Greedy Randomized Search for Scalable Compilation of Quantum Circuits

An Innovative Genetic Algorithm for the Quantum Circuit Compilation Problem

“…The results are presented in aggregated form for reasons of space; however, the complete set of makespan values, together with the complete set of solutions are available at http://pst.istc.cnr.it/ ∼ angelo/qc/. We compare our results with the best results presented in (Venturelli et al 2017) and obtained with the T F D (Eyerich, Mattmüller, and Röger 2009), SGP lan (Wah and Chen 2004;Chen, Wah, and Hsu 2006) and LP G (Gerevini, Saetti, and Serina 2003) temporal planners, and wih those presented in (Oddi and Rasconi 2018) and obtained by means of their GRS greedy random sampling procedure (see the respective columns in Table 1). The results of our implementation of the GRS procedure and of the genetic algorithm are presented in the GRS * and GA columns, respectively.…”

“…The Quantum Gate Ranking Heuristic (QGRH) The solution building approach presented in (Oddi and Rasconi 2018) is based on the chronological insertion of one gate operation at a time in the partial solution S, until all the gates in the set P-S ∪ MIX are in S. This insertion relies on the concept of chain; we define chain i as a sequence of gate operations op ∈ S that involve qstate q i . In a (partial) solution S there exist as many chains as are the qstates involved in the problem instance.…”

“…The material presented in this section describes the performance obtained with our genetic algorithm against the benchmark set originally presented in (Venturelli et al 2017) (QCCP problem version), and later expanded in (Booth et al 2018) (QCCP-V and QCCP-X problem versions). Many of the results obtained in (Venturelli et al 2017) 1 have been successively improved by (Oddi and Rasconi 2018) by means of an heuristic-based greedy randomized search procedure.…”

“…In particular, the benchmark set we tackle is composed of three benchmarks characterized by intances of three different sizes, based on quantum chips with N = 8, 21 and 40 qubits, respectively (see Figure 1). The N = 8 and the N = 21 benchmarks are solved considering P = 2 problem instances (two passes), while the N = 40 benchmark is solved in both flavors P = 1 (to allow comparison with (Venturelli et al 2017)) and P = 2 (to allow comparison with (Oddi and Rasconi 2018)). Moreover, in order to guarantee maximum fairness of comparison, we have also implemented a version of the greedy random sampling proposed in (Oddi and Rasconi 2018), named GRS * in this work, to differentiate it with the original GRS procedure.…”

“…To guarantee fairness of comparison also from the computing resources standpoint, all experiments have been performed on a 64-bit Windows10 O.S. running on Intel(R) Core(TM)2 Duo CPU E8600 @3.33 GHz with 8GB RAM, exactly as in (Oddi and Rasconi 2018).…”