, an error was thrown if this condition was not met. Now, if there are not enough groups in the training set for groups to be kept together during CV, groups are allowed to be split up across CV partitions.
        • -
      • +
      • Allow kfold >= length(groups) (#285, @kelly-sovacool).
        • When using the groups parameter, groups are kept together in cross-validation partitions when kfold <= the number of groups in the training set. Previously, an error was thrown if this condition was not met. Now, if there are not enough groups in the training set for groups to be kept together during CV, groups are allowed to be split up across CV partitions.
      • Report p-values for permutation feature importance (#288, @kelly-sovacool).

    mikropml 1.2.02021-11-10

    -
    • New parameter cross_val added to run_ml() allows users to define their own custom cross-validation scheme (#278, @kelly-sovacool). -
      • Also added a new parameter calculate_performance, which controls whether performance metrics are calculated (default: TRUE). Users may wish to skip performance calculations when training models with no cross-validation.
        • -
      • +
      • New parameter cross_val added to run_ml() allows users to define their own custom cross-validation scheme (#278, @kelly-sovacool).
        • Also added a new parameter calculate_performance, which controls whether performance metrics are calculated (default: TRUE). Users may wish to skip performance calculations when training models with no cross-validation.
      • New parameter group_partitions added to run_ml() allows users to control which groups should go to which partition of the train/test split (#281, @kelly-sovacool).
        • -
      • Modified the training_frac parameter in run_ml() (#281, @kelly-sovacool). -
        • By default, training_frac is a fraction between 0 and 1 that specifies how much of the dataset should be used in the training fraction of the train/test split.
          • +
        • Modified the training_frac parameter in run_ml() (#281, @kelly-sovacool).
          • By default, training_frac is a fraction between 0 and 1 that specifies how much of the dataset should be used in the training fraction of the train/test split.
          • Users can instead give training_frac a vector of indices that correspond to which rows of the dataset should go in the training fraction of the train/test split. This gives users direct control over exactly which observations are in the training fraction if desired.

    mikropml 1.1.12021-09-14

    -
    +

    mikropml 1.1.02021-08-10

    -
    • New correlation method option for feature importance (#267, @courtneyarmour). -
      • The default is still “spearman”, and now you can use other methods supported by stats::cor with the corr_method parameter: get_feature_importance(corr_method = "pearson") -
        • -
      • +
    @@ -114,12 +104,10 @@

    mikropml 1.1.0mikropml 1.0.02021-05-13