NMF n_components are not properly getting reflected in output when using Grid Search CV · scikit-learn/scikit-learn#10329

(25 留言) (0 反應) (0 負責人)Python (66,084 star) (27,020 fork)batch import

Bughelp wantedmodule:decompositionmodule:pipeline

描述

When I executed the example code from this link and analyzed the grid search output (grid.cv_results_['params']), n_components are not properly getting reflected in output.

Posting a small snippet of output of grid.cv_results_['params']:

{'classify__C': 1000,
  'reduce_dim': NMF(alpha=0.0, beta_loss='frobenius', init=None, l1_ratio=0.0, max_iter=200,
    **n_components=None,** random_state=None, shuffle=False, solver='cd',
    tol=0.0001, verbose=0),
  'reduce_dim__n_components': 2},
 {'classify__C': 1000,
  'reduce_dim': NMF(alpha=0.0, beta_loss='frobenius', init=None, l1_ratio=0.0, max_iter=200,
    **n_components=None**, random_state=None, shuffle=False, solver='cd',
    tol=0.0001, verbose=0),
  'reduce_dim__n_components': 4},
 {'classify__C': 1000,
  'reduce_dim': NMF(alpha=0.0, beta_loss='frobenius', init=None, l1_ratio=0.0, max_iter=200,
    **n_components=None**, random_state=None, shuffle=False, solver='cd',
    tol=0.0001, verbose=0),
  'reduce_dim__n_components': 8},

where reduce_dim__n_components are updating for NMF but not the actual n_components in NMF

Thanks, Pat

貢獻者指南

技術棧: python
領域: datamachine learning
議題類型: bug
難度: 3
預計時間: 1-2 days
活動狀態: stale
清晰度: clear
前置要求: scikit learn basicsGridSearchCV usageNMF estimator
新手友善度: 45
研究方向: Examine the NMF class's init method to see how n components is stored (likely as self.n components = n components). Look at GridSearchCV's fit method to understand how parameters are assigned to the estimator. Reproduce the bug using the example from the issue. Review the 25 comments for any proposed fixes or discussions about the root cause, such as the parameter not being set due to the way GridSearchCV handles estimator cloning.