Combinatorial optimization by weight annealing in memristive hopfield networks

被引:14
|
作者
Fahimi, Z. [1 ]
Mahmoodi, M. R. [1 ]
Nili, H. [1 ]
Polishchuk, Valentin [2 ]
Strukov, D. B. [1 ]
机构
[1] UC Santa Barbara, Santa Barbara, CA 93106 USA
[2] Linkoping Univ, S-60174 Norrkoping, Sweden
关键词
NEURAL-NETWORKS;
D O I
10.1038/s41598-020-78944-5
中图分类号
O [数理科学和化学]; P [天文学、地球科学]; Q [生物科学]; N [自然科学总论];
学科分类号
07 ; 0710 ; 09 ;
摘要
The increasing utility of specialized circuits and growing applications of optimization call for the development of efficient hardware accelerator for solving optimization problems. Hopfield neural network is a promising approach for solving combinatorial optimization problems due to the recent demonstrations of efficient mixed-signal implementation based on emerging non-volatile memory devices. Such mixed-signal accelerators also enable very efficient implementation of various annealing techniques, which are essential for finding optimal solutions. Here we propose a "weight annealing" approach, whose main idea is to ease convergence to the global minima by keeping the network close to its ground state. This is achieved by initially setting all synaptic weights to zero, thus ensuring a quick transition of the Hopfield network to its trivial global minima state and then gradually introducing weights during the annealing process. The extensive numerical simulations show that our approach leads to a better, on average, solutions for several representative combinatorial problems compared to prior Hopfield neural network solvers with chaotic or stochastic annealing. As a proof of concept, a 13-node graph partitioning problem and a 7-node maximum-weight independent set problem are solved experimentally using mixed-signal circuits based on, correspondingly, a 20 x 20 analog-grade TiO2 memristive crossbar and a 12 x 10 eFlash memory array.
引用
收藏
页数:10
相关论文
共 50 条
  • [1] Combinatorial optimization by weight annealing in memristive hopfield networks
    Z. Fahimi
    M. R. Mahmoodi
    H. Nili
    Valentin Polishchuk
    D. B. Strukov
    Scientific Reports, 11
  • [2] Classical Adiabatic Annealing in Memristor Hopfield Neural Networks for Combinatorial Optimization
    Kumar, Suhas
    Van Vaerenbergh, Thomas
    Strachan, John Paul
    2020 INTERNATIONAL CONFERENCE ON REBOOTING COMPUTING (ICRC 2020), 2020, : 76 - 79
  • [3] Combinatorial optimization by Hopfield networks using adjusting neurons
    Liang, Y
    INFORMATION SCIENCES, 1996, 94 (1-4) : 261 - 276
  • [4] Combinatorial Optimization in Hopfield Networks with Noise and Diagonal Perturbations
    Yi, Su-In
    Kumar, Suhas
    Williams, R. Stanley
    2022 IEEE INTERNATIONAL SYMPOSIUM ON CIRCUITS AND SYSTEMS (ISCAS 22), 2022, : 321 - 325
  • [5] In-Materia Annealing and Combinatorial Optimization Based on Vertical Memristive Array
    Lee, Soo Hyung
    Cheong, Sunwoo
    Cho, Jea Min
    Ghenzi, Nestor
    Shin, Dong Hoon
    Jang, Yoon Ho
    Han, Janguk
    Park, Tae Won
    Kim, Dong Yun
    Shim, Sung Keun
    Han, Joon-Kyu
    Kim, Seung Soo
    Hwang, Cheol Seong
    ADVANCED MATERIALS, 2024, 36 (40)
  • [6] SOLVING INEQUALITY CONSTRAINED COMBINATORIAL OPTIMIZATION PROBLEMS BY THE HOPFIELD NEURAL NETWORKS
    ABE, S
    KAWAKAMI, J
    HIRASAWA, K
    NEURAL NETWORKS, 1992, 5 (04) : 663 - 670
  • [7] Theoretical considerations on the dynamics of hysteresis binary Hopfield networks for combinatorial optimization
    Matsuda, S
    IJCNN 2000: PROCEEDINGS OF THE IEEE-INNS-ENNS INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS, VOL VI, 2000, : 480 - 485
  • [8] Extended Hopfield Model of Neural Networks for Combinatorial Multiobjective Optimization Problems
    Balicki, J
    Kitowski, Z
    Stateczny, A
    IEEE WORLD CONGRESS ON COMPUTATIONAL INTELLIGENCE, 1998, : 1646 - 1651
  • [9] Extended hopfield models for combinatorial optimization
    Le Gall, A
    Zissimopoulos, V
    IEEE TRANSACTIONS ON NEURAL NETWORKS, 1999, 10 (01): : 72 - 80
  • [10] Dynamical analysis of continuous higher-order hopfield networks for combinatorial optimization
    Atencia, M
    Joya, G
    Sandoval, F
    NEURAL COMPUTATION, 2005, 17 (08) : 1802 - 1819