Ground States of the Mean-Field Spin Glass with 3-Spin Couplings

被引:1
作者
Boettcher, Stefan [1 ]
Lau, Ginger E. [1 ]
机构
[1] Emory Univ, Dept Phys, Atlanta, GA 30322 USA
关键词
RANDOM-ENERGY MODEL; SOLVABLE MODEL; OPTIMIZATION;
D O I
10.1103/j159-lpfx
中图分类号
O4 [物理学];
学科分类号
0702 ;
摘要
We use heuristic optimization methods in extensive computations to determine with low systematic error ground-state configurations of the mean-field p-spin glass model with p 1/4 3. Here, all possible triplets in a system of N Ising spins are connected with a bond. This model has been of recent interest, since it exhibits the "overlap gap condition," which should make it prohibitive to find ground states asymptotically with local search methods when compared, for instance, with the p 1/4 2 case better known as the Sherrington-Kirkpatrick model (SKM). Indeed, it proves more costly to find good approximations for p 1/4 3 than for the SKM, even for our heuristic. Compared to the SKM, the ground-state behavior for p 1/4 3 is quite distinct also in other ways. For the SKM, finite-size corrections for large system sizes N-* 00, of both the ensemble average over ground-state energy densities and the width of their distribution, vary anomalously with noninteger exponents. In the p 1/4 3 case here, the energy density and its distribution appear to scale with lnN/N and 1/N corrections, respectively. The distribution itself is consistent with a Gumbel form. Even more stark is the contrast for the bond-diluted case, where the SKM has shown previously a notable variation of the anomalous corrections exponent with the bond density, while for p 1/4 3 no such variation is found here. Hence, for the 3-spin model, all measured corrections scale the same as for the random energy model (REM), corresponding to p 1/4 00. This would suggest that all p-spin models with p >= 3 exhibit the same ground-state corrections as in the REM.
引用
收藏
页数:5
相关论文
共 41 条
[1]   Reexamining classical and quantum models for the D-Wave One processor [J].
Albash, T. ;
Ronnow, T. F. ;
Troyer, M. ;
Lidar, D. A. .
EUROPEAN PHYSICAL JOURNAL-SPECIAL TOPICS, 2015, 224 (01) :111-129
[2]   Physics-Inspired Optimization for Quadratic Unconstrained Problems Using a Digital Annealer [J].
Aramon, Maliheh ;
Rosenberg, Gili ;
Valiante, Elisabetta ;
Miyazawa, Toshiyuki ;
Tamura, Hirotaka ;
Katzgraber, Helmut G. .
FRONTIERS IN PHYSICS, 2019, 7 (APR)
[3]   Finite-size corrections in the Sherrington-Kirkpatrick model [J].
Aspelmeier, T. ;
Billoire, A. ;
Marinari, E. ;
Moore, M. A. .
JOURNAL OF PHYSICS A-MATHEMATICAL AND THEORETICAL, 2008, 41 (32)
[4]   Exact enumeration of ground states in the Sherrington-Kirkpatrick spin glass [J].
Boettcher, S ;
Kott, TM .
PHYSICAL REVIEW B, 2005, 72 (21)
[5]   Nature's way of optimizing [J].
Boettcher, S ;
Percus, A .
ARTIFICIAL INTELLIGENCE, 2000, 119 (1-2) :275-286
[6]   Extremal optimization for Sherrington-Kirkpatrick spin glasses [J].
Boettcher, S .
EUROPEAN PHYSICAL JOURNAL B, 2005, 46 (04) :501-505
[7]   Optimization with extremal dynamics [J].
Boettcher, S ;
Percus, AG .
PHYSICAL REVIEW LETTERS, 2001, 86 (23) :5211-5214
[8]   Deep reinforced learning heuristic tested on spin-glass ground states: The larger picture [J].
Boettcher, Stefan .
NATURE COMMUNICATIONS, 2023, 14 (01)
[9]   Inability of a graph neural network heuristic to outperform greedy algorithms in solving combinatorial optimization problems [J].
Boettcher, Stefan .
NATURE MACHINE INTELLIGENCE, 2023, 5 (01) :24-25
[10]   Analysis of the relation between quadratic unconstrained binary optimization and the spin-glass ground-state problem [J].
Boettcher, Stefan .
PHYSICAL REVIEW RESEARCH, 2019, 1 (03)