Softwares

Machine Learning Demos

MLDemos is a tool to get interactive visualization of various machine learning algorithms. It could be very helpful to get intuitive ideas of how different algorithms, different kernels and different parameters affect the learning results. Implemented methods include classification algorithms like SVM, Neural Networks, Approximate KNN, etc.; regression algorithms like SVR, Gaussian Process, etc; as well as various clustering, projection algorithms, and more.

Here is a figure taken from the official website showing how different kernel width affect the regression result of a Gaussian Process. See the video demo of MLDemo at YouTube for more details.

Manifold Learning Demos

Dimension Reduction

mani.m is a Matlab GUI demo for various dimension reduction algorithms, including MDS, PCA, ISOMAP, LLE, Hessian LLE, Laplacian Eigenmaps, Diffusion Map and LTSA. Here is a screenshot of unrolling the Swiss Roll with Hessian LLE.

Graph Learning

GraphDemo is a Matlab GUI demo for learning with nearest neighbor graphs. Specifically, it includes demo for graph construction, as well as spectral clustering and semi-supervised learning with the constructed graph. Here is a screenshot showing the result of semi-supervised learning on the two moon dataset.