GPU-enhanced neural networks

Fancy running your SNNs 10x faster? Our GPU enhanced Neuronal Networks (GeNN) library is freely available from https://genn-team.github.io/ and provides an environment for GPU accelerated spiking neural network simulations. GeNN is capable of simulating large spiking neural network (SNN) models at competitive speeds on commodity and even embedded GPUs. In GeNN, SNN models are described using a simple model description API through which variables, parameters and C-like code snippets that describe various aspects of the model elements can be specified, e.g. neuron and synapse update equations or learning dynamics. Model elements of neuron and synapse types are combined into neuron and synapse populations to form a full spiking neural network model. GeNN takes the model description and generates optimised, event-driven code to simulate the model. Current code-generation backends include CUDA for NVIDIA GPUs as well as a C++ CPU-only mode.

We have recently completed work on GeNN 5.0.0 which adds a flexible structural plasticity framework, reduces model build times and offers a much improved user experiance. Also we have built mlGeNN (https://github.com/genn-team/ml_genn/), which lets you easily explore spike-based machine learning from a simple Keras-like API using our efficient GeNN implementations of the e-prop and EventProp learning rules.

This year at CapoCaccia, we're going to run sessions to introduce people to GeNN and mlGeNN, help them with installation problems and then walk them through our selection of tutorials that should allow them to get a flavour of the capabilities and user experience of GeNN; and enable them to start using it for their own work. CapoCaccia WiFi permitting, all our tutorials are now available as Google Collab notebooks at https://genn-team.github.io/tutorials.html so no installation is required!

Go to group wiki Go to wiki users Info

Timetable

Moderator

Members