Curriculum reinforcement learning for optimization of variational quantum circuit architectures

Mateusz Ostaszewski, Wojciech Masarczyk, Lea Marion Trenkwalder, Eleanor Scerri, Vedran Dunjko

Abstract:

As we are entering the so called Noisy Intermediate Scale Quantum (NISQ)[9] technology era, the search for more suitable algorithms under NISQ restrictions is becoming ever important. A truly compatible NISQ application must first be amenable to architecture constraints and size limits. Furthermore, to minimize the adverse effects of gate errors and decoherence, it is important that the circuits we run are as gate-frugal, and as shallow as possible.
Perhaps the most promising classes of such algorithms are based on variational circuit methods, applied to problems in quantum chemistry. A key problem in this field is the computing of ground state energies and low energy properties of chemical systems (the chemical structure problem). This problem is believed to be intractable in general, yet the quantum Variational Quantum Eigensolver (VQE)[8] algorithm can provide solutions in regimes which beyond the reach of classical algorithms, while maintaining NISQ-friendly properties.