The equivalence between the natural minimization of energy in a dynamical system and the minimization of an objective function characterizing a combinatorial optimization problem offers a promising approach to designing dynamical system-inspired computational models and solvers for such problems. For instance, the ground state energy of coupled electronic oscillators, under second harmonic injection, can be directly mapped to the optimal solution of the Maximum Cut problem. However, prior work has focused on a limited set of such problems. Therefore, in this work, we formulate computing models based on synchronized oscillator dynamics for a broad spectrum of combinatorial optimization problems ranging from the Max-K-Cut (the general version of the Maximum Cut problem) to the Traveling Salesman Problem. We show that synchronized oscillator dynamics can be engineered to solve these different combinatorial optimization problems by appropriately designing the coupling function and the external injection to the oscillators. Our work marks a step forward towards expanding the functionalities of oscillator-based analog accelerators and furthers the scope of dynamical system solvers for combinatorial optimization problems.
翻译:在动态系统中自然最大限度地减少能源与最大限度地减少组合优化问题的客观功能之间的等同性,为设计动态系统驱动计算模型和这类问题的解决方案提供了一个很有希望的方法。例如,在第二次口腔注射下,同时电动振荡器的地面状态能量可以直接映射到最大剪切问题的最佳解决办法。然而,先前的工作侧重于有限的一系列此类问题。因此,在这项工作中,我们根据同步振荡器动态模型,为从最大剪动问题的一般版本Max-K-Cut到旅行推销员问题等一系列广泛的组合优化问题制定计算模型。我们表明,可以设计同步振荡器动态,通过适当设计组合功能和对振荡器的外部注入,解决这些不同的组合优化问题。我们的工作标志着在扩大以振荡器为基础的模拟加速器的功能方面向前迈出了一步,进一步扩展了组合优化问题动态系统解析器的范围。