Python: Advanced Guide to Artificial Intelligence

Logistic regression for binary classification

For binary classification, we define the model function ϕ(z) to be the sigmoid function, written as follows:

The sigmoid function produces the value of y to lie between the range [0,1]. Thus we can use the value of y=ϕ(z) to predict the class: if y > 0.5 then class is equal to 1, else class is equal to 0.

As we saw in the previous sections in this chapter that for linear regression, the model can be trained by finding parameters that minimize the loss function and loss function could be the sum of squared error or mean squared error. For logistic regression, we want to maximize the likelihood:  ...