https://github.com/wmkouw/sample-covariateshift

Sample from synthetic covariate shift problem
https://github.com/wmkouw/sample-covariateshift

covariate-shift dataset machine-learning rejection-sampling

Last synced: 5 months ago
JSON representation

Sample from synthetic covariate shift problem

• Host: GitHub
• URL: https://github.com/wmkouw/sample-covariateshift
• Owner: wmkouw
• License: mit
• Created: 2018-05-25T16:33:23.000Z (about 8 years ago)
• Default Branch: master
• Last Pushed: 2018-05-27T17:50:01.000Z (about 8 years ago)
• Last Synced: 2025-04-04T15:17:09.366Z (about 1 year ago)
• Topics: covariate-shift, dataset, machine-learning, rejection-sampling
• Language: Matlab
• Homepage:
• Size: 4.88 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# sample-covariateshift

Scripts to sample from class-conditional distributions, under covariate shift.

To be precise:

Assume that there is a single sample space X, and a fixed number of classes Y. Let the _source domain_, referred to as pS(x,y), be one distribution over (X,Y) and the _target domain_ another, pT(x,y). In cases of covariate shift, the posterior distributions are equal in both domains; pS(y|x) = pT(y|x).

Distributions:

- p(y) is the distribution of the classes (equal in both domains).

- p(y|x) is the posterio distribution (equal in both domains; hence _covariate shift_)

- pS(x) is the source distribution of the data.

- pT(x) is the target distribution of the data.

The scripts generate class-conditional distributions for each domain, pS(x|y) and pT(x|y), using Bayes' rule. Samples are drawn using a rejection sampler.

## Usage

Have a look at `example_call.m`

## Questions

Questions, comments and bugs can be submitted in the [issues tracker](https://github.com/wmkouw/sample-covariateshift/issues).