dmcnamee / flexmodehc

Flexible modulation of sequence generation in the entorhinal-hippocampal system.

Introduction

This code was used for studying the flexible modulation of sequence generation in the entorhinal-hippocampal system.

• Nature Neuroscience
• Non-paywall version (view only)

Installation using Anaconda

conda create -n FlexModEHC python=3.8
conda activate FlexModEHC
conda install numpy scipy pandas seaborn networkx scikit-learn numba
conda install -c conda-forge gym
conda install -c pyviz holoviews
pip install git+https://github.com/zuoxingdong/mazelab.git
git clone https://github.com/dmcnamee/FlexModEHC.git

FIGURE_S8 requires torch and opencv

conda install pytorch
conda install -c conda-forge opencv

Explanation

• Package is based on a set of core classes which form a chain of inheritances:
  ENVIRONMENT -> GENERATOR -> PROPAGATOR -> SIMULATOR -> EXPLORER/LEARNER
• A GENERATOR is constructed from an ENVIRONMENT. For example, an environment transition matrix may be used to form a generator matrix.
• A PROPAGATOR takes a GENERATOR (along with several parameters as arguments) and uses eigen-decompositions to form a matrix solution to the master equation defined by the GENERATOR.
• A SIMULATOR uses a PROPAGATOR to sample trajectories in the ENVIRONMENT.
• An EXPLORER samples trajectories from SIMULATOR and performs search process analyses.
• A LEARNER samples trajectories from SIMULATOR and learns internal models and reward functions.
• Each of these classes comes equipped with documented member functions.