Diffusion map

Diffusion map of time series or similarity matrix

DiffusionMap Toolbox

This toolbox provides a simple, flexible way to perform diffusion map analysis—an approach to dimensionality reduction that preserves local data geometry. The functions included allow you to compute a similarity matrix, apply various normalization schemes, and extract diffusion map coordinates through eigenvector decomposition. An example script (`example1swissroll.m` or `example1_swissroll.mlx`) demonstrates usage on a classic Swiss roll dataset, illustrating how to reveal underlying low-dimensional structure.

Key Features

- Calculation of similarity matrices with multiple distance metrics

- Options for row or column normalization

- Different tuning parameters (e.g., number of nearest neighbors, Laplacian type)

- Example scripts to get started quickly

License

Distributed under the MIT License. See `LICENSE.txt` for details.