Effects of noise on the overparametrization of quantum neural networks

Abstract

Overparametrization is one of the most surprising and notorious phenomena in machine learning. Recently, there have been several efforts to study if, and how, Quantum Neural Networks (QNNs) acting in the absence of hardware noise can be overparametrized. In particular, it has been proposed that a QNN can be defined as overparametrized if it has enough parameters to explore all available directions in state space. That is, if the rank of the Quantum Fisher Information Matrix (QFIM) for the QNN's output state is saturated. Here, we explore how the presence of noise affects the overparametrization phenomenon. Our results show that noise can "turn on" previously-zero eigenvalues of the QFIM. This enables the parametrized state to explore directions that were otherwise inaccessible, thus potentially turning an overparametrized QNN into an underparametrized one. For small noise levels, the QNN is quasi-overparametrized, as large eigenvalues coexists with small ones. Then, we prove that as the magnitude of noise increases all the eigenvalues of the QFIM become exponentially suppressed, indicating that the state becomes insensitive to any change in the parameters. As such, there is a pull-and-tug effect where noise can enable new directions, but also suppress the sensitivity to parameter updates. Finally, our results imply that current QNN capacity measures are ill-defined when hardware noise is present.

14 + 6 pages, 11 figures

References in corpus (19)

• Noisy intermediate-scale quantum (NISQ) algorithms
• Challenges and Opportunities in Quantum Machine Learning
• Connecting ansatz expressibility to gradient magnitudes and barren plateaus
• Quantum advantage in learning from experiments
• Training variational quantum algorithms is NP-hard
• Provably efficient machine learning for quantum many-body problems
• Diagnosing Barren Plateaus with Tools from Quantum Optimal Control
• Theory of overparametrization in quantum neural networks
• Fisher Information in Noisy Intermediate-Scale Quantum Applications
• Quantum Variational Optimization of Ramsey Interferometry and Atomic Clocks
• Theoretical Guarantees for Permutation-Equivariant Quantum Neural Networks
• Representation Learning via Quantum Neural Tangent Kernels
• Analytic theory for the dynamics of wide quantum neural networks
• Progress toward favorable landscapes in quantum combinatorial optimization
• Efficient classical algorithms for simulating symmetric quantum systems
• Exploiting Landscape Geometry to Enhance Quantum Optimal Control
• Quantum simulation with just-in-time compilation
• Inference-Based Quantum Sensing
• Characterization of variational quantum algorithms using free fermions

Cited by in corpus (15)

• A Lie Algebraic Theory of Barren Plateaus for Deep Parameterized Quantum Circuits
• Towards large-scale quantum optimization solvers with few qubits
• An Expressive Ansatz for Low-Depth Quantum Approximate Optimisation
• Generalization of Quantum Machine Learning Models Using Quantum Fisher Information Metric
• Lie-algebraic classical simulations for quantum computing
• On the universality of $S_n$-equivariant $k$-body gates
• Investigating the Effect of Noise on the Training Performance of Hybrid Quantum Neural Networks
• Quantum DeepONet: Neural operators accelerated by quantum computing
• Forecasting steam mass flow in power plants using the parallel hybrid network
• Random Natural Gradient
• Simulating quantum circuits with arbitrary local noise using Pauli Propagation
• Characterization of overparametrization in the simulation of realistic quantum systems
• Measurement-based quantum machine learning
• More buck-per-shot: Why learning trumps mitigation in noisy quantum sensing
• Introducing Reduced-Width QNNs, an AI-inspired Ansatz Design Pattern