Mear Koochakie

We introduce a machine-learning-aided tensor network (TN) framework for discovering quantum error-correcting codes (QECCs). By directly representing the code space projector as a 2D TN, our approach captures standard stabilizer formalisms while extending naturally to non-additive codes. We frame QECC discovery as a hybrid optimization problem, using continuous gradient-based methods alongside discrete reinforcement learning to simultaneously optimize tensor weights and network topology. This provides a scalable, automated methodology for exploring novel QECCs.