dg_batch_minibatch_online_seq: Global Gradient descent algorithm

In Beuleup93/dgrGlm: Regression logistic binary tools

Description

This function allows the execution of the binary logistic regression according to the batch, mini batch and online mode.

Usage

dg_batch_minibatch_online_seq(
  X,
  y,
  theta,
  batch_size,
  random_state,
  leaning_rate,
  max_iter,
  tolerance,
  rho = NA,
  C = NA
)

Arguments

X	is the matrix of our predictor variables with the bias column
y	is a target variable to predict.
theta	is a vector containing the parameters or coefficient of the logistic to be estimated
batch_size	a parameter that specifies the number of observations in each mini-batch. It can significantly affect performance
random_state	this parameter defines the seed of the random number generator, use when shuffling to mix observations.
leaning_rate	is the learning rate that controls the magnitude of the vector update.
max_iter	is the number of iterations.
tolerance	an additional parameter which specifies the minimum movement allowed for each iteration
rho	hyper parameter which allows arbitration between RDIGE and LASSO.
C	parameter allowing to arbitrate between the penalty and the likelihood in the guidance of the modeling.

Value

this function returns an instance containing:

• final theta

• history cost

• iteration number

• intern iteration number

Examples

## Not run: 
 global_grad_descent(X,y,theta)

## End(Not run)

Beuleup93/dgrGlm documentation built on Dec. 17, 2021, 10:50 a.m.