/Python/Neural network hyperparameter tuning

Neural network hyperparameter tuning

# random search cross validation in neural network model

from keras.models import Sequential
from keras.layers import Dense, Dropout
from keras.wrappers.scikit_learn import KerasClassifier

def report(results, n_top=3):
    for i in range(1, n_top + 1):
        candidates = numpy.flatnonzero(results['rank_test_score'] == i)
        for candidate in candidates:
            print("Model with rank: {0}".format(i))
            print("Mean validation score: {0:.3f} (std: {1:.3f})"
                  .format(results['mean_test_score'][candidate],
                          results['std_test_score'][candidate]))
            print("Parameters: {0}".format(results['params'][candidate]))
            print("")

def nn_model(activation = 'relu', neurons = 32, optimizer = 'Adam',dropout = 0.1, init_mode = 'uniform'):
    model = Sequential()
    model.add(Dense(32, input_dim = 32, kernel_initializer = init_mode, activation= activation))
    model.add(Dense((neurons*2)//3, kernel_initializer = init_mode,activation= activation))
    model.add(Dense((neurons*4)//9,kernel_initializer = init_mode,  activation = activation))
    model.add(Dropout(dropout))
    model.add(Dense(1, kernel_initializer = init_mode, activation='sigmoid'))
    model.compile(loss='binary_crossentropy', optimizer= optimizer, metrics=['accuracy'])
    return model

# Defining grid parameters
activation = ['softmax', 'softplus', 'softsign', 'relu', 'selu', 'elu', 'tanh','sigmoid', 'linear']
neurons = range(31,39)
dropout = [0.0, 0.1, 0.2, 0.3, 0.4, 0.5]
init_mode = ['uniform', 'lecun_uniform', 'normal', 'zero', 'glorot_normal', 'glorot_uniform', 'he_normal', 'he_uniform']
optimizer = ['SGD', 'Adam', 'Adamax','RMSprop','Adagrad','Adadelta','Nadam','Ftrl']
batch_size = range(10,101,10)
param_grid = dict(activation = activation, neurons = neurons, optimizer = optimizer, dropout = dropout, init_mode = init_mode, batch_size = batch_size)

clf = KerasClassifier(build_fn= nn_model, epochs= 10, verbose= 1)

model = RandomizedSearchCV(estimator= clf, param_distributions = param_grid, n_jobs=-1,verbose = 3)
model.fit(stan_x_train,stan_y_train)

report(model.cv_results_)

Similar codes

0 vote345 views1 answer
Two labels on one axis python
Python
-1 vote307 views1 answer
Django model UUID field
Python
17 votes327 views3 answers
Create log in python
Python
0 vote285 views1 answer
Mad python
Python
0 vote310 views1 answer
Python extract email attachment
Python
0 vote382 views1 answer
Size of variable
Python
87 votes311 views5 answers
Python update pip3
Python
10 votes328 views6 answers
Pandas if else
Python
8 votes275 views1 answer
String startswith python
Python
1 vote333 views1 answer
Torch save state dict
Python