NEMO (Neural Encoding Models for Ophys Data)

Setup

Prerequisites

• Required
  • Anaconda
• Optional
  • Cuda Toolkit >= 10.2
  • Cudnn >= 7.6.5

Create an Anaconda Environment

Begginning with anaconda already installed, create an environment for this repo

conda create -n nemo python=3.7

Next, activate the newly-created conda environment

conda activate nemo

Finally, install GNU Octave, Lua, JupyterLab, and Google-Protobuf in this environment (all other libraries will be installed to this environment below when running setup.py).

conda install -c conda-forge octave lua jupyterlab protobuf

Cloning the Repository and Installing Python Packages

First, clone this repository (https shown) and run setup.py.

git clone git@github.com:MichaelTeti/NEMO.git &&
cd NEMO &&
python -m pip install --editable .

Install PetaVision

Install PetaVision with the install_openpv.sh script.

bash scripts/install_openpv.sh <dir-to-install-in>

The script takes one positional argument, which is the directory you wish to install in (make sure you have permissions in this dir). You may need to first install some of PetaVision's dependencies (listed here), which can be installed via most Linux-based package managers.

Downloading and Processing the Data

1. Download and pre-process the ImageNet video dataset (directions here)
2. Download and extract the neurodata and stimuli (directions here)

Experiments

1. LCA Convolutional Dictionary Learning on Video
2. LCA Simple Cell Dictionary Learning on Video