EIANN Documentation

Note

🚧 Work in Progress: This documentation site is currently under construction. Content may change frequently.

EIANN Documentation#

A framework for training neural networks with E/I cell types and biologically-plausible learning rules

About EIANN#

EIANN (Excitatory/Inhibitory Artificial Neural Networks) is a PyTorch-based library designed to build and train rate-based biological neural networks containing multiple layers of recurrently connected Excitatory (E) and Inhibitory (I) cell types.

EIANN was created to accelerate research within computational neuroscience with a focus on bio-inspired learning rules. We provide a neural network syntax that is intuitive from a neuroscience perspective, organized around neuronal populations and their projections.

We specifically designed EIANN to make reproducible experiments, hyperparameter optimization, and neural architecture search easier by providing a simple YAML-based configuration file interface for specifying network architecture, training parameters, learning rules, and cell type constraints. EIANN allows you to easily specify biological constraints and mechanisms such as:

  • Dale’s Law: Enforces biologically realistic constraints on connections between E and I cell types

  • Local Learning Rules: You can create arbirary learning rules at any projection in the network, or just use one of the existing gradient-based, dendrite-based, or Hebbian learning mechanisms.

  • E/I Cell Types: Explicit modeling of diverse neural populations

  • Recurrent Connections: Connections can be made either recurrent or feedforward and can arbitrarily connect any two neural populations

In addition to the above, EIANN provides a range of analysis and visualization tools to help you understand your network’s behavior and learning dynamics.

Publications#

EIANN is based on research published in:

Galloni A.R., Peddada A., Chennawar Y., Milstein A.D. (2025)
Cellular and subcellular specialization enables biology-constrained deep learning.
bioRxiv 2025.05.22.655599
https://doi.org/10.1101/2025.05.22.655599

Use Cases#

EIANN is particularly well-suited for:

  • Computational Neuroscience Research: Understanding how biological constraints affect learning

  • Biologically-Inspired AI: Developing AI systems that incorporate brain-like mechanisms

  • Educational Applications: Teaching neural network principles with biological realism

Getting Started#

  1. Installation: Set up EIANN in your environment

  2. Quick Start: Build your first E/I network

  3. Tutorials: Work through detailed examples

  4. User Guide: Learn about all features

Support#

EIANN is developed by the [Milstein Lab] at Rutgers University as part of ongoing research into biologically-constrained deep learning.

Contents