Discovery of Optimal Quantum Error Correcting Codes via Reinforcement Learning

by Mr Vincent Paul Su (University of California, Berkeley)

Thursday 28 Dec 2023, 11:00 → 12:00 Asia/Taipei

Physics/620 (NTHU)

Quantum error correction design is a difficult but important problem on the road to fault tolerance. In this talk, I will review one way to build larger quantum error correcting codes from smaller ones. By treating the individual encoding unitaries as tensors, small QECCs can be stitched together producing rich emergent behavior. For example, the surface code can be built from many identical copies of a simple 4 qubit code. By treating code building as a game, we are able to use machine learning methods to build novel codes that i) saturate bounds on distance for CSS codes and ii) outperform surface code variants for ~20 qubits at protecting logical information from biased Pauli noise. Based on https://arxiv.org/abs/2305.06378