The error function most suited for gradient descent using logistic regression is
The entropy function.
The squared error.
The cross-entropy function.
The number of mistakes.