Releases: madgik/exareme
Releases · madgik/exareme
23.4.3
Changelog
Exareme:
- (Bug 359 https://redmine.hbpmip.link/issues/359 ) Fixed exareme hanging when a node fails.
23.4.2
24.0.1
24.0.0
Changelog
Exareme:
- (Improvement) Improved Health Check on the workers to check that master is pingable.
- (Improvement) Reduced max exareme threads on all the thread pools to make it lighter.
- (Bugfix) Many smaller stabilization fixes.
Mip-Algorithms:
- (Improvement) Implementation of Oneway Anova algorithm for one categorical and one numerical variable
- Output of Anova includes:
- Anova summary table
- Tukey Honest Significant Differences test table
- Mean plot for every level in the categorical variable
- 95 Algorithm tests (80 local + 15 federated on 10 nodes)
- 5 Exareme integration tests
- 1 Privacy test
- Output of Anova includes:
- (Improvement) Implementation of Naive Bayes classifier for mixed numerical and categorical variables
- Output includes:
- Classification report (precision, recall, specificity, F score)
- ROC report (AUC per class, Gini coefficient per class)
- Accuracy report (accuracy, upper and lower c.i.)
- Confusion matrix chart
- ROC per class chart
- Output includes:
- (Improvement) CART visualization addition.
23.5.1
23.5.0
Changelog
Exareme:
- (Improvement) Added more exceptions to the bad user input exception category.
- (Improvement) Exareme reponse codes are 200 on success/user-error/warning, 400 on bad algorithm name and 500 for internal server error.
MIP-Algorithms:
- (Bugfix) Add error checking for input variables: modulo 2
- (Bugfix issue324) The user inputs the levels he wishes to assign to positive and negative outcomes. This also eliminates the need to use filters to handle multilevel variables. Only two levels are always filter from within the algorithm now (method: keep_levels).
- (Bugfix) Add method for aligning categorical variables across local nodes. Checks all categorical columns in design matrices produced by patsy and if any level is missing in some local node it adds the corresponding column (all zeros since the level was missing originally).
- (Improvement) Add more nodes to LogisticRegression and CalibrationBelt tests.