Neural network regression of car-prices

We assume that the car prices (y) can be predicted as ?=?(?) where ?=[?0..?9..] are the car features and ? is a three layer fully connected neural network. We first create the matrix X_train, whose columns are the normalized training features.

Create a three layer neural network with hidden layers of h1 and h2 neurons respectively. Use RELU activations at the hidden layers and no activation on the output layer.

class Net(nn.Module):
    def __init__(self,Nfeatures,Noutput,Nh1=10,Nh2=10):
        super(Net, self).__init__()
        
        # YOUR CODE HERE

    def forward(self, x):
        
        # YOUR CODE HERE

featureScale = np.max(np.abs(X_train),axis=0,keepdims=True)
X_train_T = X_train/featureScale
X_train_T = torch.tensor(X_train_T.astype(np.float32))

ymax = np.max(y_train,axis=0,keepdims=True)
y_train_T = torch.tensor(y_train.astype(np.float32)/ymax).unsqueeze(1)

Training

YOUR CODE BELOW 

1. Define a network with 20 features/neurons in hidden layer 1 and 25 in layer 2
2. Define optimizer to be SGD
3. Define the loss as the norm of the error
4. Train the network for 10,000 epochs with a learning rate of 1e-3

net = Net(16,1,50,50)
out = net(X_train_T)

# YOUR CODE HERE

for epoch in range(10000):
    
    #YOUR CODE HERE

    if(np.mod(epoch,1000)==0):
      print("Error =",error.detach().cpu().item())
      fig,ax = plt.subplots(1,1,figsize=(12,4))
      ax.plot(y_train_T.abs().detach().cpu(),label='Actual Price')
      ax.plot(out.abs().detach().cpu(),label='Prediction')
      ax.legend()
      plt.show()