Kim Joon Pyo, Kim Hyun Wook, Jeong Jaeyong, Park Juhyuk, Kuk Song-Hyeon, Kim Jongmin, Woo Jiyong, Kim Sanghyeon
School of Electrical Engineering, Korea Advanced Institute of Science and Technology (KAIST), 291 Daehak-ro, Yuseong-gu, Daejeon, 34141, Republic of Korea.
School of Electronic and Electrical Engineering, Kyungpook National University (KNU), 80 Daehak-ro, Buk-gu, Daegu, 41566, Republic of Korea.
Small. 2024 Dec;20(52):e2406822. doi: 10.1002/smll.202406822. Epub 2024 Oct 21.
Ising solvers are important for efficiently addressing non-deterministic polynomial-time (NP)-hard combinatorial optimization problems (COPs), where scalability and compactness are crucial for practical applications. In this study, an experimental demonstration of an oscillator-based Ising solver employing a highly scalable 4F InGaAs bi-stable resistor (biristor) is presented. It is first explored the oscillation behavior of the InGaAs biristor, establishing that classical Ising spins can be emulated using the sub-harmonic injection locking (SHIL) technique. Furthermore, capacitive and resistive coupling between two coupled InGaAs biristors is demonstrated, leading to out-of-phase and in-phase coupling, respectively. Employing this foundational technology, it is experimentally achieved a solution to the MaxCUT problem with the InGaAs biristor-based Ising solver, supplemented by simulation-based behavior evaluations. This emerging device architecture offers a viable pathway to surmount the scaling limitations faced by present hardware-based Ising solvers, representing a significant step forward in the development of efficient, scalable solutions for complex optimization challenges.
伊辛求解器对于有效解决非确定性多项式时间(NP)难组合优化问题(COP)非常重要,其中可扩展性和紧凑性对于实际应用至关重要。在本研究中,展示了一种基于振荡器的伊辛求解器的实验演示,该求解器采用了高度可扩展的4F InGaAs双稳态电阻器(双稳电阻)。首先研究了InGaAs双稳电阻的振荡行为,确定可以使用次谐波注入锁定(SHIL)技术来模拟经典伊辛自旋。此外,还展示了两个耦合的InGaAs双稳电阻之间的电容性和电阻性耦合,分别导致异相耦合和同相耦合。利用这一基础技术,通过基于InGaAs双稳电阻的伊辛求解器,通过基于模拟的行为评估,实验实现了对最大割问题的解决方案。这种新兴的器件架构为克服当前基于硬件的伊辛求解器面临的扩展限制提供了一条可行途径,代表了在开发针对复杂优化挑战的高效、可扩展解决方案方面向前迈出的重要一步。