glmfo: first-order solution for the GLM maximum likelihood

In importbq/bis557: Package for BIS557

Description

implementation of a first-order solution for the GLM maximum likelihood problem using only gradient information.

Usage

glmfo(x, y, family, method = "constant", lr = 1e-04, maxit = 10000,
  er = 1e-12, print = FALSE)

Arguments

x	feature variables
y	response variables
family	family object and link function
method	optimization method
lr	learning rate
maxit	Maximum number of iterations
er	allowed errors
print	whether to print to track progress

Examples

set.seed(100)
n <- 1000
X <- cbind(1, matrix(rnorm(n * 3), ncol = 3))
beta <- c(-1, 0.2, 0.5, 0.6 )
Y = rpois(n,exp(X %*% beta))
glmfo(X,Y,poisson(link='log'))

importbq/bis557 documentation built on Dec. 21, 2020, 3:05 a.m.