Machine learning search for quantum algorithms

Quantum algorithm design lies in the hallmark of applications of quantum computation and quantum simulation. Recent theoretical progress has established complexity-equivalence of circuit and adiabatic quantum algorithms. Here we utilize deep reinforcement learning methods to search for optimal Hamiltonian path in the framework of quantum adiabatic algorithm. We benchmark our approach in Grover search and 3-SAT problems, and find that the adiabatic algorithm obtained by our reinforcement learning approach leads to improved performance in the final state fidelity and significant computational speedups for both moderate and large number of qubits compared to conventional algorithms. Our approach offers a recipe to design quantum algorithms for generic problems through a systematic search. This approach paves a novel way to automated quantum algorithm design by artificial intelligence.