Abstract and subjects
This work develops a novel approach to exploit synthesized, approximate circuits for the ansatz of variational quantum algorithms (VQA) and demonstrates its effectiveness for NchooseK, a domain-specific language supporting quantum-based solving of constraint-based problems. Synthesis is generalized to produce parametric circuits of short depth in close approximation of the original circuit offline. This removes syn-thesis from the critical path (online) between repeated quantum circuit executions of VQA while reducing circuit depth, thereby resulting in higher fidelity results than the baseline without synthesis. Simulation experiments indicate improvements of 98% on average. Further, experiments indicate that this approach can obtain viable solutions when the baseline could not. All of this is achieved with an average variation in circuit depth of less than 10%.