Genetic Motifs as a Blueprint for Mismatch-Tolerant Neuromorphic Computing

TL;DR

The proposed method mitigates mismatch-induced noise without requiring precise device mismatch measurements, effectively outperforming alternative hardware-aware techniques proposed in the literature, and providing a more general solution for improving the robustness of SNNs in neuromorphic hardware.

Abstract

Mixed-signal implementations of SNNs offer a promising solution to edge computing applications that require low-power and compact embedded processing systems. However, device mismatch in the analog circuits of these neuromorphic processors poses a significant challenge to the deployment of robust processing in these systems. Here we introduce a novel architectural solution inspired by biological development to address this issue. Specifically we propose to implement architectures that incorporate network motifs found in developed brains through a differentiable re-parameterization of weight matrices based on gene expression patterns and genetic rules. Thanks to the gradient descent optimization compatibility of the method proposed, we can apply the robustness of biological neural development to neuromorphic computing. To validate this approach we benchmark it using the Yin-Yang classification dataset, and compare its performance with that of standard multilayer perceptrons trained with state-of-the-art hardware-aware training method. Our results demonstrate that the proposed method mitigates mismatch-induced noise without requiring precise device mismatch measurements, effectively outperforming alternative hardware-aware techniques proposed in the literature, and providing a more general solution for improving the robustness of SNNs in neuromorphic hardware.

Figures (4)

• Figure 1: Our spiking neural architecture, which exhibits mismatch-tolerant computing capabilities through bio-inspired network motifs introduced via re-parameterization of the weight matrices.
• Figure 2: A fully-connected bi-partite motif example, where two clusters of neurons--the $\bullet$ and $\bullet$ nodes--have no intra-cluster connections but are fully connected between clusters.
• Figure 3: The Yin-Yang classification dataset 10.1145/3517343.3517380. Left: the original dataset. Right: a visual representation of our rate encoded version.
• Figure 4: A box plot showing the accuracy distributions of models tested in our simulated on-chip deployment. Noise injection is characterized by a coefficient of variation of 10% ($\alpha = 0.1$). The two groups of boxes represent different network sizes, and models are color-coded. Statistical annotations are provided at the top of the plot.