Annealing Processor Search Space Reduction via Preprocessing

The patent applies preliminary action by using classical computation (CPU/GPU) to perform preprocessing of the parameter space before the annealing processor executes. This includes generating initial solutions, identifying promising regions, and filtering out unpromising parameters. By preparing and reducing the parameter space in advance, the annealing processor receives a pre-filtered subset that requires less computation time while still finding optimal or near-optimal solutions, thus resolving the contradiction between execution speed and search space coverage.

The patent applies the extraction principle by removing unpromising parameters and regions from the full parameter space before processing. Classical computation algorithms analyze the problem structure and extract only the most promising subsets of parameters that are likely to contain optimal solutions. This extracted reduced parameter space is then input to the annealing processor, which focuses its compute capacity on these high-potential regions rather than wasting resources on exhaustive search of the entire space, thereby maximizing both accuracy and compute utilization.

The patent applies the merging principle by combining classical computation algorithms with annealing processor computation in a hybrid approach. Classical algorithms provide simple preprocessing, problem transformation, and postprocessing, while the annealing processor handles the complex optimization of the reduced parameter space. This combination leverages the simplicity and flexibility of classical algorithms together with the speed and parallelism of annealing hardware, achieving both ease of implementation and superior computation performance.