/device_matched_SNN_sims

Primary LanguagePythonGNU General Public License v3.0GPL-3.0

device_matched_SNN_sims

This is an snntorch based implementation of SNN training with real spiking neuron device characteristics approximated by suitable parameters of an LIF model. A 10-layer spiking CNN is trained on CIFAR-10 data with 8-bit quantized weights

Requirements

Pytorch, numpy, snntorch, matlab

Description

  1. Spiking_CNN_quant_training_with_theta_beta_variations_8b_2.py - trains a 10-layer Spiking convolutional neural network training with 8-bit weights for CIFAR-10
  2. curr_driven_SNNmodeling.py - models spiking neuron device characteristics as LIF neurons, Capacitance determined by inspection as of now
  3. iris_classification_matlab - old matlab code for training 1 layer SNN for Fisher-iris classification with STDP, main script - testScript3.m