fit: Fit model to train data

In samiaittilat/Logistic-Regression-from-scratch-in-R: Project R SISE 2021 2022

Description

Logistic regression on train data (classication)

Usage

fit(
  formule,
  data,
  mode = "mini_batch",
  batch_size = 32,
  alpha = 0.01,
  max_iter = 500,
  tol = 0.001,
  ncores = NA,
  parallel = FALSE,
  plot = FALSE,
  lambda = 0,
  beta = 1
)

Arguments

formule	an object of class "formula" (or one that can be coerced to that class): a symbolic description of the model to be fitted. The details of model specification are given under ‘Details’
data	a dataframe containing the variables in the model.
mode	parameter is the way you feed your data to the model ( online, mini-batch or batch). The mode is set to mini-batch by default.
batch_size	the batch size when mode is mini-batch
alpha	the learning rate in gradient descent
max_iter	maximum number of iteration
tol	tolerance for accepting a model :
ncores	indicates the number of cores to use; if invalid (<= 0 or> number of cores available), the maximum capacities of the host machine are used
parallel	logical.Should we use parallel computing ? specifies whether you want to perform a parallel or a sequential computing. The default value is set to False ( meaning it's sequential). Note that parallel computing is not performed in the case of online gradient descent.
plot	logical. Should we plot the cost function while iterating ?
lambda	parameter is set to zero by default if no regularization is applied to the model. You may want to modify this value if your model is overfitting.
beta	beta parameter which takes only two values (beta = 1 by default: ridge regularization case) beta = 0 lasso case

Details

Function that trains your model via gradient descent and outputs the model weights. This function comes with a bunch of parameters.

Value

An object S3 of class Reg.Log md$coeffs : Coefficients ( model weights) md$nb_iter: Number of epochs md$cost: cost function values at each epoch md$mean_col= mean value of each feature md$sd: Standard deviation value of each feature

Author(s)

Marie Vachet, Afaf Ben Haj, Sami Ait Tilat

Examples

## Not run: 
modele = fit_grad(formule = y~., data=nos_data,max_iter = 300, mode='batch', parallel = TRUE )

## End(Not run)

samiaittilat/Logistic-Regression-from-scratch-in-R documentation built on Dec. 22, 2021, 10:11 p.m.