Reproduce “Visual Pattern Recognition with on On-chip Learning: towards a Fully Neuromorphic Approach” 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
Sandro Baumgartner, Alpha Renner, Raphaela Kreiser, Dongchen Liang, Giacomo Indiveri, Yulia Sandamirskaya
Researchers at the University of Zurich and ETH Zurich developed a spiking neural network on Intel's Loihi chip that performs visual pattern recognition with fully on-chip, spike-based learning. This system achieved near 100% classification accuracy, learned new patterns in under 200 ms, and demonstrated robustness against 130% input noise.
Key insights
- It is feasible to achieve fully on-chip, real-time visual pattern learning and recognition on neuromorphic hardware using local, spike-based plasticity rules.
- The proposed SNN architecture effectively integrates an arbitration mechanism, activity normalization, and a scaling mechanism to allow simultaneous pattern identity classification and estimation of location and size.
- Neuromorphic systems can be engineered to be highly resilient to both external input noise and internal neuronal parameter variability, demonstrating their potential for reliable deployment in noisy real-world environments.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.