论文标题
量子退火中的波动引导搜索
Fluctuation guided search in quantum annealing
论文作者
论文摘要
量子退火在利用量子力学来解决组合优化问题方面有很大的希望。但是，要在最大程度上实现这一诺言，我们必须适当利用基本的物理学。本着这种精神，我研究了量子退火器的众所周知的趋势如何寻求允许更多量子波动的解决方案，以便将解决方案的最优性交换为合成问题的最佳性，以便能够具有更灵活的解决方案的能力，而某些变量几乎可以以很少的成本更改。我使用反向退火功能D-Wave Systems QPU进行实验证明了这一折衷，这两个问题由所有二进制变量组成，以及包含一些比二进制离散变量的较高的变量。我进一步证明了如何使用Qubits上的本地控制来控制波动的水平并指导搜索。我讨论了利用这种权衡的地方实际上很重要，即在混合算法中，无法直接在退火器上实施某些罚款，并提供了一些概念概念证据的证据，证明了这些算法如何起作用。
Quantum annealing has great promise in leveraging quantum mechanics to solve combinatorial optimisation problems. However, to realize this promise to it's fullest extent we must appropriately leverage the underlying physics. In this spirit, I examine how the well known tendency of quantum annealers to seek solutions where more quantum fluctuations are allowed can be used to trade off optimality of the solution to a synthetic problem for the ability to have a more flexible solution, where some variables can be changed at little or no cost. I demonstrate this tradeoff experimentally using the reverse annealing feature a D-Wave Systems QPU for both problems composed of all binary variables, and those containing some higher-than-binary discrete variables. I further demonstrate how local controls on the qubits can be used to control the levels of fluctuations and guide the search. I discuss places where leveraging this tradeoff could be practically important, namely in hybrid algorithms where some penalties cannot be directly implemented on the annealer and provide some proof-of-concept evidence of how these algorithms could work.
