Skip to content

Latest commit

 

History

History
executable file
·
37 lines (21 loc) · 1.44 KB

README.MD

File metadata and controls

executable file
·
37 lines (21 loc) · 1.44 KB
Raw

SPARKS 🎆

This is the official repository for SPARKS: a Sequential Predictive Autoencoder for the Analysis of spiKing Signals.

SPARKS includes a novel self-attention mechanism using Hebbian learning to generate reliable latent representations from single spike timings. SPARKS trains a variational autoencoder with a novel criterion inspired by predictive coding for temporal coherence.

sparks is implemented in PyTorch and includes demos for a quickstart. It can perform supervised or unsupervised to produce low-dimensional latent embeddings which allows to gain biological insights from neural data.

Make sure to 👀 watch or ⭐️ star this repository to keep updated!

Reference

  • 📄 Preprint:

Available now on BiorXiv!

License

  • SPARKS is an open source software under a GLPv3 license.