How to increase a loss

autograd
1

Hi,

I want to increase the specific loss while learning the model.

I have the following code:

criterion_MSE = nn.MSELoss()
criterion_CE = nn.CrossEntropyLoss()

...
loss_MSE = criterion_MSE(predict1, target1)

...
loss_CE = criterion_CE(predict2, target2)

total_loss = loss_MSE + loss_CE

total_loss.backward()
optimizer.step()

If I want to maximize the loss_CE here (with minimize loss_MSE), just multiply the loss_CE by negative one?
If not, please tell me how I can do it.

2

Yes, remember that you are solving a optimization problem, so max f(x) <=> min -f(x)