model = random.choice(models)
architecture, step_size, termination, dropout = model
models.remove(model)
current_params = {'layer_ns': architecture, 'step_size': step_size, 'termination': termination, 'batch_size': batch_size, 'max_example': max_example, 'dropout': dropout}
print "##########################################################################"
print "Model #%d" % remaining_models
print current_params
print "##########################################################################"
start = time.time()
scores = []
cv = 1
cv_start = time.time()
for train_index, test_index in KFold(train_inputs[:max_example].shape[0], n_folds=3):
X_train, X_test = train_inputs[train_index], train_inputs[test_index]
y_train, y_test = train_outputs[train_index], train_outputs[test_index]
ffnn = FFNN(X_train[:batch_size], architecture, step_size, termination, dropout=dropout)
ffnn.fit(X_train, y_train, X_test, y_test, batch_size=batch_size)
score = ffnn.score(X_test, y_test)
print "CV " + str(cv) + " score: " + str(score)
print "CV " + str(cv) + " time: " + str(time.time() - cv_start)
scores += [score]
cv += 1
cv_start = time.time()
score_avg = sum(scores)/len(scores)
current_params['score_avg'] = score_avg
current_params['train_time'] = str((time.time() - start))
print "Score average: " + str(score_avg)
if score_avg > best_score:
print "NEW MAX SCORE: " + str(score_avg)
best_score = score_avg
best_params = current_params
import numpy as np from ffnn.ffnn import FFNN X = np.asarray([[0,0],[0,1],[1,0],[1,1]]) y = np.asarray([[1,0],[0,1],[0,1],[1,0]]) ffnn = FFNN(X[0],[400,200,100,10,2],.00000000000000001,.01) ffnn.fit(X,y,X,y,batch_size=1)