Scikit-learn compatible tools using theano
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
benchmarks
build_tools/circle
doc
examples
sklearn_theano
.gitignore
AUTHORS.rst
LICENSE
MANIFEST.in
README.rst
circle.yml
setup.py

README.rst

sklearn-theano

This repository contains experiments with scikit-learn compatible estimators, transformers, and datasets which use Theano internally.

Important Links

Setup

The HTML docmentation linked above has a variety of resources for installation and example use. To get started immediately, clone this repo then do:

python setup.py develop

When using the examples, there will be some fairly large downloads (~1GB) to get weights, sample datasets, and other useful tools. The default directory for this is $HOME/sklearn_theano_data.

The key packages required are:
  • numpy
  • scipy
  • theano
  • scikit-learn
  • pillow

and a soft dependency on matplotlib for the examples.

Documentation is sparse but we are working to improve unclear modules. Feel free to raise issues on GitHub with any problems found!

License

The License for sklearn-theano is 3-clause BSD. See the LICENSE file in the top level of the repository https://github.com/sklearn-theano/sklearn-theano/blob/master/LICENSE

This project provides downloaders for models that are distributed under their own terms, namely:

The model specification for the BVLC Caffe GoogLeNet model are taken from a protocol buffer file, https://raw.githubusercontent.com/BVLC/caffe/master/src/caffe/proto/caffe.proto which is distributed under the same licence as the Caffe code (2-clause BSD).