GLM_gradient_adaptive: GLM_gradient_adaptive

In Z1chenZhao/bis557: BIS557

Description

Implement a first-order solution for the GLM maximum likelihood problem using only gradient information, avoiding the Hessian matrix. Standard adaptive update Momentum.

Usage

GLM_gradient_adaptive(X, Y, mu_fun, mom = 0.2, maxit = 1e+05, tol = 1e-05)

Arguments

X	The design matrix
Y	Response variable
mu_fun	A function to use eta get mu
mom	The momentum update size, set default as 0.2
maxit	The maximum iteration times to allow the loop to exit, set default as 1e5
tol	The difference between previous beta and new beta to allow the loop to exit, set default as 1e-5.

Details

The code is adapted from lecture notes.

Value

A list of beta coefficients.

Z1chenZhao/bis557 documentation built on Dec. 21, 2020, 10:03 p.m.