Reproduce “Vestibular reservoir computing” 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
Smita Deb, Shirin Panahi, Mulugeta Haile, Ying-Cheng Lai
A bio-inspired framework termed vestibular reservoir computing utilizes an uncoupled network architecture for time-series prediction. This system achieves predictive performance on chaotic Lorenz and food-chain systems comparable to traditionally coupled networks, with similar memory capacities observed when their eigenvalue spectra are matched.
Key insights
- Uncoupled reservoir network architectures can achieve predictive performance comparable to, or even exceeding, coupled networks, simplifying physical hardware implementation.
- The memory function and total memory capacity of linear reservoirs are determined by the full eigenvalue spectrum of the internal weight matrix, not solely its spectral radius.
- For nonlinear vestibular reservoir computing, matching the eigenvalue spectra of coupled and uncoupled networks leads to similar memory capacities and thus comparable predictive performance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.