new_gradient_descent: New Gradient Descent

In BillyTian/bis557: Package containing assignments, projects or other products in BIS557

Description

This is a function fitting the OLS model using gradient descent which calculate the loss based on the out-of-sample accuracy.

Usage

new_gradient_descent(form, data, gamma = 1e-04, beta0 = 1,
  fold.num = 10, stop.diff = 1e-08, max.itera = 5e+05)

Arguments

form	a formula with the legal format.
data	a dataframe provided by the user.
gamma	a factor defined by user as a pace or speed of gradient descent (default=0.0001).
beta0	an initializer for the beta vector that can be changed by user (default=1).
fold.num	a fold number of k-fold cv.
stop.diff	an absolute difference between loss and the updated one set by user to stop the iteration at some satisfying point (default=1e-8).
max.itera	a maximum number of iterations set by user, which is an alternative criterion to stop the iterations (default=1e6).

Value

a list including the coefficient estimates.

Examples

data(iris)
fit <- new_gradient_descent(Sepal.Length ~ ., iris)
fit$coefficients

BillyTian/bis557 documentation built on Dec. 19, 2020, 7:30 a.m.