Abstract:The constantly increasing dimensionality of artificial quantum systems demands for highly efficient methods for their characterization and benchmarking. Conventional quantum tomography fails for larger systems due to the exponential growth of the required number of measurements. The conceptual solution for this dimensionality curse relies on a simple idea - a complete description of a quantum state is excessive and can be discarded in favor of experimentally accessible information about the system. The probably approximately correct (PAC) learning theory has been recently successfully applied to a problem of building accurate predictors for the measurement outcomes using a dataset which scales only linearly with the number of qubits. Here we present a constructive and numerically efficient protocol which learns a tensor network model of an unknown quantum system. We discuss the limitations and the scalability of the proposed method.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to efficiently learn and reconstruct the tensor network model of an unknown quantum state in a quantum system, especially in a multi - qubit system. Traditional quantum tomography methods become impractical for larger systems because the number of measurements required grows exponentially with the number of qubits. This paper proposes a new algorithm that uses only single - qubit measurements and is numerically efficient and can learn the tensor network model of an unknown quantum system. ### Specific Problem Description 1. **Characterization and Benchmarking of High - Dimensional Quantum Systems**: - As the dimensions of artificial quantum systems continue to increase, there is an increasingly urgent need for efficient characterization and benchmarking of these systems. - Traditional quantum tomography methods fail when dealing with larger systems because the number of measurements required grows exponentially. 2. **Reducing the Number of Measurements**: - By using the tensor network model and shadow tomography techniques, reduce the number of required measurements so that it is linearly related to the number of qubits rather than exponentially. 3. **Learnability and Predictive Power of the Model**: - Research how to use limited measurement data to build a model that can accurately predict future measurement results. - Explore the limitations and extensibility of the model. ### Solutions 1. **Tensor Network Representation**: - Use tensor networks (especially matrix product states and matrix product operators) to represent quantum states, thereby reducing storage and computational complexity. 2. **Shadow Tomography**: - Utilize shadow tomography techniques to learn the tensor network model of a quantum state through a small number of single - qubit measurements. - This method is based on the Probably Approximately Correct (PAC) learning theory and can extract useful information from experimental data in a relatively short time. 3. **Numerical Optimization**: - Use the stochastic gradient descent algorithm to minimize the loss function and optimize the model parameters to improve prediction accuracy. ### Experimental Verification - The effectiveness of this algorithm has been verified through numerical simulation and actual quantum circuit experiments. - Simulation results show that the minimum size of the training data set increases linearly with the number of qubits and can successfully predict the results of future measurements. ### Conclusion This paper proposes a practical algorithm that can efficiently learn and reconstruct the tensor network model of a multi - qubit system under experimentally feasible conditions. This algorithm not only reduces the number of required measurements but is also numerically efficient and is suitable for current NISQ devices.

Learning the tensor network model of a quantum state using a few single-qubit measurements

Predictive Modelling of Quantum Process with Neural Networks

Quantum Machine Learning Tensor Network States

Learning topological states from randomized measurements using variational tensor network tomography

Neural-network quantum state tomography

Experimental characterization of quantum processes: a selective and efficient method in arbitrary finite dimension

Local-measurement-based Quantum State Tomography Via Neural Networks

Many-body quantum state tomography with neural networks

Quantum State Tomography of Multi-Qubit Systems — a Comparative Study

Efficient Quantum Mixed-State Tomography with Unsupervised Tensor Network Machine Learning

High Dimensional Quantum Machine Learning With Small Quantum Computers

The Learnability of Unknown Quantum Measurements

Scalable tomography of many-body quantum environments with low temporal entanglement

On how neural networks enhance quantum state tomography with constrained measurements

Learning quantum many-body systems from a few copies

Deep learning of quantum entanglement from incomplete measurements

Quantum state preparation using tensor networks

Finite-Depth Preparation of Tensor Network States from Measurement

Deep Learning Quantum States for Hamiltonian Estimation

Quantum state tomography with locally purified density operators and local measurements

Experimental Realization of a Quantum Image Classifier Via Tensor-Network-based Machine Learning