Reproduce “Paired Competing Neurons Improving STDP Supervised Local Learning In Spiking Neural Networks” 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
Gaspard Goupy, Pierre Tirilly, Ioan Marius Bilasco
Stabilized Supervised STDP (S2-STDP) and a Paired Competing Neurons (PCN) architecture are introduced to improve local learning in single-spike Spiking Neural Networks for classification. These methods enhance accuracy by up to 2.34 percentage points on MNIST and significantly increase the frequency of weight updates, addressing key limitations of prior supervised STDP techniques.
Key insights
- Dynamically setting specific target and non-target firing timestamps, instead of ranges, is critical for enabling frequent and precise STDP weight adjustments, leading to more robust and accurate learning.
- Introducing paired competing neurons encourages specialization, allowing individual neurons to better adapt to either target or non-target firing conditions, thereby improving overall classification performance.
- Local learning rules can achieve competitive classification performance on standard image datasets when carefully designed to control neuron firing times and integrate architectural enhancements that manage learning conflicts.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.