Ronan Collobert (collober [at]
Release 3.1
August 11, 2004

This is a minor update (bug corrections).
See the ChangeLog.

Torch 3 Vision

A full additional package for machine learning applied to vision applications is now available.
Have a look here.


Please, read the installation notes in the documentation section before downloading anything.

    Archive     Description
Torch3 src Torch3 for Unix/Linux
Torch3 doc Torch3 documentation
Torch3 win    Torch3 for MS Windows   


We strongly encourage you to use from now xmake (a python script designed for Torch) instead of the GNU make software for compiling Torch.

Note that the sources for Unix/Linux and MS Windows are the same... only the packaging method is different.
If for some reasons you want the previous version of Torch, it is still available here.

Short description of packages

Package Description Dependencies Maintainers
core The core of Torch Ronan Collobert
gradients Gradient machines Ronan Collobert
kernels Machine based on kernel methods,
such as SVM
Ronan Collobert
distributions Stuff related to distributions gradients Samy Bengio
Johnny Mariéthoz
convolutions Spatial and temporal convolutions gradients Ronan Collobert
matrix Matrix manipulations Ronan Collobert
datasets Several classes to handle data
such as HTK
Johnny Mariéthoz
 non-parametrics  Non-parametrics algorithms,
such as KNN or Parzen Windows
Samy Bengio
speech Tools for speech recognition distributions
Samy Bengio
Johnny Mariéthoz
decoder Large vocabulary speech decoder distributions
Darren Moore
examples Some main program examples all packages
if you want to run all
all coders