# Cost

# Mean Square Error

Average of all the squared errors. Used for problems where the output is not confined to specific classes like a regression problem.

# Root Mean Square

Square Root of Average of all the squared errors. Used for problems where the output is not confined to specific classes like a regression problem.

# Sigmoid Cross Entropy

Cross entropy loss with sigmoid activation applied on the logits before the loss calculation.

# Softmax Cross Entropy

Cross entropy loss with softmax activation applied on the logits before the loss calculation.

# weighted Cross Entropy

Cross entropy loss with sigmoid activation and also each class if weighted by a scale during the loss calculation to give priority to one class over another.