Training a quantum optimizer Article Swipe
YOU?
·
· 2016
· Open Access
·
· DOI: https://doi.org/10.1103/physreva.94.022309
· OA: W4239969927
We study a variant of the quantum approximate optimization algorithm [ E.\nFarhi, J. Goldstone, and S. Gutmann, arXiv:1411.4028] with slightly different\nparametrization and different objective: rather than looking for a state which\napproximately solves an optimization problem, our goal is to find a quantum\nalgorithm that, given an instance of MAX-2-SAT, will produce a state with high\noverlap with the optimal state. Using a machine learning approach, we chose a\n"training set" of instances and optimized the parameters to produce large\noverlap for the training set. We then tested these optimized parameters on a\nlarger instance set. As a training set, we used a subset of the hard instances\nstudied by E. Crosson, E. Farhi, C. Yen-Yu Lin, H.-H. Lin, and P. Shor (CFLLS)\n[arXiv:1401.7320]. When tested on the full set, the parameters that we find\nproduce significantly larger overlap than the optimized annealing times of\nCFLLS. Testing on other random instances from $20$ to $28$ bits continues to\nshow improvement over annealing, with the improvement being most notable on the\nhardest instances. Further tests on instances of MAX-3-SAT also showed\nimprovement on the hardest instances. This algorithm may be a possible\napplication for near-term quantum computers with limited coherence times.\n
