Combinatorial optimization problems are notoriously hard for conventional computers to solve efficiently. While quantum and analog hardware have been explored to tackle these problems, they often face challenges such as high power use, complexity, or limited scalability. This study introduces a novel approach using commercial vertical NAND (V‐NAND) flash memory, commonly found in everyday devices, as the basis for solving these problems. By creatively adjusting how the memory cells operate, we implement a Hopfield Neural Network that can mimic simulated annealing, a method for finding near‐optimal solutions. Our system achieves high accuracy in solving the max‐cut problem while consuming significantly less energy than conventional graphics processing unit‐ or field‐programmable gate array‐based solutions. Unlike emerging technologies, our design uses existing V‐NAND flash memory without any structural changes, making it highly practical for large‐scale and energy‐efficient applications. This work demonstrates that V‐NAND flash memory is not just for storage but can also serve as a powerful tool for solving complex optimization problems.
Park et al. (Sun,) studied this question.