IQIM Postdoctoral and Graduate Student Seminar

Abstract: Learning and sensing are among the most promising applications of quantum devices, yet noise threatens to erase the quantum advantage before it can be realized at scale. To address this challenge, we model the quantum computer's access to an unknown external system as an oracle, and propose oracle distillation (OD) — constructing a single effectively error-free oracle from multiple noisy ones.

In this talk, I will explain OD from two complementary perspectives. First, we introduce OD as a practical tool that utilizes quantum error correction to robustify quantum learning and sensing algorithms. The key idea is the weak query: a quantum computer deliberately interrogates its environment only weakly, balancing the information gained per query against the noise incurred. This leads to near-optimal protocols for a broad class of Boolean oracles and, in particular, preserves Grover speedup under local depolarizing noise with negligible overhead. Second, we establish OD as a conceptual bridge between quantum learning and error correction. It allows query lower bounds for noisy learning tasks to be translated into structural constraints on quantum codes, yielding information-theoretic lower bounds on the block length of exact codes with certain transversal higher-level Clifford hierarchy gates. Our work paves the way toward a general framework for a robust interface between quantum computers and noisy environments, opening a path to scalable quantum advantages in learning and sensing.

Following the talk, lunch will be provided on the lawn outside East Bridge.