If we work with regression problem and we have sigmoid function in the hidden layers but the output layer is linear function, what the best cost function to deal with this problem , Mean Squared Error cost function or Logistic cost function ?

Thanks

The cost function is determined by the output layer, or by the optimizer settings.

what do you mean by optimizer settings ?

For example in TensorFlow, if specifying crossentropy or from_logits.

1 Like

I got it , many thanks