Sigma-Delta Neural Network Conversion on Loihi 2

2025-05-09

AI Summary

This work addresses temporal redundancy and low hardware efficiency in artificial neural network (ANN) to spiking neural network (SNN) conversion. We propose a novel conversion method based on Sigma-Delta (ΣΔ) modulation—the first application of ΣΔ principles to SNN spike encoding design. Leveraging Loihi 2’s hierarchical spike encoding capability, our approach replaces conventional binary rate coding with multi-level quantized spikes, preserving spatiotemporal sparsity while substantially reducing the number of inference time steps. Experiments demonstrate that deploying our method on Loihi 2 achieves zero-accuracy-loss inference with significantly lower energy consumption and latency compared to Jetson Xavier—an edge AI platform. The core contribution is the introduction of a ΣΔ-driven hierarchical spike conversion paradigm, establishing a scalable, co-designed software–hardware optimization pathway for efficient neuromorphic inference.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Machine Learning: Learning on the Edge & Model Compression
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Neuromorphic computing aims to improve the efficiency of artificial neural networks by taking inspiration from biological neurons and leveraging temporal sparsity, spatial sparsity, and compute near/in memory. Although these approaches have shown efficiency gains, training these spiking neural networks (SNN) remains difficult. The original attempts at converting trained conventional analog neural networks (ANN) to SNNs used the rate of binary spikes to represent neuron activations. This required many simulation time steps per inference, which degraded efficiency. Intel's Loihi 2 is a neuromorphic platform that supports graded spikes which can be used to represent changes in neuron activation. In this work, we use Loihi 2's graded spikes to develop a method for converting ANN networks to spiking networks, which take advantage of temporal and spatial sparsity. We evaluated the performance of this network on Loihi 2 and compared it to NVIDIA's Jetson Xavier edge AI platform.

Problem

• Convert analog neural networks to efficient spiking networks
• Leverage Loihi 2's graded spikes for improved conversion
• Compare performance between neuromorphic and edge AI platforms

Innovation

• Uses Loihi 2's graded spikes for ANN to SNN conversion
• Leverages temporal and spatial sparsity for efficiency
• Compares performance with NVIDIA's Jetson Xavier platform