Quantum neural networks form Gaussian processes

Abstract

It is well known that artificial neural networks initialized from independent and identically distributed priors converge to Gaussian processes in the limit of a large number of neurons per hidden layer. In this work we prove an analogous result for Quantum Neural Networks (QNNs). Namely, we show that the outputs of certain models based on Haar random unitary or orthogonal deep QNNs converge to Gaussian processes in the limit of large Hilbert space dimension $d$. The derivation of this result is more nuanced than in the classical case due to the role played by the input states, the measurement observable, and the fact that the entries of unitary matrices are not independent. Then, we show that the efficiency of predicting measurements at the output of a QNN using Gaussian process regression depends on the observable's bodyness. Furthermore, our theorems imply that the concentration of measure phenomenon in Haar random QNNs is worse than previously thought, as we prove that expectation values and gradients concentrate as $\mathcal{O}\left(\frac{1}{e^d\sqrt{d}}\right)$. Finally, we discuss how our results improve our understanding of concentration in $t$-designs.

arXiv comment14+37 pages, 4+6 figures

References in corpus (27)

• Quantum Machine Learning
• Variational Quantum Algorithms
• Supervised learning with quantum enhanced feature spaces
• Barren plateaus in quantum neural network training landscapes
• Natural Language Processing: State of The Art, Current Trends and Challenges
• Predicting Many Properties of a Quantum System from Very Few Measurements
• Cost Function Dependent Barren Plateaus in Shallow Parametrized Quantum Circuits
• The foundations of statistical mechanics from entanglement: Individual states vs. averages
• Challenges and Opportunities in Quantum Machine Learning
• Data re-uploading for a universal quantum classifier
• Random Quantum Circuits are Approximate 2-designs
• Is quantum advantage the right goal for quantum machine learning?
• Matchgates and classical simulation of quantum circuits
• Theory of overparametrization in quantum neural networks
• Group-Invariant Quantum Machine Learning
• Equivalence of quantum barren plateaus to cost concentration and narrow gorges
• Bayesian Deep Convolutional Networks with Many Channels are Gaussian Processes
• Entanglement Devised Barren Plateau Mitigation
• Barren plateaus preclude learning scramblers
• Approximate unitary $t$-designs by short random quantum circuits using nearest-neighbor and long-range gates
• Decoupling with random quantum circuits
• Random quantum circuits are approximate unitary $t$-designs in depth $O\left(n{t}^{5+o(1)}\right)$
• Representation Learning via Quantum Neural Tangent Kernels
• Bayesian Deep Learning on a Quantum Computer
• Interpretable Quantum Advantage in Neural Sequence Learning
• Unscrambling Quantum Information with Clifford decoders
• Preparation of ordered states in ultra-cold gases using Bayesian optimization

Cited by in corpus (12)

• Does provable absence of barren plateaus imply classical simulability?
• Quantum Convolutional Neural Networks are Effectively Classically Simulable
• Quantum continual learning on a programmable superconducting processor
• Adversarial Robustness Guarantees for Quantum Classifiers
• Exact spectral gaps of random one-dimensional quantum circuits
• Real randomized measurements for analyzing properties of quantum states
• Architectures and random properties of symplectic quantum circuits
• Analyzing the free states of one quantum resource theory as resource states of another
• Detecting high-dimensional entanglement by randomized product projections
• Adaptive Interpolating Quantum Transform: A Quantum-Native Framework for Efficient Transform Learning
• Moments of Quantum Channel Ensembles
• A graph-theoretic approach to chaos and complexity in quantum systems