Elefant 0.4.0

Elefant (Efficient Learning, Large-scale Inference, and Optimisation Toolkit) includes modules for many common optimisation problems arising in machine learning and inference.

Elefant is designed to be modular and easy to use. Framework provides easy to use Python interface, which can be use for quick prototyping and testing inference algorithms.

Following machine learning algorithms are implemented in the Elefant:
· Support Vector Machine (SVM) for classification, regression, quantile and novelty detection, online learning, Epsilon Insensitive and Laplacian support vector regression.
· Heteroscedastic Gaussian Process regression, Gaussian Process Regression, Multi-class transductive classification with Gaussian Process.
· Solvers for the quadratic programming problem
· BAHSIC feature selection
· Algorithms for fast computation and manipulation of kernel matrices. Linear, RBF, Dot Product and String Kernels
· Loopy Belief Propagation and Junction Tree algorithms.
· Cover Tree for calculating the nearest neighbor

Elefant is currently being tested on the following platforms:
· Mac OS X (10.4)
· Linux (Ubuntu 7.04 - Feisty Fawn)
· Windows XP (SP2), Windows Vista

Here are some key features of "Elefant":

· Light weight component based system design, plug and play kind of a architecture
· Component suite for basic as well as advanced machine learning algorithms
· Support for various data source formats
· Components for data visualizations
· Easy to use graphical user interface for visual programming and quick prototyping
· Intuitive application programming interface for advanced prototyping
· Python wrappers for high-performance parallel scientific packages like PETSc, TAO, and SLEPc
· Interface to external systems like UIMA using jpype
· Comprehensive system documentation
· Open source and licensed under the Mozilla Public License (MPL)

Requirements:

· Python 2.5 or later

What's New in This Release: [ read full changelog ]

· Now compatible with Python version 2.6, and prepared for the move to Python3000.