snnTorch learning

Personal playground to learn how snnTorch works and to create some interesting models for the future.

It might not have the best practices, but it might help someone.

Current status

• Iris dataset with SNN and 100% accuracy, original measures
• Iris dataset, but with spikegen used to create spikes for the measures
  • With rate coding, accuracy dropped to 46.67% due to lost information
  • When we do rate coding with linear interpolation, we're at 100% again (in this case, no need for MinMaxScaler)
  • Latency coding gives 46.67% again
  • Interestingly, delta coding gives 100% (no need for MinMaxScaler)
  • Note that the 100% numbers can vary from time to time by one sample, while this wasn't observer with original measures