This is an snntorch implementation of Liquid State Machine (LSM) networks. The parameters of the example network implemented in main.py are derived from the paper "MAdapter: A Multimodal Adapter for Liquid State Machines configures the Input Layer for the same Reservoir to enable Vision and Speech Classification" link
Pytorch, Tonic, numpy, sklearn and snntorch
- lsm_weight_definitions.py - contains definitions of connectivity (Input->Reservoir and Recurrent Reservoir weights)
- lsm_models.py - contains the LSM model definition.
- main.py - contains an example implementation with the N-MNIST dataset. Network execution must be run with torch.no_grad() for LSM operation
- main.ipynb - same as main.py but for running in Google Colab. (Make sure to upload this repository folder to in Google Drive and modify the path in the 3rd cell accordingly to run)