Reproduce “A theoretical basis for efficient computations with noisy spiking neurons” 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
Zeno Jonke, Stefan Habenschuss, Wolfgang Maass
The authors design a noisy spiking network that shapes probabilities of short-window states to explore constrained assignments without fixed neuron codes. Winner-take-all and OR motifs construct energy landscapes, while spike timing alters transitions. In TSP tests, transitions beyond ±15 were frequent, and fewer state changes reached given-cost tours than with matched continuous-time Gibbs sampling, without showing faster runtime or energy savings.
Key insights
- Noise in spiking neurons can be a computational resource, driving exploration in an energy landscape defined by the network's stationary state distribution.
- The asymmetric dynamics of spike generation (stochastic on-transition, deterministic off-transition) provide a computational advantage over symmetric samplers by facilitating the bypassing of high energy barriers.
- Complex energy functions, representing constraint satisfaction problems, can be modularly constructed by combining simple, stereotypical spiking circuit motifs.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.