param_grid = {
'epochs': [42],
'batch_size': [50],
'optimizer': optimizer_list,
'n1': [32, 36, 40, 42, 46],
'n2': [10, 15, 20, 25, 30],
'activation': activation
} #'n1':range(37,40), 'n2':range(20,45,5)}
grid_search = GridSearchCV(estimator=model,
param_grid=param_grid,
n_jobs=-1,
cv=3,
scoring=['accuracy'],
verbose=10)
grid_result = grid_search.fit(X_train, y_train)
report(grid_result.cv_results_, n_top=10)
print("GridSearchCV took %.2f seconds." % (time.time() - time1))
random_search = False
if random_search:
time1 = time.time()
# specify parameters and distributions to sample from
param_dist = {
"optimizer":
['SGD', 'RMSprop', 'Adagrad', 'Adadelta', 'Adam', 'Adamax', 'Nadam'],
"activation":
['softmax', 'softplus', 'softsign', 'tanh', 'sigmoid', 'hard_sigmoid'],
"n1":
randint(1, 50),
"n2":
randint(5, 50),
# specify parameters and distributions to sample from
param_dist = {"n1": sp_randint(15, 80),
"n2": sp_randint(10, 80),
"n3": sp_randint(10, 80),
"epochs": sp_randint(30, 60),
"batch_size": sp_randint(20, 100),
"optimizer":['rmsprop', 'nadam', 'adagrad'],
"activation": ['softmax', 'sigmoid', 'softplus']
}
n_iter_search = 100
random_search = RandomizedSearchCV(model, param_distributions = param_dist, n_iter=n_iter_search, cv = 3, scoring=scorer, verbose=10)
random_search.fit(X_train, y_train)
report(random_search.cv_results_)
#scores = random_search.cv_results_['mean_test_score']
if useGridCV:
# grid setup
optimizers = ['adam']
activations = ['softplus', 'sigmoid'] #['sigmoid', 'softmax', 'softplus']
#inits = ['glorot_uniform', 'normal', 'uniform']
epochs = [15] #range(10, 100, 20)
batches = [33] #range(50, 500, 50)
n1s = [73, 75, 77]
n2s = [12, 14, 16]
n3s = [14, 16, 18] #3 x 3 x 3 x 2 = 54