"Wunda Machine learning" Learning note--2.7 First learning algorithm = linear regression + gradient descent

gradient descent algorithm: linear regression Model:　　　　　　　　　　　　　　

Linear hypothesis:

Squared difference cost function:

By substituting each formula, the θ0 and θ1 are respectively biased:

By substituting the partial derivative into the gradient descent algorithm, we can realize the process of finding the local optimal solution.

The cost function of linear regression is always a convex function, so the gradient descent algorithm only has a minimum value after execution.

" Batch " gradient descent: use all training samples for each step

"Wunda Machine learning" Learning note--2.7 First learning algorithm = linear regression + gradient descent