This repository contains experiments with scikit-learn compatible estimators, transformers, and datasets which use Theano internally.
- Official source code repo: https://github.com/sklearn-theano/sklearn-theano
- HTML Documentation: http://sklearn-theano.github.io
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!
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 OverFeat model http://cilvr.nyu.edu/doku.php?id=code:start
- The BVLC Caffe GoogLeNet model http://caffe.berkeleyvision.org/
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).