Twisty-puzzle-inspired Approach to Clifford Synthesis

The problem of decomposing an arbitrary Clifford element into a sequence of Clifford gates is known as Clifford synthesis. Drawing inspiration from similarities between this and the famous Rubik's cube twisty puzzle, we develop a machine learning approach for Clifford synthesis based on learning an approximation to the distance to the identity. This approach is probabilistic and computationally intensive. However, when a decomposition is successfully found, it often involves fewer gates than the decomposition methods used in the Qiskit decomposition protocol, which uses a combination of several well-known Clifford decomposition schemes. Additionally, our approach is much more flexible than existing algorithms in that arbitrary gate sets, device topologies, and gate fidelities may be incorporated, thus allowing for the approach to be tailored to a specific device.