Generalized Linear Models
var glm_model = glm;var feature_vectors = 1 2;var target_values = 1 2;glm_model;
We will soon have support for regularization and Probit regression. After this, we plan on optimizing the runtime performance of the system. It would be neat to have support for MAP or fully Bayesian GLMs, but we currently don't see any reason to work on this functionality.
There is one main function called GLM which expects a distribution to be passed in to it. The families can be found in the families attribute of this GLM function. For example: GLM.families.Gaussian() will initialize a gaussian distribution object.
This is essentially a port of a python GLM implementation that uses the iteratively reweighted least squares algorithm in the excellent statsmodels library.
$ npm install node$ node> var glm = require('glm');
Just include 'glm.js' as a script in your HTML code.
python -m SimpleHTTPServer in the root of this repo and navigate your browser to
To compile, first install the dependencies with npm and then run make. To test, run