Adding MSE and Cross entropy loss


I have one question about the regression and classification loss. I have trained a network which outputs denoised data and its corresponding events picked in binary(0 or 1). At the moment I am using only MSE loss and then set some threshold criteria for binary data, to convert to 0 or 1.

Is it possible that I have two types of losses(one classification and other regression) added and backpropagated for this type of network? If not, what are the other alternatives?


Yes, you can calculate and sum different losses, which can then be used to compute the gradients and train the model.

@ptrblck . Thank you for the helpful response.