Nonlinear Dimensionality Reduction Techniques and Their Application in Neural Networks

Taken by: Michael Hobbs
Completed by: Michael Hobbs
Download Project Report in PDF Format

Project Description

Dimensionality reduction is a widely used technique in machine learning and data processing. Since dealing with large amount of high dimensional data is difficult, one could use dimension reduction to reduce the number of variables (dimensions) and then use this coarse version of input data for further processing.

It is believed that human neural system also performs lots of dimensionality reduction to efficiently deal with natural stimuli.

In this project we focus on studying and implementing some of the widely used nonlinear dimensionality reduction techniques using neural networks. The implemented approaches are then applied to a dataset of natural stimuli (images or sounds) to extract important features from the datasets. The final goal of this project would be to see if such features could help us increase the storage capacity of various artificial neural memories.

The implementation can be done in MATLAB or C/C++ (MATLAB is preferred though).

To read more about nonlinear dimensionality reduction, here are the corresponding wikipedia entries Dimension Reduction and Nonlinear Dimensionality Reduction.