Abstract:The No-Free-Lunch (NFL) theorem, which quantifies problem- and data-independent generalization errors regardless of the optimization process, provides a foundational framework for comprehending diverse learning protocols' potential. Despite its significance, the establishment of the NFL theorem for quantum machine learning models remains largely unexplored, thereby overlooking broader insights into the fundamental relationship between quantum and classical learning protocols. To address this gap, we categorize a diverse array of quantum learning algorithms into three learning protocols designed for learning quantum dynamics under a specified observable and establish their NFL theorem. The exploited protocols, namely Classical Learning Protocols (CLC-LPs), Restricted Quantum Learning Protocols (ReQu-LPs), and Quantum Learning Protocols (Qu-LPs), offer varying levels of access to quantum resources. Our derived NFL theorems demonstrate quadratic reductions in sample complexity across CLC-LPs, ReQu-LPs, and Qu-LPs, contingent upon the orthogonality of quantum states and the diagonality of observables. We attribute this performance discrepancy to the unique capacity of quantum-related learning protocols to indirectly utilize information concerning the global phases of non-orthogonal quantum states, a distinctive physical feature inherent in quantum mechanics. Our findings not only deepen our understanding of quantum learning protocols' capabilities but also provide practical insights for the development of advanced quantum learning algorithms.

What problem does this paper attempt to address?

This paper discusses the differences between classical and quantum learning protocols, particularly within the framework of the No-Free-Lunch (NFL) theorem. The NFL theorem quantifies the generalization error of different learning protocols without considering the optimization process. Although there is not sufficient research on the NFL theorem for quantum machine learning models, the paper categorizes various quantum learning algorithms into three types of learning protocols and establishes their NFL theorems. These three protocols are classical learning protocols (CLC-LPs), restricted quantum learning protocols (ReQu-LPs), and quantum learning protocols (Qu-LPs), which have different access rights to quantum resources. The paper points out that these protocols significantly reduce the sample complexity for non-orthogonal quantum states and diagonal observables. The advantage of quantum learning protocols lies in their ability to indirectly utilize global phase information of non-orthogonal quantum states, which is an inherent characteristic of quantum mechanics. Through these theorems, the research not only deepens the understanding of the capabilities of quantum learning protocols but also provides practical guidance for the development of more advanced quantum learning algorithms. The paper also emphasizes the importance of the general separation between quantum and classical learning protocols, especially in handling large-scale data, where quantum computers may offer more efficient solutions. Although quantum machine learning has shown advantages in specific tasks, the universality gap in their performance across all datasets and problems remains unclear. Therefore, establishing NFL theorems for these protocols helps to fairly compare the performance of different learning protocols and reveal the advantages of quantum learning. In conclusion, this paper strives to fill the gaps in the theoretical framework of quantum machine learning, providing a new perspective for understanding and improving quantum learning algorithms by analyzing the generalization abilities of different protocols.

Separable Power of Classical and Quantum Learning Protocols Through the Lens of No-Free-Lunch Theorem

Transition Role of Entangled Data in Quantum Machine Learning

Non-Unitary Quantum Machine Learning

Quantum-classical separations in shallow-circuit-based learning with and without noises

Unifying (Quantum) Statistical and Parametrized (Quantum) Algorithms

Entanglement-induced provable and robust quantum learning advantages

Arbitrary Polynomial Separations in Trainable Quantum Machine Learning

Towards understanding the power of quantum kernels in the NISQ era

Quantum delegated and federated learning via quantum homomorphic encryption

Experimental Machine Learning of Quantum States

The Dilemma of Quantum Neural Networks

Information-theoretic generalization bounds for learning from quantum data

Quantum Private Distributed Learning Through Blind Quantum Computing

Classical Verification of Quantum Learning

Problem-Dependent Power of Quantum Neural Networks on Multi-Class Classification

Problem-Dependent Power of Quantum Neural Networks on Multiclass Classification

Practical distributed quantum information processing with LOCCNet

On establishing learning separations between classical and quantum machine learning with classical data

Adversarial Learning in Quantum Artificial Intelligence

Quantum Local Differential Privacy and Quantum Statistical Query Model