Gijs van Tulder

Convolutional classification restricted Boltzmann machines

These scripts have been used to train and evaluate convolutional classification restricted Boltzmann machines.

The code provided here was used for the experiments in

Combining Generative and Discriminative Representation Learning for Lung CT Analysis with Convolutional Restricted Boltzmann Machines
by Gijs van Tulder and Marleen de Bruijne
in IEEE Transactions on Medical Imaging (2016)
http://dx.doi.org/10.1109/TMI.2016.2526687

The code is available at GitHub: https://github.com/gvtulder/ccrbm.

If you have any questions, comments or are using this code for something interesting, we would love to know. We would also appreciate it if you cite our paper if you use this code for your own publications.

February 2016

Gijs van Tulder
http://vantulder.net/

Biomedical Imaging Group Rotterdam
Erasmus MC, Rotterdam, the Netherlands
http://www.bigr.nl/

Data

The experiments were done on data from a public dataset for interstital
lung diseases, which is described in

Building a reference multimedia database for interstitial lung diseases
by Adrien Depeursinge et al.
in Computerized Medical Imaging and Graphics (April 2012)
http://dx.doi.org/10.1016/j.compmedimag.2011.07.003

Once you have a copy of this dataset, you can use the MATLAB scripts in
patch-preprocessing/ to extract patches.

Requirements

We have used this code with:

The code uses a modified version of Morb, a modular RBM implementation in Theano. The modifications include support for classification RBMs.

You will need Ruby to run the glue scripts that generate the parameter sets and schedule the experiments.

See morb-repo/ for details, or go to https://github.com/gvtulder/morb

The original version of Morb by Sander Dieleman can be found at https://github.com/benanne/morb

Components

The main components are:

These scripts are fairly generic and take a set of parameters to run. To run the actual experiments, we used two other scripts:

As baselines, we used Leung-Malik and Schmidt filter banks. Code to generate these filters can be found in the filter-banks/ directory.

Copyright (c) 2016 Gijs van Tulder / Erasmus MC, the Netherlands. This code is licensed under the MIT license:

The MIT License (MIT)

Copyright (c) 2016 Gijs van Tulder / Erasmus MC, the Netherlands

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.