WorksPhysics and Astronomy
Environment-assisted imaginary time evolution in feedback-based quantum algorithms
Abstract
The Feedback-based Algorithm for Quantum Optimization (FALQON) prepares ground states without classical parameter optimization, but requires deep circuits and can stagnate away from the ground state as a consequence of barren plateaus and symmetry protection. In order to avoid stagnation, imaginary time evolution (ITE) layers can be incorporated into FALQON layers, enhancing FALQON convergence. However, applying a non-unitary ITE layer in a closed quantum system is not possible, and approximations must be made to use ITE on real hardware; using an open quantum system formulation to incorporate ITE into FALQON represents a better approach. We describe ITE as a quantum operation induced by an environment, and combine it with the FALQON feedback law. In simulations of a frustrated Ising chain (ANNNI) and of MaxCut, the resulting algorithm prepares the ground state starting from any initial state, where FALQON can fail, and needs fewer layers on average when the initial state is fixed for the two investigated problems. The CITE-FALQON advantage strongly depends on the system phase for the ANNNI chain, and on the graph weights for the MaxCut problem.