Fit Error OneVsOneClassifier, OneVsRestClassifier Logistic Regression for digits recognition multilabel classifier model fit error

Question

Problem

New error on problem:

UserWarning: X has feature names, but LogisticRegression was fitted without feature names
  warnings.warn(UserWarning: X has feature names, but LogisticRegression was fitted without feature names

Goal is to use RandomisedSearchCV for tune parameter, then fit two models, one each for OneVsOneClassifier and OneVsRestClassifier. Then check accuracy performance of each models using tuned parameter from RandomizedSearchCV. Define two models to fit, to fit on digit recognition MNIST dataset for multi-label classification prediction.

I setup a tune hyper=parameters using GridSearchCV, just simple for estimator__C': [0.1, 1, 100, 200] for LogisticRegression. For audit, I print the computed grid parameters. Provide a scaled X-train object to the fit model. Then run the fit model.

Problem was running on Kaggle GPU P100. When I execute code: ovr_grid_search.fit() & ovo_grid_search.fit(), finish running and next step error is this by Adding verbose=1 and error_score="raise" to RandomainedSearchCV classifier, and determined that needed to relocate StandardScaler with MinMaxScaler

Error Error is LogisticRegression fit with out features.

/Users/matthew/opt/anaconda3/lib/python3.9/site-packages/sklearn/base.py:443: 
UserWarning: X has feature names, 
but LogisticRegression was fitted without feature names

Code

from sklearn.linear_model import LogisticRegression
from sklearn.multiclass import OneVsOneClassifier, OneVsRestClassifier
from sklearn.model_selection import GridSearchCV
from sklearn.preprocessing import StandardScaler

scaler = StandardScaler()
X_train_scaled = scaler.fit_transform(X_train.astype(np.float64))

ovr_model = OneVsRestClassifier(LogisticRegression())
ovo_model = OneVsOneClassifier(LogisticRegression())

param_grid = {
    'estimator__C': [0.1, 1, 100, 200]
}

ovr_grid_param = RandomizedSearchCV(ovr_model, param_grid, cv=5, n_jobs=8)
ovo_grid_param = RandomizedSearchCV(ovo_model, param_grid, cv=5, n_jobs=8)

print("OneVsRestClassifier best params: ", ovr_grid_param)
print("OneVsOneClassifier best params: ", ovo_grid_param)

min_max_scaler = preprocessing.MinMaxScaler()
X_train_scaled = min_max_scaler.fit_transform(X_train)

### below code is the problem area **

ovr_grid_param.fit(X_train_scaled, y_train)
ovo_grid_param.fit(X_train_scaled, y_train)

Data The digit recognition MNIST dataset. X_train scaled data

array([[0., 0., 0., ..., 0., 0., 0.],
       [0., 0., 0., ..., 0., 0., 0.],
       [0., 0., 0., ..., 0., 0., 0.],
       ...,
       [0., 0., 0., ..., 0., 0., 0.],
       [0., 0., 0., ..., 0., 0., 0.],
       [0., 0., 0., ..., 0., 0., 0.]])

y_train data

33092    5
30563    2
17064    4
16679    9
30712    0
30177    0
11735    3
1785     8
4382     3
21702    7
37516    3
9476     6
4893     5
22117    0
12646    8

RandomisedSearch execution results

OneVsRestClassifier best params:  RandomizedSearchCV(cv=5, error_score='raise',
                   estimator=OneVsRestClassifier(estimator=LogisticRegression()),
                   n_jobs=3,
                   param_distributions={'estimator__C': [0.1, 1, 100, 200],
                                        'estimator__max_iter': [2500, 4500,
                                                                6500, 9500,
                                                                14000]})
OneVsOneClassifier best params:  RandomizedSearchCV(cv=5, error_score='raise',
                   estimator=OneVsOneClassifier(estimator=LogisticRegression()),
                   n_jobs=3,
                   param_distributions={'estimator__C': [0.1, 1, 100, 200],
                                        'estimator__max_iter': [2500, 4500,
                                                                6500, 9500,
                                                                14000]})

Answer 1

I am unaware of sklearn GPU support https://scikit-learn.org/stable/faq.html#will-you-add-gpu-support

Is there a way to run ‘nvidia-smi’? You can check if there is a process active on P100. Otherwise, it may be just be taking too long because the grid search space is too large. Try increasing the n_jobs parameter to take advantage of more cpus.