def train_regression_predictor(train_x, train_y, learning_rule='sgd', learning_rate=0.002, n_iter=20, units=4):
mlp = Regressor(layers=[Layer('Rectifier', units=units),
Layer('Linear')],
learning_rule=learning_rule,
learning_rate=learning_rate,
n_iter=n_iter)
mlp.fit(train_x, train_y)
print mlp.score(train_x, train_y)
return mlp
def CreateNetwork(data, predicates):
# входная размерность
dim_in = len(predicates)
# выходная размерность
dim_out = len(data[0]) - 1
# конфигурация сети
neural_network = Regressor(
layers=[
Layer("Rectifier", units=50),
Layer("Linear")],
learning_rate=0.001,
n_iter=5000)
# формирование обучающей выборки
x_train = np.array([CalcPredicates(row[0], predicates) for row in data])
y_train = np.array([apply(float, row[1:]) for row in data])
# обучение
logging.info('Start training')
logging.info('\n'+str(x_train))
logging.info('\n'+str(y_train))
try:
neural_network.fit(x_train, y_train)
except KeyboardInterrupt:
logging.info('User break')
pass
logging.info('Network created successfully')
logging.info('score = '+str(neural_network.score(x_train, y_train)))
# сохранение обученной сети
pickle.dump(neural_network, open(datetime.datetime.now().isoformat()+'.pkl', 'wb'))
return neural_network
print "Generate X_test and y_test" n_input = 11 print "X_test..." n_sample2 = np.asarray([[raw_data.ix[t-i][4] for i in range(1,n_input)] for t in np.arange (289*400,289*410)]) print "y_test..." n_test2 = np.asarray([raw_data.ix[t][4] for t in np.arange(289*400+1,289*410+1)]) # <codecell> print "Training time!!!!" nn.fit(X_training,y_training) # # # <codecell> # n_input = 11 n_sample2 = np.asarray([[raw_data.ix[t-i][4] for i in range(1,n_input)] for t in np.arange (289*400,289*410)]) n_test2 = np.asarray([raw_data.ix[t][4] for t in np.arange(289*400+1,289*410+1)]) print nn.score(n_sample2,n_test2) # # # <codecell> # # pred = np.asarray(nn.predict(n_sample2)) # ax = pl.subplot() # ax.set_color_cycle(['blue','red']) # pl.plot(n_test2) # pl.plot(pred) # pl.show() # # # <codecell> # # pd.DataFrame(zip(pred,n_test2),columns=["Prediction","Real"]) # # # <codecell>
#TODO: add autoencoder-pretrain
######################
# PREDICTION #
######################
y_predicted = nn.predict(X_test) # predict
#################
# EVALUATION # (to evaluate how well the REGRESSION did). For now we evaluate in the TRAINING DATA
#################
#TEST DATA
# R^2 measure
R2 = nn.score(X_test, y_test)
print('R^2_test= ',R2) #evaluating predictions with R^2
# My EVALUATION metric (mean cosine similarity)
cos = 0
for i in range(1,y_test.shape[0]):
#cos = cos + np.dot(np.array(y_predicted[i,]), np.array(y_test[i,]))/ (np.linalg.norm(np.array(y_test[i,])) * np.linalg.norm(np.array(y_predicted[i,])))
cos = cos + np.dot(y_predicted[i,], y_test[i,]) / (np.linalg.norm(y_test[i,]) * np.linalg.norm(y_predicted[i,]))
meanCos = cos/y_train.shape[0]
print('mean cos similarity_test= ', meanCos)
#TRAIN DATA:
y_predicted_train = nn.predict(X_train) # predict
# R^2 measure
R2_train = nn.score(X_train, y_train)
print('R^2_train= ',R2_train) #evaluating predictions with R^2
# My EVALUATION metric (mean cosine similarity)
