CODEWITHSUNDEEP
pythonscikit-learndatasetgridsearchcv
PtfDigital
PtfDigital

Reputation: 11

Scikit-Learn TypeError: Invalid parameter activation for estimator Pipeline

I was trying to use scikit-learn package 0.24.2 with python-3.8.5 to do a grid search but I get the following error:

Error : Invalid parameter activation for estimator Pipeline(steps=[('scaler', StandardScaler()), ('MLPRegressor', MLPRegressor())]). Check the list of available parameters with estimator.get_params().keys().

My code is the following:

mpl_pipe = Pipeline(steps=[
    ('scaler',  StandardScaler()),
    ('MLPRegressor', MLPRegressor())
])

parameters = {
    'hidden_layer_sizes': [(900,700,500,300,150,) ,(700,500,300,150, ),(500,300,150, ),(300,150, )],
    'max_iter': [20000], 
    'activation' : ['identity', 'logistic', 'tanh', 'relu'],
    'solver' : ['lbfgs', 'sgd', 'adam']
}

mlp_grid = GridSearchCV(mpl_pipe,
                        parameters,
                        cv = 5,
                        n_jobs = -1,
                        verbose=True)

mlp_grid.fit(x_train, y_train)

print(mlp_grid.best_score_)
print(mlp_grid.best_params_)

Upvotes: 1

Views: 1222

Answers (2)

user11989081
user11989081

Reputation: 8663

The pipeline parameters can be accessed as pipeline step name + __ + parameter name, which means that you will need to add MLPRegressor__ before each of the parameter names in your search grid.

from sklearn.pipeline import Pipeline
from sklearn.datasets import make_regression
from sklearn.preprocessing import StandardScaler
from sklearn.neural_network import MLPRegressor
from sklearn.model_selection import GridSearchCV

X, y = make_regression(n_samples=200, n_targets=1, n_features=5, n_informative=5)

pipe = Pipeline(steps=[
    ('scaler', StandardScaler()),
    ('MLPRegressor', MLPRegressor())
])

param_grid = {
    'MLPRegressor__hidden_layer_sizes': [(900, 700, 500, 300, 150,), (700, 500, 300, 150, ), (500, 300, 150,), (300, 150,)],
    'MLPRegressor__max_iter': [1000],
    'MLPRegressor__activation': ['identity', 'tanh', 'relu'],
    'MLPRegressor__solver': ['adam']
}

search = GridSearchCV(pipe, param_grid, cv=5, n_jobs=-1, verbose=True)
search.fit(X, y)

print(search.best_score_)
print(search.best_params_)

Upvotes: 2

lpounng
lpounng

Reputation: 640

I think the problem is that the parameters e.g. activation are only for MLPRegressor().

You may specify which estimator the parameters are for, with say {'MLPRegressor__activation': ['identity', 'logistic', 'tanh', 'relu'], ...}

Reference: https://scikit-learn.org/stable/modules/compose.html#nested-parameters

Upvotes: 1

Related Questions