Linear regression is to use linear equation, the characteristics of the data all fit into the equation, write the cost function: (yi - (mxi + b)) in the form of the same cost function, and then use the gradient descent or normal equation to select feature (the number of less than ten thousand) the method to adjust offset and weights

CodePudding user response:

It can be understand that
The neural network is do "nonlinear" effect,

CodePudding user response:

Using neural network (Relu, Tanh, sigmoid activation function... ) do "nonlinear" effect,