What loss function should I use?

#1

Hello everyone,

I’m doing a task that has targets within a range of 0 and 1 (like probabilities).
I thought about Binary Cross-Entropy, but the task is not the binary classification problem. So from my point of view, the BCE loss may not be fit.
I have tried the MSE loss function but the results were not good.

Can you please suggest me other loss functions? I’d like to try them.

Thank you so much.

#2

You have a regression problem. But not logistic regression because logistic regression is a classification problem as well. So you need some loss functions for real regression problem. L1, L2 respectable called MAE, and MSE for instance. If you tried MSE, maybe you should try more. :slight_smile:

1 Like