Do you work with PyTorch? Would you like to have smarter, smaller, cheaper ML models? Would you like to have them today rather than next year? Then join us for Perforated AI's second Dendritic Optimization Hackathon as a part of Open Source Week!

​Neuroscience has come a long way since the artificial neuron was first invented in 1943. In 2012 researchers discovered that this original design was missing something important, the additional computation happening within biological dendrites. This hackathon is an opportunity to use Perforated AI's open source repository to add artificial dendrites to your own PyTorch projects. Empowering them to be even more powerful on the same datasets.

​Mandatory for Participation: Bring a project with you!  This is not a typical hackathon where you build something from scratch.  You will be applying Dendritic Optimization.  When you walk in the door you should already have a functional PyTorch project you want to improve.

​Schedule:

• ​October 23rd

  • ​6:30: Welcome / Networking

  • ​7: Dinner and Presentation on Dendritic Optimization

  • ​7:30: Dendritic Optimization Implementation Walkthrough

  • ​7:45: Weights and Biases Hyperparameter Sweeps Implementation Walkthrough

  • ​8: Hacking

  • ​10:45: Closeout

• ​October 24rd - January 4th

  • ​Experiments!  The hackathon day is just to get the initial system implemented.  You shouldn’t have to do any more coding at this point, but you do have to actually run the sweeps. 

• ​January 5th

  • ​Final due date for project submissions

• ​January 8th

  • ​The hackathon wrap-up will be entirely online to support everyone coming in from out of town for the PyTorch conference.  This will be a one hour session to discuss top projects and award prizes for the winners.

​​Judges:

• ​Rorry Brenner - Perforated AI

• ​Mohammad Bakir - Weights and Biases

• ​Arisa Chelsea Ueno - DraperU Ventures

​Additional details here.

• General Suggestions
  Link here with details: https://docs.google.com/document/d/1MJcxo7tTPXfAky8qrIPrav3WzR_tfdfUDPKpp1BDN_Y/edit?tab=t.0