automl / auto-sklearn

Automated Machine Learning with scikit-learn
https://automl.github.io/auto-sklearn
BSD 3-Clause "New" or "Revised" License
7.6k stars 1.28k forks source link

Configuration for `SimpleClassificationPipelineTest.test_configurations_signed_data` gives undeterministic error #1310

Closed eddiebergman closed 2 years ago

eddiebergman commented 2 years ago
 Configuration:
  balancing:strategy, Value: 'weighting'
  classifier:__choice__, Value: 'lda'
  classifier:lda:shrinkage, Value: 'auto'
  classifier:lda:tol, Value: 0.038890093430048595
  data_preprocessor:__choice__, Value: 'feature_type'
  data_preprocessor:feature_type:categorical_transformer:categorical_encoding:__choice__, Value: 'encoding'
  data_preprocessor:feature_type:categorical_transformer:category_coalescence:__choice__, Value: 'minority_coalescer'
  data_preprocessor:feature_type:categorical_transformer:category_coalescence:minority_coalescer:minimum_fraction, Value: 0.001521146558163954
  data_preprocessor:feature_type:numerical_transformer:imputation:strategy, Value: 'mean'
  data_preprocessor:feature_type:numerical_transformer:rescaling:__choice__, Value: 'none'
  feature_preprocessor:__choice__, Value: 'fast_ica'
  feature_preprocessor:fast_ica:algorithm, Value: 'deflation'
  feature_preprocessor:fast_ica:fun, Value: 'exp'
  feature_preprocessor:fast_ica:whiten, Value: 'False'
eddiebergman commented 2 years ago

Just a note that this error seems to now occur once every test run. We need to identify which parameter causes the issue and add some constraint/modify the HP space.

This will likely change meta-data and require a benchmark test.

eddiebergman commented 2 years ago

Closing, documented in #1337