Complex Priors and Flexible Inference in Recurrent Circuits with Dendritic Nonlinearities

This repo contains code that implements the model proposed in the paper "Complex Priors and Flexible Inference in Recurrent Circuits with Dendritic Nonlinearities", in ICLR 2024.

Here is an overview of the different directories:

• the demos directory contains jupyter notebooks that generate figures found in the main text and the supplementary.
• the core directory contains:
  • saved_arrays (data used for analysis)
  • saved_weights (the saved weights of the trained model)
  • utils (utilities used throughout the repo)
  • models.py (collection of model architectures)
  • model_description (descriptions of the parameters of every trained model)
  • a collection of python scripts for training models or generating samples from models on a high performance computing cluster.