# @author:Akash # @package:bayesiandnn from os import sys, path sys.path.append(path.dirname(path.dirname( path.abspath(__file__) ) ) ) import numpy as np import argparse from neuralnetwork.learning.sgd import * from dataio.pfileio import PfileIO from neuralnetwork.DNN import DNN from datasets import mnist from datasets import timit parser = argparse.ArgumentParser(description='setting up hyperparameters by command line ...') parser.add_argument('--percent', '-p', type=float, default=0.9999) parser.add_argument('--lr', '-l', type=float, default=0.08) args = parser.parse_args() percent = args.percent lr = args.lr mnist = mnist.load_mnist_theano('mnist.pkl.gz', percent_data=percent) # print mnist rng = np.random.RandomState(1111) nn = DNN(rng, [100], 784, 10) bsgd(nn, mnist, epochs=15, batch_size=20, lr=lr)
import numpy as np from neuralnetwork.learning.sgd import * from neuralnetwork.DNN import DNN from neuralnetwork.LadderAE import LadderAE from theano.tensor.shared_randomstreams import RandomStreams from datasets import mnist from datasets import timit mnist = mnist.load_mnist_theano('mnist.pkl.gz') print mnist train_set_x, train_set_y = mnist[0] valid_set_x, valid_set_y = mnist[1] test_set_x, test_set_y = mnist[2] numpy_rng = np.random.RandomState(1111) theano_rng = RandomStreams(numpy_rng.randint( 2**30 )) train_x_label = train_set_x[:20000,:] train_y_label = train_set_y[:20000] train_x_unlabel = train_set_x[20000:30000,:] LadderNetwork = LadderAE(rng, train_x_label, train_x_unlabel, train_y_label, [784, 500, 500, 100, 50, 10], ) # nn = DNN(rng, [3096, 3096], 784, 10) bsgd(nn, mnist, epochs=40)
from os import sys, path sys.path.append(path.dirname(path.dirname( path.abspath(__file__) ) ) ) import numpy as np from neuralnetwork.learning.sgd import * from neuralnetwork.DNN import DNN from neuralnetwork.SdA import SdA from datasets import mnist from datasets import timit from theano.tensor.shared_randomstreams import RandomStreams BATCH_SIZE = 700 NUM_EPOCHS = 40 mnist_data = mnist.load_mnist_theano('mnist.pkl.gz', percent_data=0.1) mnist_full = mnist.load_mnist_theano('mnist.pkl.gz', percent_data=1.0) print mnist # train_set_x, train_set_y = mnist_data[0] valid_set_x, valid_set_y = mnist_data[1] test_set_x, test_set_y = mnist_data[2] train_set_x, train_set_y = mnist_full[0] numpy_rng = np.random.RandomState(1111) theano_rng = RandomStreams(numpy_rng.randint( 2**30 )) # nn_ae = DNN(numpy_rng, [1024, 1024], 429, 144) # configuration for mnist