LSTM (Long Short Term Memory) networks

# Define LSTM 
class Lstm(nn.Module):
    
    def __init__(self, input_size, hidden_size=2, output_size=1, num_layers=1):
        super().__init__()
 
        self.layer1 = nn.LSTM(input_size, hidden_size, num_layers) 
        self.layer2 = nn.Linear(hidden_size, output_size)
 
    def forward(self, _x):
        x, _ = self.layer1(_x) 
        s, b, h = x.shape 
        
        x = x.view(s*b, h)
        x = self.layer2(x)
        x = x.view(s, b, -1)
        return x

# Generate data
n = 100
t = np.linspace(0,10.0*np.pi,n)
X = np.sin(t) 
X = X.astype('float32')
# Set window of past points for LSTM model
input_N = 5
output_N = 1
batch = 5
# Split into train/test data
last = int(n/2.5)
Xtrain = X[:-last]

Xtest = X[-last-input_N:]

# Store window number of points as a sequence
xin = []
next_X = []
for i in range(input_N,len(Xtrain)):
    xin.append(Xtrain[i-input_N:i])
    next_X.append(Xtrain[i])

# Reshape data to format for LSTM
xin, next_X = np.array(xin), np.array(next_X)
xin = xin.reshape(-1,1,input_N)


train_x = torch.from_numpy(xin)
train_y = torch.from_numpy(next_X)

train_x_tensor = train_x.reshape(-1,batch,input_N) # set batch size to 5
train_y_tensor = train_y.reshape(-1,batch,output_N) # set batch size to 5


model = Lstm(input_N, 5, output_N,1) # 5 hidden units
loss_function = nn.MSELoss()
optimizer = torch.optim.Adam(model.parameters(), lr=1e-2)

max_epochs = 1000
for epoch in range(max_epochs):
	output = model(train_x_tensor)
	loss = loss_function(output, train_y_tensor)
	
	loss.backward()
	optimizer.step()
	optimizer.zero_grad()
	if loss.item() < 1e-4:
		print('Epoch [{}/{}], Loss: {:.5f}'.format(epoch+1, max_epochs, loss.item()))
		print("The loss value is reached")
		break
	elif (epoch+1) % 100 == 0:
		print('Epoch: [{}/{}], Loss:{:.5f}'.format(epoch+1, max_epochs, loss.item()))
        
    # prediction on training dataset
predictive_y_for_training = model(train_x_tensor)
predictive_y_for_training = predictive_y_for_training.view(-1, 1).data.numpy()

# ----------------- plot -------------------
plt.figure()
plt.plot(train_y, label='Actual')
plt.plot(predictive_y_for_training, 'm--', label='Lstm')
plt.title('Train')
plt.legend(loc=1)


x_test_in = []
test_X =[]
for i in range(input_N,len(Xtest)):
    x_test_in.append(Xtrain[i-input_N:i])
    test_X.append(Xtrain[i])
    
x_test_in = np.array(x_test_in) 
test_x_tensor = x_test_in.reshape(-1, batch, input_N) # set batch size to 5
test_x_tensor = torch.from_numpy(test_x_tensor)
 
predictive_y_for_testing = model(test_x_tensor)
predictive_y_for_testing = predictive_y_for_testing.view(-1, 1).data.numpy()

plt.figure()
plt.plot(test_X,label = 'Actual')
plt.plot(predictive_y_for_testing,'r-.',label='Lstm')
plt.title('Test')
plt.legend(loc=0)
plt.show()

标签：Term,Short,tensor,train,Long,plt,test,input,size
来源： https://www.cnblogs.com/conpi/p/16114080.html