I have 10000 tuples of numbers (x1, x2, y) generated from the equation: y = np.cos(0.583 * x1) + np.exp(0.112 * x2). I want to use a neural network, trained with gradient descent, in PyTorch, to find the 2 parameters, i.e. 0.583 and 0.112

Here is my code:

class NN_test(nn.Module):
    def __init__(self):
        self.a = torch.nn.Parameter(torch.tensor(0.7))
        self.b = torch.nn.Parameter(torch.tensor(0.02))

    def forward(self, x):
        y = torch.cos(self.a*x[:,0])+torch.exp(self.b*x[:,1])
        return y

model = NN_test().cuda()

lrs = 1e-4
optimizer = optim.SGD(model.parameters(), lr = lrs)
loss = nn.MSELoss()

epochs = 30
for epoch in range(epochs):
    for i, dtt in enumerate(my_dataloader):

        inp = dtt[0].float().cuda()
        output = dtt[1].float().cuda()

        ls = loss(model(inp),output)

    if epoch%1==0:
        print("Epoch: " + str(epoch), "Loss Training: " + str(ls.data.cpu().numpy()))

where x contains the 2 numbers x1 and x2. In theory, it should work easily, but the loss doesn't go down. What am I doing wrong?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.