bowni commented 3 years ago

The fully flexible splitter enables the definition of a very precise split of a dataset among partners. One can indicate for each partner the % of data samples of each class that the partner should receive.

codecov-io commented 3 years ago

Codecov Report

Merging #320 (91f915d) into master (2e28840) will decrease coverage by 0.99%. The diff coverage is 17.07%.

@@            Coverage Diff             @@
##           master     #320      +/-   ##
==========================================
- Coverage   82.63%   81.64%   -1.00%     
==========================================
  Files          13       13              
  Lines        2637     2669      +32     
==========================================
  Hits         2179     2179              
- Misses        458      490      +32

Impacted Files	Coverage Δ
mplc/splitter.py	`81.53% <13.51%> (-14.76%)`	:arrow_down:
mplc/scenario.py	`82.60% <50.00%> (-0.48%)`	:arrow_down:
mplc/dataset.py	`85.83% <0.00%> (-0.21%)`	:arrow_down:

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 2e28840...c08e5e2. Read the comment docs.

bowni commented 3 years ago

What do you think of rewrite random and stratified on the top of this method ?

Agreed! But maybe not in this PR. It could be an enhancement (or refacto let's say) afterwards.

bowni commented 3 years ago

I will have only one question: can a flexible splitter be used with a config file ? Did you test it ?

Good point - just did it, enabled to catch an issue when the Scenario copies the Splitter 👍

bowni commented 3 years ago

@RomainGoussault could you check if your "requested changes" are now OK?

LabeliaLabs / distributed-learning-contributivity

Create a fully flexible splitter #320

Codecov Report