Sigmoid function g$($x$)$ is often used as the activation function in neural networks. Which of the following statements is NOT true regarding sigmoid function?

Sigmoid function g$($x$)$ is often used as the activation function in neural networks. Which of the following statements is NOT true regarding sigmoid function?

The derivative of g$($x$)$ is g$($x$)(1$ g$($x$)).$

Sometimes the Sigmoid function can also be replaced by other activation functions such as ReLu and Tanh. Different activation functions have different impacts on performance.

With the sigmoid function, the initial values of parameters must be small during training.

None of the above

Which of the following is NOT true?$$

ANN can be used for either regression or classification.

In classification ANN, the output layer may use Sigmoid $($for a single class$)$ or Softmax $($for multiclass$)$ as the activation function.

We usually use the same cost function to train regression ANN and classification ANN.$*$

Stochastic gradient descent is helpful in reducing the chance of local minima.