Supervised Transfer Sparse Coding

A MATLAB implementation of the STSC algorithm that makes use of LIBSVM and LIBLINEAR for prediction. Additionally, we provide handwritten digits data used in the experiments of the original paper, a few example scripts for running experiments and cross-validating in parallel on a multi-core machine, and utility functions for visualization.

Download

The code and data can be donwloaded directly as a zip-file (200MB). Alternatively, you may clone the repository:

$ git clone --recursive git@github.com:alshedivat/stsc.git

Note: You need to clone recursively as STSC depends on LIBSVM and LIBLINEAR submodules.

Installation

After downloading/cloning the code, you just need to mex-compile LIBSVM and LIBLINEAR libraries. This is done from MATLAB by executing make script in code/libsvm and code/liblinear subfolders:

$ matlab -nodisplay -nosplash
[...]
>> cd code/liblinear/matlab
>> make
>> cd ../libsvm/matlab
>> make

Usage

The code is executed by simply running scripts given in the code subfolder. First, generate a config using GetConfiguration function. Next, run one of the scripts, e.g., SingleRun or CrossValidation. For more details, please refer to the code; it's written in a clean and neat way.

Citation

If you use this code (in full or in part) for academic purposes, please consider citing our paper:

@inproceedings{alshedivat2014stsc,
  title={Supervised Transfer Sparse Coding},
  author={Al-Shedivat, Maruan and Wang, Jim Jing-Yan and Alzahrani, Majed and Huang, Jianhua Z and Gao, Xin},
  booktitle={Proceedings of the Twenty-Eighth AAAI Conference on Artificial Intelligence},
  year={2014},
  organization={The AAAI Press}
}

License

MIT (for details, please refer to LICENSE)

Copyright (c) 2014 Maruan Al-Shedivat