In alexpghayes/choosy: Exploration of Optimization Methods for Multinomial Logistic Regression

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "README-"
)

choosy

The ideal of this package is to get some practice with optimization algorithms to fit multinomial logistic regression models.

To Do / Roadmap

1. (Backtracking?) Line search for gradient descent / selection of optimization method
2. BFGS / L-BFGS
3. Newton's method
4. ADMM
5. Cpp once xtensor-r is easy

Eventually we may want the following methods:

• accept recipe objects?
• tidy method
• augment method
• plot method

Installation

Once we get underway, you'll be able to install this package with:

# install.packages('devtools')
# devtools::install_github('alexpghayes/choosy')

Motivation

...

Assumptions

• Independence of Irrelevant Alternatives: what it is, and what are some options to get around it, where those are implemented in R (i.e. mlogit)
• What type of data you have: data about observed individuals rather than alternatives to select from
• Other

Example Usage

...

alexpghayes/choosy documentation built on May 23, 2019, 11:35 p.m.