Source code for the experiments and figures of the paper "Adaptive Acceleration of Sparse Coding via Matrix Factorization ".
All the development was done with python3.4 and might not work for earlier versions.
Use the main script NIPS_figures.py to launch the experiements. Various option are available from the command line. See python NIPS_figures.py --help for more information.
To generate the 4 figures from the paper, use:
python NIPS_figures.py --data artificial --save_dir layer1
python NIPS_figures.py --data artificial --rho .2 --save_dir layer2
python NIPS_figures.py --data mnist --lmbd .1 -K 100 --save_dir mnist
python NIPS_figures.py --data images --lmbd .05 --save_dir images