Epilepsy Seizure Prediction

This is the project page for the Kaggle UPenn and Mayo Clinic’s Seizure Detection Challenge. The NuPIC community is interesting in putting together a formal team to address this challenge.

This project is a joint endeavor of the NuPIC community and the Epilepsy Seizure Prediction Season of NuPIC Project, submitted by Anubhav Chaturvedi.

Kaggle Rules

Please read the entire set of rules linked above.

Repo

https://github.com/fergalbyrne/nupic.kaggle-eeg

Team Members

Kaggle Team

Planning Meeting: May 30

Video here (unlisted).

Meeting Notes (taken by Matt)

Data

  • James create a tool to convert kaggle data into csv
  • Fergal thinks RDSE is the right encoder

Swarming

  • Will take too long to swarm over all the columns
  • Might give useful params to swarm over a subset of columns or even just one column

Classification

  • two models: one regular patterns, one for pre-seizure
  • training data fed to each model
  • test data sent to both
  • generate classification based upon anomaly scores

Spitting up the data

  • training
  • cross-validation
  • test

Tasks

  • James will put up code to parse input data into CSV
  • All of use will try to come up with model params
  • Matt will post meeting minutes and email ML
  • Fergal created tickets