Related Experiment Video
Updated: May 4, 2026
A Simple Stimulatory Device for Evoking Point-like Tactile Stimuli: A Searchlight for LFP to Spike Transitions
Published on: March 25, 2014
Solving Max-Cut Problem Using Spiking Boltzmann Machine Based on Neuromorphic Hardware with Phase Change Memory
Yu Gyeong Kang1, Masatoshi Ishii2, Jaeweon Park1
1Department of Material Science & Engineering, Inter-University Semiconductor Research Center, Research Institute of Advanced Materials, Seoul National University, Seoul, 08826, Republic of Korea.
This study introduces a novel hardware-friendly method using spiking neural networks (SNNs) to efficiently solve complex combinatorial optimization problems like Max-Cut on neuromorphic chips. The approach demonstrates effective convergence and high accuracy for large-scale problems.
Area of Science:
- Neuromorphic Engineering
- Computational Neuroscience
- Optimization Algorithms
Background:
- Combinatorial optimization problems (COPs), such as Max-Cut, are computationally intensive, with resource requirements growing exponentially with problem size.
- Existing methods struggle with scalability and efficiency for large-scale COPs.
- Neuromorphic hardware offers a promising platform for energy-efficient computation.
Purpose of the Study:
- To propose a hardware-friendly method for solving the Max-Cut problem using spiking neural networks (SNNs) implemented on neuromorphic hardware.
- To analyze the stochastic dynamics of leaky integrate-and-fire (LIF) neurons for hardware implementation of spiking Boltzmann machines (sBMs).
- To develop an innovative algorithm for efficient and accurate solutions to large-scale COPs.
Main Methods:
- Implementation of a spiking neural network (SNN)-based Boltzmann machine (BM) in neuromorphic hardware.
- Analysis of stochastic dynamics of leaky integrate-and-fire (LIF) neurons with random walk noise.
- Development of an innovative algorithm utilizing overlapping time windows for sBM.
- Hardware validation on a 6-transistor/2-resistor (6T2R) neuromorphic chip with phase change memory (PCM) synapses.
Main Results:
- Demonstrated effective convergence and high accuracy for large-scale Max-Cut problems through simulations.
- Successful hardware implementation and validation on a custom neuromorphic chip.
- Proposed annealing techniques and bias split methods to enhance convergence.
- Introduced circuit design ideas for efficient sampling convergence evaluation.
Conclusions:
- The proposed SNN-based neuromorphic hardware approach offers a potential solution for energy-efficient and hardware-implementable solving of COPs.
- This work represents the first known instance of solving the Max-Cut problem using an SNN neuromorphic hardware chip.
- The findings pave the way for practical applications of neuromorphic computing in complex optimization tasks.
