Abstract:A quadrillion dimensional Hilbert space hosted by a quantum processor with over 50 physical qubits has been expected to be powerful enough to perform computational tasks ranging from simulations of many-body physics to complex financial modeling. Despite few examples and demonstrations, it is still not clear how we can utilize such a large Hilbert space as a computational resource; in particular, how a simple and small quantum system could solve non-trivial computational tasks. In this paper, we show a simple Ising model capable of performing such non-trivial computational tasks in a quantum neural network model. An Ising spin chain as small as ten qubits can solve a practical image classification task with high accuracy. To evaluate the mechanism of its computation, we examine how the symmetries of the Hamiltonian would affect its computational power. We show how the interplay between complexity and integrability/symmetries of the quantum system dictates the performance as quantum neural network.

What problem does this paper attempt to address?

This paper discusses how to use simple quantum systems, such as the Ising model, to perform non-trivial computational tasks, particularly in the context of quantum neural network models. The researchers demonstrate, through the Quantum Extreme Reservoir Computing (QERC) benchmark test, that Ising chains containing only 10 qubits can accurately perform image classification tasks. They analyze how the symmetry of the Hamiltonian affects computational capability and point out that the complexity of the quantum system and its interaction with integrability/symmetry determine its performance as a quantum neural network. The paper mentions that although processors with hundreds of qubits can simulate various physical, chemical, and biological systems, how to effectively utilize these large Hilbert spaces for computation remains unclear. The authors propose that even small-scale quantum systems, such as the Ising model, can handle complex computational tasks, challenging some traditional views in quantum machine learning and quantum reservoir computing. Through QERC, information is first encoded into the quantum system, then mapped to complex states through quantum dynamics, and finally extracted and converted into classical data for classical neural network processing through measurement steps. The researchers chose QERC as the benchmark because it does not require optimized control of the quantum components and can demonstrate the required accuracy with a small number of qubits. Experimental results show high accuracy in handling handwritten digit recognition tasks with different types of Ising models (such as ZZ-Ising and XX-Ising) as well as the ZZ-X model (which includes ZZ-Ising dynamics and single qubit rotations). In particular, the XX-Ising model works effectively even without a transverse magnetic field, indicating the importance of multi-body interactions for computational tasks. Additionally, the study found that the non-periodicity and non-integrability of the quantum system may be advantageous for the performance of quantum neural networks, while simple periodic dynamics may lead to performance degradation. The paper also discusses the impact of entanglement and system symmetry (or integrability) on the performance of quantum reservoir computing. Overall, this work provides a new perspective for understanding and utilizing quantum systems for computation, and provides guidance for designing quantum reservoirs.

Simple Hamiltonian dynamics is a powerful quantum processing resource

Variational Quantum Simulation of Chemical Dynamics with Quantum Computers

Qubits on programmable geometries with a trapped-ion quantum processor

Robustly learning the Hamiltonian dynamics of a superconducting quantum processor

Ground state energy and magnetization curve of a frustrated magnetic system from real-time evolution on a digital quantum processor

Quantum Supremacy for Simulating A Translation-Invariant Ising Spin Model

Practical and Efficient Hamiltonian Learning

Probing critical states of matter on a digital quantum computer

Robust and Efficient Hamiltonian Learning

Quantum simulation of one-dimensional fermionic systems with Ising Hamiltonians

Noisy intermediate scale quantum simulation of time dependent Hamiltonians

Small-world complex network generation on a digital quantum processor

Simulating spin dynamics with quantum computers

Riemannian quantum circuit optimization for Hamiltonian simulation

Programmable quantum simulations on a trapped-ions quantum computer with a global drive

Hamiltonian simulation for time-evolving partial differential equation by scalable quantum circuits

Hybrid Oscillator-Qubit Quantum Processors: Simulating Fermions, Bosons, and Gauge Fields

Simulation of the 1d XY model on a quantum computer

Classically optimized Hamiltonian simulation

Simulating nonlinear optical processes on a superconducting quantum device