MLC@Home: Machine Learning Comprehension @ Home

Frequently Asked Questions

  • What is this?

    MLC@Home is a distributed computing project dedicated to understanding and interpreting complex machine learning models, with an emphasis on neural networks. It uses the BOINC distributed computing platform.

  • What is BOINC?

    BOINC is the Berkeley Open Infrastructure for Network Computing. It is a volunteer-based distributed computing platform, where anyone can run the official BOINC client on their computer, and project like MLC@home can send out chunks of information to process (called "Work Units" or "WUs"). BOINC was created to power SETI@Home, but now there are many projects volunteers can choose to donate their computer time to. BOINC only runs when the computer is idle, so your ordinary work shouldn't be interrupted.

  • Who are you?

    MLC@Home is a project of the Cognition, Robotics, and Learning (CORAL) Lab at the University of Maryland, Baltimore County (UMBC). It was created by John Clemens, a Doctoral Candidate in CS, and a member of CORAL. You can view his Google Scholar profile, and his homepage. John is studying ways to extract information from neural networks so we can better understand their uses and limitations.

  • How can I contact you?

    You can contact the MLC Admins at the email address mlcathome2020 at gmail.com. Or you can sign up on the project site (click "BOINC Project Page" above) and post on the project forums.

  • You're going to download code and run it on my computer? Is that safe?
  • I can refer you to BOINC's own page about this here. I've been running BOINC projects on my own computers since the late 1990s with no issue.

  • I want to volunteer, how do I start?

    If you're running Linux or 64-bit Windows, follow the directions under "How you can help" on the homepage.

  • I want to help, but don't run Linux or Windows, what can I do?

    We'd like to support a Mac OSX client, but we do not have access to a mac to develop on, nor any mac development experience. The client is pure posix C++17, however, so it should be "easy" to do for an experienced developer. Other Linux platforms are also a possibility, but a low priority. So for now, stay tuned. Please consider signing up for one of the other worthwhile BOINC projects until MLC@home is ready to support your platform.

  • What science are you doing?

    Our first task is generating a dataset of thousands of similar neural network models. We want to study how networks trained on similar data differ from from each other. This will be used to test if we can determine if a network has been trained with malicious data can be spotted by comparing it to known good networks. A more complete description is available by clicking on "MLDS Dataset" above.

    But that's just the first subproject ready to go. We see MLC@Home supporting more research, like federated learning, neuro-evolution, and hyperparameter search. Stay tuned.