def model(eta, batch_size):
filename = str.format("LR_BS_Trial\\loss_{0}_{1}.pkl", eta,
batch_size).replace('.', '', 1)
filepath = Path(filename)
if filepath.exists():
return filename
dataReader = MnistImageDataReader(mode="vector")
dataReader.ReadData()
dataReader.NormalizeX()
dataReader.NormalizeY(NetType.MultipleClassifier, base=0)
dataReader.Shuffle()
dataReader.GenerateValidationSet(k=12)
n_input = dataReader.num_feature
n_hidden1 = 64
n_hidden2 = 16
n_output = dataReader.num_category
eps = 0.01
max_epoch = 30
hp = HyperParameters_3_0(n_input, n_hidden1, n_hidden2, n_output, eta,
max_epoch, batch_size, eps,
NetType.MultipleClassifier, InitialMethod.Xavier)
net = NeuralNet_3_0(hp, "LR_BS_Trial")
net.train(dataReader, 0.5, True)
net.DumpLossHistory(filename)
return filename
def load_data():
dataReader = MnistImageDataReader(mode="vector")
dataReader.ReadData()
dataReader.NormalizeX()
dataReader.NormalizeY(NetType.MultipleClassifier, base=0)
dataReader.Shuffle()
dataReader.GenerateValidationSet(k=12)
x_train, y_train = dataReader.XTrain, dataReader.YTrain
x_test, y_test = dataReader.XTest, dataReader.YTest
x_val, y_val = dataReader.XDev, dataReader.YDev
x_train = x_train.reshape(x_train.shape[0], 28 * 28)
x_test = x_test.reshape(x_test.shape[0], 28 * 28)
x_val = x_val.reshape(x_val.shape[0], 28 * 28)
return x_train, y_train, x_test, y_test, x_val, y_val
# Copyright (c) Microsoft. All rights reserved.
# Licensed under the MIT license. See LICENSE file in the project root for full license information.
from HelperClass2.MnistImageDataReader import *
from HelperClass2.NeuralNet_3_0 import *
if __name__ == '__main__':
dataReader = MnistImageDataReader(mode="vector")
dataReader.ReadData()
dataReader.NormalizeX()
dataReader.NormalizeY(NetType.MultipleClassifier, base=0)
dataReader.Shuffle()
dataReader.GenerateValidationSet(k=12)
n_input = dataReader.num_feature
n_hidden1 = 64
n_hidden2 = 16
n_output = dataReader.num_category
eta = 0.2
eps = 0.01
batch_size = 128
max_epoch = 40
hp = HyperParameters_3_0(n_input, n_hidden1, n_hidden2, n_output, eta,
max_epoch, batch_size, eps,
NetType.MultipleClassifier, InitialMethod.Xavier)
net = NeuralNet_3_0(hp, "MNIST_64_16")
net.train(dataReader, 0.5, True)
net.ShowTrainingHistory(xline="iteration")
print(m.weight)
def metric(pred, label):
'''
:param pred: batch_size * num_classes, numpy array
:param label: [batch_size,]
:return: accuracy
'''
real_len = label.shape[0]
pred_y = np.argmax(pred, axis=1)
return sum(label == pred_y) / real_len
if __name__ == '__main__':
# reading data
dataReader = MnistImageDataReader(mode="vector")
dataReader.ReadData()
dataReader.NormalizeX()
dataReader.NormalizeY(NetType.MultipleClassifier, base=0)
dataReader.Shuffle()
dataReader.GenerateValidationSet(k=12)
# data preprocessing
dataReader.XTrain = np.reshape(dataReader.XTrain, [-1, 28 * 28])
dataReader.YTrain = np.argmax(dataReader.YTrain, axis=1)
dataReader.XDev = np.reshape(dataReader.XDev, [-1, 28 * 28])
dataReader.YDev = np.argmax(dataReader.YDev, axis=1)
max_epoch = 500 # max_epoch
batch_size = 64 # batch size
lr = 1e-4 # learning rate