Open MorrisNein opened 7 months ago
Some models appear to behave differently in different environments.
For example, a pipeline consisting of multiple logistic regressions gives different outputs for the same data:
https://github.com/aimclub/FEDOT/actions/runs/7086728823/job/19285547590?pr=1214
This does not allow testing binary classification values (in the test test_metrics). The source of the stochastic behaviour should be eliminated, at least for the test.
test_metrics
Some models appear to behave differently in different environments.
For example, a pipeline consisting of multiple logistic regressions gives different outputs for the same data:
https://github.com/aimclub/FEDOT/actions/runs/7086728823/job/19285547590?pr=1214
This does not allow testing binary classification values (in the test
test_metrics
). The source of the stochastic behaviour should be eliminated, at least for the test.