Spikefinder submission

Reproduces my final submission to Spikefinder.

Steps to retrain model from scratch:

• Clone this repo
• cd into spikefinder_submission and clone mixture density network model via git clone https://github.com/patrickmineault/write-rnn-tensorflow.git write_rnn_tensorflow
• Train mixture density network latent features via python train_and_eval_mdn.py --operation=train
• Eval mixture density network latent features via python train_and_eval_mdn.py --operation=eval
• Fit core model and refine for each of the first 5 recording sets by running ./train_and_refine.sh
• Eval for the unrefined and the refined model via ./eval_all.sh

Steps to predict using pretrained model via Docker.

• Clone this repo.
• sudo docker build -t patrickmineault:spikefinder .
• sudo docker run -it patrickmineault:spikefinder