Quick summary

This is a general framework for implementing neural networks. The main feature is that the behavior of neurons and synapses is explicitly implemented by the user by providing the appropriate differential equations and event handlers. This makes nnet2 highly flexible, allowing to implement anything from firing-rate-based simulations to networks of Hodgkin-Huxley neurons with complicated STDP rules.

The architecture of nnet2 is strongly inspired by Brian. Brain, however, uses strings to describe the equations governing the evolution of the neural network, while nnet2 uses normal Python functions. The advantage of the former is conciseness, but the disadvantage is more awkwardness in interacting with the Python environment, and significantly higher overheads.


This is now version 0.5.

How do I get set up?

Summary of set up

Just copy in the folder where you want to use it, and you're good to go.


The only requirement for nnet2 is numpy.

How to run tests

Run the script.