Skip to content

README.md

Latest commit

 

History

History
41 lines (36 loc) · 3.26 KB

README.md

File metadata and controls

41 lines (36 loc) · 3.26 KB

deep-learning

deep-learning contains code to train and evaluate deep neural networks on data from the CMS Experiment at CERN's LHC. It also contains some scripts to graph the performance of networks.

Installing:

  1. Get the stuff we wrote
    • git clone https://github.com/klannon/deep-learning.git deep_learning
    • It is VERY IMPORTANT that you clone the repository into the folder "deep_learning" (which will happen by default) instead of some other directory of your choosing. The imports depend on the top-level module being "deep_learning."
  2. Create your virtual environment

    To keep this repository's dependencies separate from the rest of the python packages you have installed on your system, we recommend setting up a virtual envirtonment to contain the software you install in step 3.

    1. Install virtualenv
    2. Because some of the software in step 3 depends on scipy and numpy, each of which is tightly coupled with your system BLAS libraries, set up your virtual environment like this: virtualenv --system-site-packages env so virtualenv will know about your system installations of numpy and scipy.
    3. Since your virtual environment wouldn't be very useful if you didn't activate it, source env/bin/activate if you use BASH or source env/bin/activate.csh if you use CSH
    </li>
<li>Get the stuff we <em>didn't</em> write
    <ul>
        <li><a href="http://keras.io/">Keras</a>: <code>pip install keras</code> Deep learning library used for designing/executing experiments</li>
        <li><a href="https://developers.google.com/protocol-buffers/">Google Protocol Buffers</a>: <code>pip install protobuf==3.0.0b2</code> Data serialization library used to save the results of Keras experiments in an intelligible format</li>
        <li><a href="http://www.pytables.org/downloads.html">PyTables</a>: <code>pip install tables</code> Library we use to store and access arrays of data very efficiently, even if it exceeds your RAM</li>
    </ul>
    Note: If you want to visualize the results of the networks you train, you should use <a href="https://github.com/mdkdrnevich/deep_viz">Deep Viz</a>
</li>
<li>Set the <code>PYTHONPATH</code> environment variable to the <em>PARENT</em> directory of the directory that you cloned this repository into (whatever directory containes the folder "deep_learning" by default) so python can resolve our imports correctly.
    <ul>
        <li>CSH: <code>setenv PYTHONPATH /path/to/parent/directory</code></li>
        <li>BASH: <code>export PYTHONPATH="/path/to/parent/directory"</code></li>
    </ul>
</li>

If you're planning on using this code frequently, consider adding the PYTHONPATH environment variable to your .cshrc, .bashrc, .bash_profile, or .profile (depending on your shell of choice) so you don't have to keep setting it