ibm-research-tokyo / diffsnn

An implementation of a differentiable point process and a differentiable spiking neural network.

Differentiable Spiking Neural Networks

This repository contains Python implementation of a spiking neural network based on point processes. In particular, a differentiable point process is implemented, which enables us to train a spking neural network with hidden units efficiently.

How to install

Run

pip install .

Experiments

See tasks/synthetic for experiments using a synthetic data set.

Reference

Hiroshi Kajino: "A Differentiable Point Process with Its Application to Spiking Neural Networks", ICML-21 [preprint].