Skip to content
sumitsomans edited this page Aug 17, 2017 · 1 revision

EigenSample

Welcome to the EigenSample wiki!

About EigenSample

EigenSample is a method for augmenting small datasets by synthetically generated samples. The new samples generated by this method preserve the eigen-structure of the data, i.e., new samples are added after projecting the data onto the principal eigenvectors, clustering in the low-dimensional space and adding samples along lines joining the input samples to their cluster centers.

The newly added samples are projected back into input space by a non-iterative least squares approach, which allows for parameters to control the approximation error and tolerance of the reverse projection.

EigenSample Paper

The paper is published in the Elsevier Applied Soft Computing journal special issue on Non-iterative Methods in Learning (DOI: 10.1016/j.asoc.2017.08.017).

Clone this wiki locally