Reproduce “End-to-End Learning of Neuromorphic Wireless Systems for Low-Power Edge Artificial Intelligence” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Nicolas Skatchkovsky, Hyeryung Jang, Osvaldo Simeone
Researchers at King’s College London’s KCLIP lab developed NeuroJSCC, an "all-spike" end-to-end neuromorphic system for remote wireless inference that natively handles data in the spike domain, integrating SNNs and Impulse Radio. This approach achieved high classification accuracy on the MNIST-DVS dataset, such as 88% at -6 dB SNR, enabling low-latency, energy-efficient processing for edge AI applications.
Key insights
- An "all-spike" system design, where sensing, processing, and communication occur natively in the spike domain, offers inherent energy efficiency and low-latency benefits for edge AI.
- Probabilistic Spiking Neural Networks (SNNs) based on the Generalized Linear Model enable end-to-end training of complex neuromorphic systems, overcoming the non-differentiability challenge of deterministic SNNs.
- Joint Source-Channel Coding (JSCC) implemented via SNNs (NeuroJSCC) outperforms traditional separate source-channel coding (SSCC) and uncoded transmission, particularly in low Signal-to-Noise Ratio (SNR) wireless environments.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.