def DigitRecognitionTest():
X_train, y_train, X_val, y_val, X_test, y_test = load_and_extract_mnist_data.load_dataset()
Y_train = []
for e in y_train:
tmp = [0]*10
tmp[e]=1
Y_train.append(tmp)
Y_train = np.array(Y_train)
Y_val = []
for e in y_val:
tmp = [0]*10
tmp[e]=1
Y_val.append(tmp)
Y_val = np.array(Y_val)
Y_test = []
for e in y_test:
tmp = [0]*10
tmp[e]=1
Y_test.append(tmp)
Y_test = np.array(Y_test)
myann = LeNet5()
myann.setTrain(X_train,Y_train)
myann.setCVD(X_val,Y_val)
myann.setTest(X_test,Y_test)
#myann.MiniBatch_Train(64,1000,ifshow=True,gradiantCheck=False) #use cross validation data to test
myann.Final_Train_and_Evaluate(64,1000,ifshow=True) #use test data to test
def DigitRecognitionTest():
X_train, y_train, X_val, y_val, X_test, y_test = load_and_extract_mnist_data.load_dataset()
X_train = X_train.reshape(np.size(X_train,0),28*28).transpose()
X_val = X_val.reshape(np.size(X_val,0),28*28).transpose()
X_test = X_test.reshape(np.size(X_test,0),28*28).transpose()
Y_train = []
for e in y_train:
tmp = [0]*10
tmp[e]=1
Y_train.append(tmp)
Y_train = np.array(Y_train).transpose()
Y_val = []
for e in y_val:
tmp = [0]*10
tmp[e]=1
Y_val.append(tmp)
Y_val = np.array(Y_val).transpose()
Y_test = []
for e in y_test:
tmp = [0]*10
tmp[e]=1
Y_test.append(tmp)
Y_test = np.array(Y_test).transpose()
myann = NeuralNetwork('Digital Recognition')
myann.model.SetInputLayer(28*28)
myann.model.SetOutputLayer('SoftMax')
myann.model.SetFullConnectLayer([(50,'ReLu'),(20,'ReLu'),(20,'ReLu'),(10,None)])
myann.setTrain(X_train,Y_train)
myann.setCVD(X_val,Y_val)
myann.setTest(X_test,Y_test)
myann.MiniBatch_Train(1000,2500,True)
def DigitRecognitionTest():
X_train, y_train, X_val, y_val, X_test, y_test = load_and_extract_mnist_data.load_dataset(
)
Y_train = []
for e in y_train:
tmp = [0] * 10
tmp[e] = 1
Y_train.append(tmp)
Y_train = np.array(Y_train)
Y_val = []
for e in y_val:
tmp = [0] * 10
tmp[e] = 1
Y_val.append(tmp)
Y_val = np.array(Y_val)
Y_test = []
for e in y_test:
tmp = [0] * 10
tmp[e] = 1
Y_test.append(tmp)
Y_test = np.array(Y_test)
myann = LeNet5()
myann.setTrain(X_train, Y_train)
myann.setCVD(X_val, Y_val)
myann.setTest(X_test, Y_test)
#myann.MiniBatch_Train(64,1000,ifshow=True,gradiantCheck=False) #use cross validation data to test
myann.Final_Train_and_Evaluate(64, 1000,
ifshow=True) #use test data to test
def main_data():
X_train, y_train, X_val, y_val, X_test, y_test = im.load_dataset()
m_Xtrain = my_data(X_train) ##### (50000,785)
m_Xval = my_data(X_val) ##### (10000, 785)
m_Xtest = my_data(X_test) #### (10000, 785)
mm_xtrain = m_Xtrain[0:5000, :]
mm_ytrain = y_train[0:5000]
mm_xval = m_Xval[0:1000, :]
mm_yval = y_val[0:1000]
train_y = y_lab(mm_ytrain)
val_y = y_lab(mm_yval)
return mm_xtrain, train_y, mm_xval, val_y
"type": "relu"
},
{
"type": "fc",
"neurons": n_classes
},
{
"type": "relu"
},
{
"type": "softmax",
"categories": n_classes
}
])
X_train, y_train, X_val, y_val, X_test, y_test = im.load_dataset()
mm_xtrain, mm_ytrain, mm_xval, mm_yval = proce.main_data()
"""
train_data, test_data, train_target, test_target = train_test_split(mm_xtrain[0:3000,], mm_ytrain[0:3000], train_size=0.7)
train_data = train_data.reshape((len(train_data),1, 28, 28))
test_data = test_data.reshape((len(test_data), 1, 28, 28))
"""
train_data = X_train
test_data = X_val
train_target = mm_ytrain
test_target = mm_yval
#print test_data.shape, test_target.shape
net.train(train_data, train_target, n_epochs=1000, batch_size=64)
"""
###Parallel python cpu