The material I learned from while writing the Spiking Neural Networks notes. Each post also lists the original papers behind its content at the end.
Course
- Munich Neuromorphic Hackathon, with Michael Neumeier (fortiss). A hands-on hackathon run by fortiss together with the student initiative NeuroTUM, in which interdisciplinary student teams build neuromorphic systems for practical industry challenges, alongside expert talks. Michael Neumeier, a researcher in fortiss’s Neuromorphic Computing team (profile), works on energy-efficient AI built from neuromorphic sensors, spiking algorithms and hardware, such as real-time gesture and action recognition with spiking networks on Intel’s Loihi chip. His talk at the hackathon covers neuromorphic computing for industrial use cases.
Books and papers
- Training Spiking Neural Networks Using Lessons From Deep Learning: Jason K. Eshraghian et al. Proceedings of the IEEE, 2023. arXiv:2109.12894. A tutorial-style paper on surrogate gradients, neural coding and online learning, with the snnTorch tutorials.
- Spiking Neuron Models: Wulfram Gerstner and Werner M. Kistler. Cambridge University Press, 2002. The neuron models (integrate-and-fire and beyond) in depth.
- Event-based Vision: A Survey: Guillermo Gallego et al. IEEE TPAMI, 2022. arXiv:1904.08405. Event cameras, their properties and algorithms.
- Surrogate Gradient Learning in Spiking Neural Networks: Emre O. Neftci, Hesham Mostafa and Friedemann Zenke. IEEE Signal Processing Magazine, 2019. arXiv:1901.09948. The short paper that named and explained the method behind Training Spiking Networks.
- The NeuroBench Framework for Benchmarking Neuromorphic Computing Algorithms and Systems: Jason Yik et al. Nature Communications, 2025. arXiv:2304.04640. How to count and compare the work of neuromorphic models, used in Counting Energy.
Datasets
- DVS Gesture: Arnon Amir et al., CVPR 2017. 11 gestures by 29 people, recorded with a 128 × 128 event camera; CC BY 4.0. Used in Training Spiking Networks and Counting Energy.
- N-MNIST: Garrick Orchard et al., Frontiers in Neuroscience, 2015. MNIST digits recorded by an event camera making three saccades, the idea simulated in From ReLU to Spikes.
Tools used in these notes
- BrainChip Akida (akida, CNN2SNN, QuantizeML) for quantizing and converting a CNN for the Akida neuromorphic processor and running it in its simulator, used in From ANN to Akida.
- PyTorch for every network trained here, and snnTorch for the spiking neurons that NeuroBench recognises.
- NeuroBench for counting operations, activation sparsity and neuron updates.
- Tonic for event-camera datasets, and v2e for converting video to events.