gradient_mini_batch: Mini_batch gradient descent
In AlexandreRives/RegLog-EDA: Logistic Regression

Mini_batch gradient descent computes the gradient using a batch size of the dataset

gradient_mini_batch(
  df,
  var_X,
  var_y,
  nb_batch,
  learning_rate,
  max_iter,
  graph,
  epsilon
)

`df`	dataset
`var_X`	names of the X columns
`var_y`	names of the Y columns
`nb_batch`	batch size
`learning_rate`	the learning rate
`max_iter`	number of iterations
`graph`	TRUE if you want to plot the cost list while the gradient descent is running.
`epsilon`	Tolerance's threshold of the cost list convergence.