/toUniform

A simple neural model for unlabelled data

Primary LanguageJupyter Notebook

A simple neural model for unlabelled data

This model transforms data to uniformly distributued points. This is achived just by mapping nearby points to distants image points. By doing this transformation, it is easy to understand the structure behind the data and even generate new data. Details can be found in our preprint https://www.researchsquare.com/article/rs-2671009/latest.

Alt text

For demonstration we transform two datasets onto the plane so that the underlying pattern behind the data is visible. Just give a try!

First one is Swis Roll dataset shown in fugures below and the implementation is here: https://github.com/unverciftci/toUniform/blob/main/DatatoUniformSR.ipynb.

Alt text Alt text

Second example is MNIST dataset shown in fugures below and the implementation is here: https://github.com/unverciftci/toUniform/blob/main/DataToUniformMnist.ipynb

Alt text Alt text