Information Scrambling in Quantum Neural Networks

Abstract

The quantum neural network is one of the promising applications for near-term noisy intermediate-scale quantum computers. A quantum neural network distills the information from the input wavefunction into the output qubits. In this Letter, we show that this process can also be viewed from the opposite direction: the quantum information in the output qubits is scrambled into the input. This observation motivates us to use the tripartite information, a quantity recently developed to characterize information scrambling, to diagnose the training dynamics of quantum neural networks. We empirically find strong correlation between the dynamical behavior of the tripartite information and the loss function in the training process, from which we identify that the training process has two stages for randomly initialized networks. In the early stage, the network performance improves rapidly and the tripartite information increases linearly with a universal slope, meaning that the neural network becomes less scrambled than the random unitary. In the latter stage, the network performance improves slowly while the tripartite information decreases. We present evidences that the network constructs local correlations in the early stage and learns large-scale structures in the latter stage. We believe this two-stage training dynamics is universal and is applicable to a wide range of problems. Our work builds bridges between two research subjects of quantum neural networks and information scrambling, which opens up a new perspective to understand quantum neural networks.

arXiv comment6 pages, 4 figures + 9 pages of supplemental material

References in corpus (25)

• Quantum Computing in the NISQ era and beyond
• Quantum Machine Learning
• Topological entanglement entropy
• A bound on chaos
• Barren plateaus in quantum neural network training landscapes
• Quantum Circuit Learning
• On the Convergence of Adam and Beyond
• Black holes and the butterfly effect
• Circuit-centric quantum classifiers
• Opening the Black Box of Deep Neural Networks via Information
• Chaos in quantum channels
• Efficient Learning for Deep Quantum Neural Networks
• A generative modeling approach for benchmarking and training shallow quantum circuits
• Hierarchical quantum classifiers
• Towards Quantum Machine Learning with Tensor Networks
• The Expressive Power of Parameterized Quantum Circuits
• Out-of-Time-Order Correlation for Many-Body Localization
• Differentiable Learning of Quantum Circuit Born Machine
• Machine Learning Topological Invariants with Neural Networks
• Quantum-Assisted Learning of Hardware-Embedded Probabilistic Graphical Models
• Unitary quantum perceptron as efficient universal approximator
• Learning and Inference on Generative Adversarial Quantum Circuits
• A Universal Training Algorithm for Quantum Deep Learning
• Making Trotters Sprint: A Variational Imaginary Time Ansatz for Quantum Many-body Systems
• Mutual Information Scaling and Expressive Power of Sequence Models

Cited by in corpus (38)

• Entanglement Devised Barren Plateau Mitigation
• Efficient measure for the expressivity of variational quantum algorithms
• Biology and medicine in the landscape of quantum advantages
• Thermofield dynamics: Quantum Chaos versus Decoherence
• Resource theory of quantum scrambling
• Generative quantum machine learning via denoising diffusion probabilistic models
• Expressivity of Quantum Neural Networks
• Imaginary components of out-of-time correlators and information scrambling for navigating the learning landscape of a quantum machine learning model
• Quantum Information Scrambling in Quantum Many-body Scarred Systems
• Scrambling Ability of Quantum Neural Networks Architectures
• A brief introduction to quantum algorithms
• Chaos and Complexity from Quantum Neural Network: A study with Diffusion Metric in Machine Learning
• Information scrambling in a collision model
• Barren plateaus from learning scramblers with local cost functions
• Quantifying scrambling in quantum neural networks
• Speed limits to the growth of Krylov complexity in open quantum systems
• Quantum steering as a witness of quantum scrambling
• Avoiding Barren Plateaus with Entanglement
• Dissipation induced information scrambling in a collision model
• Subsystem Information Capacity in Random Circuits and Hamiltonian Dynamics
• Thermal squeezing and nonlinear spectral shift of magnons in antiferromagnetic insulators
• Adversarial Robustness Guarantees for Quantum Classifiers
• Phase-transition-like behavior in information retrieval of a quantum scrambled random circuit system
• The Roles of Kerr nonlinearity in a Bosonic Quantum Neural Network
• Generative Invertible Quantum Neural Networks
• Information scrambling and entanglement in quantum approximate optimization algorithm circuits
• Non-parametric Semi-Supervised Learning in Many-body Hilbert Space with Rescaled Logarithmic Fidelity
• Robust Estimation of Nonlinear Properties of Quantum Processes
• Quantum speed limit for the OTOC from an open systems perspective
• Nearly query-optimal classical shadow estimation of unitary channels
• Information scrambling in quantum walks: Discrete-time formulation of Krylov complexity
• Optimal quantum reservoir learning in proximity to universality
• Dynamical transition in controllable quantum neural networks with large depth
• Equilibration of Non-interacting Photons and Quantum Signatures of Chaos
• Quantum Machine Learning with SQUID
• Direct Gradient Computation for Barren Plateaus in Parameterized Quantum Circuits
• Superior resilience to poisoning and amenability to unlearning in quantum machine learning
• A graph-theoretic approach to chaos and complexity in quantum systems