The Stochastic Gradient Descent and Mini-batch gradient descent

[frameproject]

hello Aurélien,

From https://github.com/amitanalyste/aurelienGeron/blob/master/04_training_linear_models.ipynb

The Stochastic Gradient Descent and Mini-batch gradient descent you use :

    gradients = 2 * xi.T.dot(xi.dot(theta) - yi) 

it gives a good result but the cost function should be:

  gradients = 2/m * xi.T.dot(xi.dot(theta) - yi) 

then you use this function for :

Linear regression using batch gradient descent and :

gradients = 2/m * X_b.T.dot(X_b.dot(theta) - y) 

What are the differences ?

Many thanks

Frederic