glm_new: glm_new() function

In Yannuo10/bis557: Homework 1

Description

another way to build a generalized linear model using only gradient information

Usage

glm_new(
  X,
  y,
  mu_fun,
  var_fun,
  itr = 50,
  step,
  method,
  friction,
  tolerance = 1e-10
)

Arguments

X	a data matrix used for the function
y	a vector of response values
mu_fun	expectation function of the distribution
var_fun	variance function of the distribution
itr	number of iteration
step	learning rate for gradient descent
method	constant step size or momentum update
friction	a number to control the velocity and prevents overshooting
tolerance	allowed tolerance for two residuals

Examples

 n <- 5000; p <- 3; beta <- c(-1, 0.2, 0.1); X <- cbind(1, matrix(rnorm(n * (p- 1)), ncol = p - 1))
 eta <- X %*% beta; lambda <- exp(eta); y <- rpois(n, lambda = lambda)
   fit_linear_model <- glm_new(X, y, mu_fun = function(eta) exp(eta), var_fun = identity, step = 0.0001, method = "constant")

Yannuo10/bis557 documentation built on Dec. 21, 2020, 10:02 p.m.