We are glad to announce release 2.4 of the Modular toolkit for Data
MDP is a Python library of widely used data processing algorithms that
can be combined according to a pipeline analogy to build more complex
data processing software. The base of available algorithms includes,
to name but the most common, Principal Component Analysis (PCA and
NIPALS), several Independent Component Analysis algorithms (CuBICA,
FastICA, TDSEP, and JADE), Slow Feature Analysis, Restricted Boltzmann
Machine, and Locally Linear Embedding.
What's new in version 2.4?
- The new version introduces a new parallel package to execute the MDP
algorithms on multiple processors or machines. The package also offers
an interface to develop customized schedulers and parallel algorithms.
Old MDP scripts can be turned into their parallelized equivalent with
one simple command.
- The number of available algorithms is increased with the Locally
Linear Embedding and Hessian eigenmaps algorithms to perform
dimensionality reduction and manifold learning (many thanks to Jake
VanderPlas for his contribution!)
- Some more bug fixes, useful features, and code migration towards