SSG Software

Locally Linear Embedding


[LLE figure]
LLE on SDSS spectra

Author: Jake VanderPlas

Download: DimReduce version 3.1 (tgz file)

About: DimReduce is a C++ package for performing nonlinear dimensionality reduction of very large datasets with Locally Linear Embedding (LLE) and its variants. DimReduce is built for speed, using the optimized linear algebra packages BLAS, LAPACK, and ARPACK. Because of the need for storing very large matrices (1000 by 10000, for our SDSS LLE work), DimReduce is designed to use binary FITS files as inputs and outputs. This means that using the code is a bit more cumbersome. For smaller-scale LLE, where speed of computation is not as much of an issue, the Modular Data Processing toolkit may be a better choice. It is a python toolkit with some LLE functionality, which VanderPlas contributed.

If you find this software useful in your research please reference VanderPlas, J. and Connolly, A.J. "Reducing the Dimensionality of Data: Locally Linear Embedding of Sloan Galaxy Spectra", 2009, AJ (submitted) (ADS link)

eLSST Software

Note: Much of this software is for internal use, and is password protected.

eLSST software page can be found here.