Post-Selection-Free Quantum Automated Learning (in Preparation)

Author: Wang, Junkai; Deng, Dong-Ling; Liu, Jin-Peng.
[First author.]

Published in Manuscript in preparation, 2026

Summary: We identify post-selection as the central serial bottleneck of quantum automated learning (QAL) and show that projector-mixing QAL updates admit an explicit block-encoding with a certified success floor. This enables a branch-preserving fixed-point amplitude amplification wrapper that turns long training trajectories into a near-deterministic process with predictable query cost, and the same certified-floor logic extends to normalized imaginary-time protocols and a microcanonical-double lifting regime.