Unitary Synthesis of Clifford+T Circuits with Reinforcement Learning

• DOI: 10.48550/arxiv.2404.14865
• Publication Date: 2024-04-23
• Authors: Sebastian Rietsch; Abhishek Y. Dubey; Christian Ufrecht; Maniraman Periyasamy; Axel Plinge; Christopher Mutsc...; Daniel D. Scherer
• Keywords: Quantum Physics; FOS: Physical sciences; Quantum Physics (quant-ph)

This paper presents a deep reinforcement learning approach for synthesizing unitaries into quantum circuits. Unitary synthesis aims to identify circuit that represents given unitary while minimizing depth, total gate count, specific or combination of these factors. While past research has focused predominantly on continuous sets, from the parameter-free Clifford+T set remains challenge. Although time complexity this task will inevitably remain exponential in number qubits general unitaries, reducing runtime simple problem instances still poses significant In study, we apply tree-search method Gumbel AlphaZero solve subset exactly synthesizable unitaries. Our can synthesize up five generated randomized circuits with 60 gates. Furthermore, our inference times are around 30 seconds single GPU average, surpassing state-of-the-art algorithms QuantumCircuitOpt and MIN-T-SYNTH higher qubit numbers. work provides competitive baseline be developed upcoming years.