TY - GEN
T1 - A Novel Classical-Ising Hybrid Annealing Method with QUBO Model Cutting
AU - Atobe, Yuta
AU - Tawada, Masashi
AU - Togawa, Nozomu
N1 - Publisher Copyright:
© 2024 IEEE.
PY - 2024
Y1 - 2024
N2 - Most quantum annealers, or Ising machines have the hardware limitations in its qubit size and thus the computable problem size is much limited. This paper proposes a novel classical-Ising hybrid annealing method which virtually extends the computable size of Ising machines based on a theoretical analysis. Given a quadratic unconstrained binary optimization (QUBO) model, the proposed method selects the cutting binary variable that is most likely to have the same value as the ground-state solution. Then, we recursively cut the QUBO model based on the cutting binary variable until we obtain sufficiently small-sized subQUBO models. Experimental evaluations demonstrate that the proposed hybrid annealing method can give much better quasi-ground-state solutions than state-of-the-art existing methods for large-sized QUBO models.
AB - Most quantum annealers, or Ising machines have the hardware limitations in its qubit size and thus the computable problem size is much limited. This paper proposes a novel classical-Ising hybrid annealing method which virtually extends the computable size of Ising machines based on a theoretical analysis. Given a quadratic unconstrained binary optimization (QUBO) model, the proposed method selects the cutting binary variable that is most likely to have the same value as the ground-state solution. Then, we recursively cut the QUBO model based on the cutting binary variable until we obtain sufficiently small-sized subQUBO models. Experimental evaluations demonstrate that the proposed hybrid annealing method can give much better quasi-ground-state solutions than state-of-the-art existing methods for large-sized QUBO models.
KW - Hybrid annealing method
KW - Ising machine
KW - Ising model
KW - QUBO model
KW - subQUBO model
UR - http://www.scopus.com/inward/record.url?scp=85204957091&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85204957091&partnerID=8YFLogxK
U2 - 10.1109/MWSCAS60917.2024.10658806
DO - 10.1109/MWSCAS60917.2024.10658806
M3 - Conference contribution
AN - SCOPUS:85204957091
T3 - Midwest Symposium on Circuits and Systems
SP - 1154
EP - 1157
BT - 2024 IEEE 67th International Midwest Symposium on Circuits and Systems, MWSCAS 2024
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 67th IEEE International Midwest Symposium on Circuits and Systems, MWSCAS 2024
Y2 - 11 August 2024 through 14 August 2024
ER -