v20.2.0

Changelog:

• Performance improvements in the exareme dfl scripts.
• Compression of the db turned off in exareme.
• New defaultValue property in the algorithm parameters.
• Added the possibility to log algorithm messages to portainer.
• Python iterative algorithms can limit their iterations through max_iter.
• Refactored default outputs for sql algorithms to match with Front end requirments.
• Removed Kmeans, ID3, naivebayes standalone algorithms, they will be added in the next release.

v20.1.1

Changelog:

• Bug fix for Naive Bayes Training - Testing. Dataset property should be on every algorithm.

v20.1.0

Changelog:

• Galaxy algorithms added (Cross Validation, Hold Out Validation, Naive Bayes Training + Testing).
• Naive Bayes training also added as standalone.
• Tests added for these algorithms created from UoA.
• Fixes for dbIdentifier in exareme.

v20.02

Changelog:

• Fixes a bug, in Logistic Regression, where missing values were not dismissed + test for this case.
• Adds an iteration counter to main loop and terminates if counter exceeds MAX_ITER=20 in python iteration algorithms.

v20.0

Changes in

Exareme:
- Added functionality in exareme to allow integration of PYTHON algorithms. Types:
1) Local
2) Local Global
3) Multiple Local Global
4) Iterative
- Added privacy error handling for the algorithms.
- Added error handling for custom errors provided from the algorithms.
- Added extra error handling on the algorithm properties.
- Upon node startup the dataset csv is converted to a db file and then given to the algorithms as input.
- Metadata turned into a db file and added on the local database, that is provided to the algorithms.
- Formula type added in the algorithm properties.
- Refactoring on the exareme algorithm classes.
- Fix for filtering on the dataset value when providing the DB query.
- Fixes for race condition and throwing errors when running iterative algorithms.
- Fixes in the metadata file to match with AUEB.
- United the 4 different repos into one. (Mip-federation, Exareme-Docker, exareme, mip-algorithms)
- Bugfix for iterative algorithms not throwing errors.
- Bugfix for iterative algorithms not finishing properly.
- Bugfix for exception when calling an algorithm that doesn't exist.

Algorithms:
- Added Pearson Correlation + Tests added from CHUV.
- Added Anova algorithm + Tests added from CHUV.
- Added KMEANS algorithm + Tests created by us.
- Added ID3 algorithm + Tests created by us.
- Added Logistic Regression + Tests created by us.
- Refactored Linear Regression into a flattable approach + Tests added from CHUV.
- Refactored the Histograms algorithm into a flattable approach + Tests added from CHUV.
- Tests for privacy errors added.
- Pearson Correlation and Logistic Regression are returning highcharts in the new format decided.

v20.01

Changelog:

• iterations_condition_query_provided removed as a requirement in iterative algorithms
• Fixed bug in linear regression unit tests, where 2 tests had the same name
• Added error handling at the end of ANOVA, and ID3 to check if the maxnumberofiterations is enough.
• Rename type to sstype in ANOVA.
• Modify test of histograms algorithm

v19.0

Requires:

• mip-federation v1.1+
• Exareme-docker v2.0+

Changes:

• The datasets are being read from a local csv file and not from RAW.
• The composer is reading from flattable and not from eav model.
• Changes to add a DBIdentifier so we can pass a local DB instance as a parameter.
• The properties.json of the algorithms have new properties to be more dynamic.
• Basic error handling on the parameters of a request using the properties.json of an algorithm.
• Added QueryBuilder to parse the filters into sql queries.
• Refactoring on the Algorithms, Composer, and request handler classes.

v18.0

Merge pull request #87 from madgik/dev_release_v18

Build status fix