Tokyo Institute of Science Builds Quantum Trainable Inequality Constraint Solver

Researchers from the Institute of Science Tokyo and RIKEN Center for Advanced Intelligence Project have developed a new approach to solving complex combinatorial optimization problems without increasing the demands on quantum computing hardware. Their work centers on the deep-unfolded unbalanced penalization Ohzeki method, or DU-UPOM, which learns the most effective way to find solutions from example problems, a departure from earlier methods requiring manual adjustments. Numerical experiments using random knapsack problems show that DU-UPOM improves over original UP and reaches optimal solutions in fewer iterations than fixed-step UPOM and other baseline methods. This framework, the authors report, reduces the tuning and embedding burdens while making the Ohzeki method trainable for problems with inequality constraints, potentially expanding the scope of solvable problems for quantum annealers. Achieving optimal solutions…