We present a recurrent neural network-based approach for ground state preparation utilizing mid-circuit measurement and feedback. Unlike previous methods that use machine learning solely as an optimizer, our approach dynamically adjusts quantum circuits based on real-time measurement outcomes and learns distinct preparation protocols for different Hamiltonians. Notably, our machine learning algorithm consistently identifies a state preparation strategy wherein all initial states are first steered toward an intermediate state before transitioning to the target ground state. We demonstrate that performance systematically improves as a larger fraction of ancilla qubits are utilized for measurement and feedback, highlighting the efficacy of mid-circuit measurements in state preparation tasks.
We present a recurrent neural network-based approach for ground state preparation utilizing mid-circuit measurement and feedback. Unlike previous methods that use machine learning solely as an optimizer, our approach dynamically adjusts quantum circuits based on real-time measurement outcomes and learns distinct preparation protocols for different Hamiltonians. Notably, our machine learning algorithm consistently identifies a state preparation strategy wherein all initial states are first steered toward an intermediate state before transitioning to the target ground state. We demonstrate that performance systematically improves as a larger fraction of ancilla qubits are utilized for measurement and feedback, highlighting the efficacy of mid-circuit measurements in state preparation tasks.
This work introduces a self-learning protocol that incorporates measurement and feedback into variational quantum circuits for efficient quantum state preparation. By combining projective measurements with conditional feedback, the protocol learns state preparation strategies that extend beyond unitary-only methods, leveraging measurement-based shortcuts to reduce circuit depth. Using the spin-1 Affleck-Kennedy-Lieb-Tasaki state as a benchmark, the protocol learns high-fidelity state preparation by overcoming a family of measurement induced local minima through adjustments of parameter update frequencies and ancilla regularization. Despite these efforts, optimization remains challenging due to the highly non-convex landscapes inherent to variational circuits. The approach is extended to larger systems using translationally invariant ansätze and recurrent neural networks for feedback, demonstrating scalability. Additionally, the successful preparation of a specific AKLT state with desired edge modes highlights the potential to discover new state preparation protocols where none currently exist. These results indicate that integrating measurement and feedback into variational quantum algorithms provides a promising framework for quantum state preparation.
Ground-state preparation is a critical task in quantum simulation and quantum computing, as it enables the study of correlated phases and the generation of entangled resource states. While measurement--feedback control has emerged as a promising route to state preparation, existing schemes either rely on handcrafted, task-specific policies or are designed using full quantum-state information that is unavailable in real experiments and becomes impractical for large many-body systems. Here we develop an adaptive measurement--feedback protocol based on reinforcement learning under partial observability. The controller uses only the history of experimentally accessible measurement outcomes to choose both the measurement operator and the feedback action in real time. To make training compatible with experiments, we introduce a stochastic terminal reward built from one-shot measurements of randomly sampled Hamiltonian components, avoiding unphysical full-state reconstruction while remaining an unbiased estimator of the target energy. We demonstrate the method by preparing ground states of the Bose--Hubbard model and by generating GHZ states, establishing a scalable and hardware-compatible route to quantum state preparation.
Dynamic quantum circuits (DQCs) incorporate mid-circuit measurements and gates conditioned on these measurement outcomes. DQCs can prepare certain long-range entangled states in constant depth, making them a promising route to preparing complex quantum states on devices with a limited coherence time. Almost all constructions of DQCs for state preparation have been formulated analytically, relying on special structure in the target states. In this work, we approach the problem of state preparation variationally, developing scalable tensor network algorithms which find high-fidelity DQC preparations for generic states. We apply our algorithms to critical states, random matrix product states, and subset states. We consider both DQCs with a fixed number of ancillae and those with an extensive number of ancillae. Even in the few ancillae regime, the DQCs discovered by our algorithms consistently prepare states with lower infidelity than a static quantum circuit of the same depth. Notably, we observe constant fidelity gains across system sizes and circuit depths. For DQCs with an extensive number of ancillae, we introduce scalable methods for decoding measurement outcomes, including a neural network decoder and a real-time decoding protocol. Our work demonstrates the power of an algorithmic approach to generating DQC circuits, broadening their scope of applications to new areas of quantum computing.
Recently developed quantum computers support mid-circuit measurements, allowing qubit measurement, resetting, and reuse during the execution of quantum circuits. This capability opens new opportunities for optimizing quantum circuit compilation, particularly in the context of Noisy Intermediate-Scale Quantum devices. This paper explores the role of machine learning in integrating mid-circuit measurements into the compilation process, focusing on strategies to minimize errors. Specifically, we propose a reinforcement learning method to enable physically-aware qubit reuse and a machine learning method to support physically-aware synthesis-based qubit reuse. We conduct an extensive experimental evaluation to demonstrate the effectiveness of our methods using various quantum algorithms executed on an IBM quantum computer.
In the Hamiltonian-based quantum dynamics, to estimate Hamiltonians from the measured data is a vital topic.In this work, we propose a recurrent neural network to learn the target Hamiltonians from the temporal records of single-qubit measurements, which does not require the ground states and only measures single-qubit observables.It is applicable on both time-independent and time-dependent Hamiltonians and can simultaneously capture the magnitude and sign of Hamiltonian parameters.Taking the Hamiltonians with the nearest-neighbor interactions as numerical examples, we trained our recurrent neural networks to learn different types of Hamiltonians with high accuracy.The study also shows that our method has good robustness against the measurement noise and decoherence effect.Therefore, it has widespread applications in estimating the parameters of quantum devices and characterizing the Hamiltonian-based quantum dynamics.
This thesis introduces and systematically analyzes novel measurement-based quantum state preparation protocols in one dimension, leveraging non-unitary dynamics, quantum measurements, and adaptive feedback mechanisms. Additionally, a stochastic sampling-based approach for ground state search using Projected Entangled Pair States (PEPS) in two dimensions is investigated, further expanding the computational toolkit for simulating complex quantum states. A central contribution is the development of a protocol that employs periodic measurement and resetting of ancilla systems, explicitly applied to the preparation of the one-dimensional Affleck-Kennedy-Lieb-Tasaki (AKLT) state. We identify an optimal reset interval, balancing entanglement generation and convergence speed, and demonstrate robustness against realistic experimental noise such as dephasing. Furthermore, this thesis introduces a self-learning adaptive measurement-feedback framework within variational quantum circuits (VQCs). The proposed framework autonomously discovers efficient, constant-depth strategies for deterministic preparation of specific AKLT edge states, outperforming existing analytical protocols. This analysis uncovers and mitigates previously unknown optimization challenges unique to measurement-feedback-based variational quantum circuits. The thesis also advances two-dimensional ground state search by investigating a sampling-based optimization approach for finite PEPS. It demonstrates that PEPS representing physically realistic states are computationally tractable within this sampling framework, in contrast to random states, whose complexity makes them prohibitively difficult to contract. To quantitatively distinguish between computationally easy and hard-to-contract states, a novel diagnostic measure is introduced. The developed methodology is successfully applied to physically significant states, including chiral spin liquids and Rydberg atom arrays with long-range interactions, highlighting its effectiveness in simulating complex states relevant to state-of-the-art cold atom experiments. Overall, the thesis provides methodological and conceptual insights into measurement-based quantum state preparation, optimization landscapes, and tensor network simulations.