global-mean-amplitude-snn-cim

name

global-mean-amplitude-snn-cim

description

Global mean-amplitude feedback-enhanced spiking neural network coherent ising machine (GFSNN-CIM) with physics-driven amplitude stabilization. Solves Max-Cut with 27% improvement vs conventional SNN-CIM, validated on traffic assignment problems. Based on Jiang, Ma, Wang & Wang (arXiv: 2509.13917). Use when solving combinatorial optimization with spiking neural networks, implementing coherent ising machines, or applying mean-amplitude feedback stabilization to SNN optimizers.

Global Mean-Amplitude Enhanced SNN Coherent Ising Machine

Physics-driven amplitude stabilization for spiking neural network-based coherent Ising machines. Based on Jiang et al. (arXiv: 2509.13917).

Core Methodology

• Global mean-amplitude feedback enhances spiking neural network CIM
• Physics-driven amplitude stabilization prevents oscillation divergence
• 27% improvement in Max-Cut solution success rates vs conventional SNN-CIM
• Validated on traffic assignment problems (generalizes beyond Max-Cut)

Architecture

1. SNN encodes Ising model spins as neuron states
2. Global mean-amplitude feedback stabilizes collective dynamics
3. Physics-driven amplitude correction enforces energy landscape convergence
4. Readout maps final spiking states to optimization solution

When to Use

• Combinatorial optimization (Max-Cut, graph partitioning, traffic assignment)
• Coherent Ising Machine implementations
• Spiking neural network optimizers
• Physics-inspired neural computation
• Problems requiring amplitude stabilization in oscillatory networks

Key Advantages

1. 27% success rate improvement over baseline SNN-CIM
2. Generalizes beyond Max-Cut to real-world optimization
3. Physics-driven (not learned) stabilization — no training required
4. Compatible with existing SNN hardware implementations

Related Concepts

• Coherent Ising Machines (CIM)
• Spiking Neural Networks (SNN)
• Combinatorial Optimization
• Mean-Field Feedback
• Amplitude Stabilization

Activation: coherent ising machine, GFSNN-CIM, mean-amplitude feedback, spiking neural optimizer, physics-driven stabilization, Max-Cut SNN, arXiv:2509.13917