SNN-IIR

This repo contains the implementation of paper Exploiting Neuron and Synapse Filter Dynamics in Spatial Temporal Learning of Deep Spiking Neural Network. It trains spiking neural network to learn spatial temporal patterns.

Run

Just simply clone the repo. A few examples are provided:

• snn_mlp_1.py
  A multi-layer fully connected SNN to classify MNIST. It uses dual exponential PSP kernel.
  Following command trains the SNN using the configurations stored in snn_mlp_1.yaml

  python snn_mlp_1.py --train

Following command loads the pretrained model ./checkpoint/pretrained_snn_mlp_1 and and test it.

  python snn_mlp_1.py --test

• snn_mlp_2.py
  A multi-layer fully connected SNN to classify MNIST. It uses first order low pass PSP kernel. A pretrained checkpoint locates in ./checkpoint/pretrained_snn_mlp_2. Configuration file is snn_mlp_2.yaml.

  python snn_mlp_2.py --train

Following command loads the pretrained model and test it.

  python snn_mlp_2.py --test

• associative_memory.py
  A multi-layer fully connected SNN which reconstructs input patterns at output layer. A pretrained model locates in ./associative_memory_checkpoint/pretrained_associative_memory. associative_memory.ipynb is the notebook version to inspect input and output.

File Organization

• snn_lib: Spiking neural network layers, data loaders,utility functions etc.
• checkpoint: pretrained models.
• dataset: data used for associative memory experiments.
• associative_memory_checkpoint: pre-trained model of associative_memory.py.

Dependencies

torch==1.2.0
numpy==1.19.0
omegaconf==1.4.1

Citation

If this is helpful for you, please cite the following paper:

@inproceedings{ijcai2020-0388,
  title     = {Exploiting Neuron and Synapse Filter Dynamics in Spatial Temporal Learning of Deep Spiking Neural Network},
  author    = {Fang, Haowen and Shrestha, Amar and Zhao, Ziyi and Qiu, Qinru},
  booktitle = {Proceedings of the Twenty-Ninth International Joint Conference on
               Artificial Intelligence, {IJCAI-20}},
  publisher = {International Joint Conferences on Artificial Intelligence Organization},             
  editor    = {Christian Bessiere}	
  pages     = {2799--2806},
  year      = {2020},
  month     = {7},
  note      = {Main track}
  doi       = {10.24963/ijcai.2020/388},
  url       = {https://doi.org/10.24963/ijcai.2020/388},
}