Scalable In-Memory Clustered Annealer with Temporal Noise of FinFET for the Travelling Salesman Problem
Anni Lu, Jae Hyun Hur, Yuan-Chun Luo, Hai Li, Dmitri E. Nikonov, Ian A. Young, Yang‐Kyu Choi, Shimeng Yu · 2022 International Electron Devices Meeting (IEDM) · 2022
We propose a scalable in-memory annealer for solving the large-scale travelling salesman problem (TSP), based on a Hopfield neural network (HNN) implemented with a crossbar array of FinFET, using its intrinsic temporal noise of the drain current caused by trapping/detrapping to realize the annealing process. A hierarchical clustered approach is adopted to overcome the scalability challenge of solving large-scale TSP. We speed up the system convergence by only allowing feasible permutation states for the spins update, and updating non-adjacent clusters in parallel with equal size k-means clustering. The proposed hardware is tested with up to 150-city problems and evaluated by the system-level simulation. It shows benefits on fast convergence and ultra-low energy compared with state-of-the-art annealers.