Title:Multi-objective evolutionary algorithms for quantum circuit discovery

Abstract:Quantum hardware continues to advance, yet finding new quantum algorithms - quantum software - remains a challenge, with classically trained computer programmers having little intuition of how computational tasks may be performed in the quantum realm. As such, the idea of developing automated tools for algorithm development is even more appealing for quantum computing than for classical. Here we develop a robust, multi-objective evolutionary search strategy to design quantum circuits 'from scratch', by combining and parameterizing a task-generic library of quantum circuit elements. When applied to 'ab initio' design of quantum circuits for the input/output mapping requirements of the quantum Fourier transform and Grover's search algorithm, it finds textbook circuit designs, along with alternative structures that achieve the same functionality. Exploiting its multi-objective nature, the discovery algorithm can trade off performance measures such as accuracy, circuit width or depth, gate count, or implementability - a crucial requirement for first-generation quantum processors and applications.