Unifying (Quantum) Statistical and Parametrized (Quantum) Algorithms

TL;DR

This paper introduces a unified framework bridging classical and quantum learning paradigms, providing new insights into the complexity and learnability of quantum states and machine learning tasks using evaluation oracles.

Contribution

It develops a novel framework connecting statistical and parametrized quantum algorithms, establishing lower bounds and characterizing query complexity in quantum learning.

Findings

Extended learnability results from SQ to QSQ for quantum circuit outputs.

Identified exponential separations in learning stabilizer states with QSQs versus quantum samples.

Provided a new perspective on the hardness of quantum machine learning tasks.

Abstract

Kearns' statistical query (SQ) oracle (STOC'93) lends a unifying perspective for most classical machine learning algorithms. This ceases to be true in quantum learning, where many settings do not admit, neither an SQ analog nor a quantum statistical query (QSQ) analog. In this work, we take inspiration from Kearns' SQ oracle and Valiant's weak evaluation oracle (TOCT'14) and establish a unified perspective bridging the statistical and parametrized learning paradigms in a novel way. We explore the problem of learning from an evaluation oracle, which provides an estimate of function values, and introduce an extensive yet intuitive framework that yields unconditional lower bounds for learning from evaluation queries and characterizes the query complexity for learning linear function classes. The framework is directly applicable to the QSQ setting and virtually all algorithms based on loss function optimization. Our first application is to extend prior results on the learnability of output distributions of quantum circuits and Clifford unitaries from the SQ to the (multi-copy) QSQ setting, implying exponential separations between learning stabilizer states from (multi-copy) QSQs versus from quantum samples. Our second application is to analyze some popular quantum machine learning (QML) settings. We gain an intuitive picture of the hardness of many QML tasks which goes beyond existing methods such as barren plateaus and the statistical dimension, and contains crucial setting-dependent implications. Our framework not only unifies the perspective of cost concentration with that of the statistical dimension in a unified language but exposes their connectedness and similarity.