Why do we use sigmoid function?
It works only when it is a binary classification problem
They have to be chosen so it makes the best fit regression, which makes the smallest mean squared error
Used for binary classification only because, as shown in the diagram, the y value can either be 0 or 1, which is binary. The 0 or 1 can represent yes or no, positive or negative, pass or fail, etc, respectively.
Some data sets have a general behavior that fits the sigmoid function, which is used in logistic regression