Not able to reach the same visualization of MLP as the problem set
With : learning rate = 0,01; hidden_size = 2; num_epoch =300; loss function = MSE; 2 linear layers; 1 non linear (ReLU or Sigmoid). I don't find a solution that looks like the problem set figure. I tried to play with the hyperparameters, yet can't see any progress. Could it be the loss function?
Not able to reach the same visualization of MLP as the problem set
With : learning rate = 0,01; hidden_size = 2; num_epoch =300; loss function = MSE; 2 linear layers; 1 non linear (ReLU or Sigmoid). I don't find a solution that looks like the problem set figure. I tried to play with the hyperparameters, yet can't see any progress. Could it be the loss function?
Is this figure with sigmoid or ReLU activation ? If its with ReLU is seems coherent.
1
Yeah, it was with ReLU.
And this is with Sigmoid:
I can't reach the sigmoid curves like on the figure of the exercise. (Same parameters as above)
Add comment