https://github.com/dirkschumacher/tfjs-glm

Generalized linear models in tensorflow.js (WIP)
https://github.com/dirkschumacher/tfjs-glm

generalized-linear-models statistics tensorflow tensorflow-js

Last synced: 2 months ago
JSON representation

Generalized linear models in tensorflow.js (WIP)

• Host: GitHub
• URL: https://github.com/dirkschumacher/tfjs-glm
• Owner: dirkschumacher
• License: mit
• Created: 2019-06-23T19:21:23.000Z (almost 6 years ago)
• Default Branch: master
• Last Pushed: 2023-05-06T03:15:53.000Z (about 2 years ago)
• Last Synced: 2025-03-10T15:11:31.106Z (2 months ago)
• Topics: generalized-linear-models, statistics, tensorflow, tensorflow-js
• Language: JavaScript
• Size: 20.5 KB
• Stars: 3
• Watchers: 1
• Forks: 0
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# tfjs-glm

Generalized linear models in tensorflow.js using Iteratively Re-Weighted Least Squares (WIP)

Experimental and work in progress. Use at own risk. Still numerical problems, especially with log link.

## Families

* Gaussian (links: identity)

* Binomial (links: logit)

* Poisson (links: log)

## API

```js

const {glm, gaussian, binomial, poisson} = require("tfjs-glm")

// linear regression

const coefficents = glm(designMatrix, response, gaussian("identity"))

// logistic regression

const coefficents = glm(designMatrix, response, binomial("logit"))

// poisson regression

const coefficents = glm(designMatrix, response, poisson("log"))

```

## References

1: Arnold, T., Kane, M., & Lewis, B. W. (2019). A Computational Approach to Statistical Learning. CRC Press.