use_validation = False nb_input = 2 nb_hidden = 10 nb_labels = 2 nb_iters = 10000 nb_epochs = 1 alpha = 3.0 lbda = 0.0 momentum = 0.0 precision = 1e-6 timer = time.clock if (sys.platform == 'win32') else time.time Plots.scatterplot(X, Y) for nb_hidden in range(1, 7): nn = NeuralNetwork(nb_input, nb_hidden, nb_labels, cost_function='cross_entropy') a,b,c,d = [],[],[],[] for i in range(nb_epochs): X_train, Y_train = X, Y Y_vec = Utils.vectorize_output(Y_train, nb_labels) print('Epoch %d' % (i+1)) print("-"*30) nn.initialize_weights()
return self.train_binary(X, Y, alpha, lbda, precision, nb_iters, verbose) if __name__ == '__main__': # X, Y = Reader.read_data('dados/xor.txt', ignore_line_number=False) # data = {'X':X, 'Y':Y} # Reader.save_mat('dados/xor.mat', data) mat = Reader.load_mat('dados/xor.mat') X_orig, Y = np.matrix(mat['X']), mat['Y'] # X, mu, sigma = Features.normalize(X) Plots.scatterplot(X_orig, Y) pyplot.show() max_degree = 60 iters, times, accs, alphas = [], [], [], [] result_times = open("reg_result_times.txt", "w") result_times.write("alpha tempo\n") result_iters = open("reg_result_iters.txt", "w") result_iters.write("alpha iters\n") result_accs = open("reg_result_accs.txt", "w") result_accs.write("alpha acc\n") result_alphas = open("reg_result_alphas.txt", "w")
if __name__ == '__main__': # X, Y = Reader.read_data('dados/xor.txt', ignore_line_number=False) # data = {'X':X, 'Y':Y} # Reader.save_mat('dados/xor.mat', data) mat = Reader.load_mat('dados/xor.mat') X_orig, Y = np.matrix(mat['X']), mat['Y'] # X, mu, sigma = Features.normalize(X) Plots.scatterplot(X_orig, Y) pyplot.show() max_degree = 60 iters, times, accs, alphas = [], [], [], [] result_times = open("reg_result_times.txt", "w") result_times.write("alpha tempo\n") result_iters = open("reg_result_iters.txt", "w") result_iters.write("alpha iters\n") result_accs = open("reg_result_accs.txt", "w") result_accs.write("alpha acc\n")
use_validation = False nb_input = 2 nb_hidden = 10 nb_labels = 2 nb_iters = 10000 nb_epochs = 1 alpha = 3.0 lbda = 0.0 momentum = 0.0 precision = 1e-6 timer = time.clock if (sys.platform == 'win32') else time.time Plots.scatterplot(X, Y) for nb_hidden in range(1, 7): nn = NeuralNetwork(nb_input, nb_hidden, nb_labels, cost_function='cross_entropy') a, b, c, d = [], [], [], [] for i in range(nb_epochs): X_train, Y_train = X, Y Y_vec = Utils.vectorize_output(Y_train, nb_labels) print('Epoch %d' % (i + 1))