One of the key research focuses in quantum annealing is the design and optimization of annealing schedules to enhance computational efficiency, enabling large-scale applications. QuantumZero (QZero) pioneered the integration of Monte Carlo Tree Search (MCTS) with neural networks to autonomously design annealing schedules within a hybrid quantum-classical framework. This approach is distinguished by its ability to enhance the performance of Monte Carlo Tree Search through the integration of neural networks, enabling the efficient design of annealing paths even with limited annealing time. The paper presents an optimized QZero method based on intuitive reasoning theory and MindSpore, which further enhances QZero's ability to conserve computational resources and resist noise. In terms of learning efficiency, the optimized QZero algorithm improves the convergence speed of the neural network by 93% compared to the original algorithm. Notably, the average number of quantum annealing queries required to achieve 99% fidelity is reduced by 45.09%. Regarding noise resistance, the optimized QZero algorithm requires 34.27% fewer quantum annealing queries to reach 990% fidelity compared to the original algorithm. The optimized QZero algorithm demonstrates strong competitiveness in optimizing quantum annealing schedules.
One of the key research focuses in quantum annealing is the design and optimization of annealing schedules to enhance computational efficiency, enabling large-scale applications. QuantumZero (QZero) pioneered the integration of Monte Carlo Tree Search (MCTS) with neural networks to autonomously design annealing schedules within a hybrid quantum-classical framework. This approach is distinguished by its ability to enhance the performance of Monte Carlo Tree Search through the integration of neural networks, enabling the efficient design of annealing paths even with limited annealing time. The paper presents an optimized QZero method based on intuitive reasoning theory and MindSpore, which further enhances QZero's ability to conserve computational resources and resist noise. In terms of learning efficiency, the optimized QZero algorithm improves the convergence speed of the neural network by 93% compared to the original algorithm. Notably, the average number of quantum annealing queries required to achieve 99% fidelity is reduced by 45.09%. Regarding noise resistance, the optimized QZero algorithm requires 34.27% fewer quantum annealing queries to reach 990% fidelity compared to the original algorithm. The optimized QZero algorithm demonstrates strong competitiveness in optimizing quantum annealing schedules.
Quantum annealing is a practical approach to execute the native instruction set of the adiabatic quantum computation model. The key of running adiabatic algorithms is to maintain a high success probability of evolving the system into the ground state of a problem-encoded Hamiltonian at the end of an annealing schedule. This is typically done by executing the adiabatic algorithm slowly to enforce adiabacity. However, properly optimized annealing schedule can accelerate the computational process. Inspired by the recent success of DeepMind's AlphaZero algorithm that can efficiently explore and find a good winning strategy from a large combinatorial search with a neural-network-assisted Monte Carlo Tree Search (MCTS), we adopt MCTS and propose a neural-network-enabled version, termed QuantumZero (QZero), to automate the design of an optimal annealing schedule in a hybrid quantum-classical framework. The flexibility of having neural networks allows us to apply transfer-learning technique to boost QZero's performance. We find both MCTS and QZero to perform very well in finding excellent annealing schedules even when the annealing time is short in the 3-SAT examples we consider in this study. We also find MCTS and QZero to be more efficient than many other leading reinforcement leanring algorithms for the task of desining annealing schedules. In particular, if there is a need to solve a large set of similar problems using a quantum annealer, QZero is the method of choice when the neural networks are first pre-trained with examples solved in the past.
Quantum-driven Bayesian optimization (QBO) presents a novel approach to neural network architecture search, employing quantum annealing or QAOA to efficiently explore the vast design space. This research investigates the potential of quantum algorithms to accelerate the discovery of optimal neural network architectures, addressing limitations of traditional methods like grid search. The core mechanism centers on utilizing quantum annealing to sample neural network designs, enabling a significantly broader exploration of potential architectures compared to classical methods. The study will analyze the impact of quantum optimization on the search process, focusing on speed and the ability to uncover unexpected architectural patterns. The paper will detail the implementation of a quantum algorithm, its performance compared to classical approaches, and the potential benefits of leveraging quantum mechanics for neural network design. This work aims to provide a fundamental understanding of the application of quantum optimization to neural network architecture discovery.
Over the past decades, the operations research community has developed numerous effective optimization algorithms, yet quantum computing is emerging as a new computational paradigm with the potential to approach optimization problems more efficiently. Grover's algorithm offers a provable speedup for combinatorial optimization, but its circuit depth places it beyond current noisy intermediate-scale quantum (NISQ) devices. A more accessible alternative is to reformulate the optimization problem as a quadratic unconstrained binary optimization (QUBO) problem and apply quantum annealing; however, practical problem instances remain out of reach for existing hardware. We introduce AtomTreeSearch, a hybrid classical-quantum algorithm that integrates a quantum subroutine natively implementable on neutral-atom quantum computers within a Monte Carlo Tree Search framework. At each expansion step, a maximal weighted independent set of candidate actions provided by the quantum processor is selected, and these collective actions are performed to obtain a child node. We benchmark our method on the Traveling Salesman Problem, with instances of up to 60 cities on random Euclidean instances and up to 100 cities on TSPLIB instances. Our hybrid algorithm generally matches or outperforms both OR-Tools and simulated annealing on these instances, and we find that the quantum subroutine produces more diverse and higher-quality branches compared to classical alternate subroutines. These results suggest that carefully scoped quantum subroutines embedded in classical search frameworks represent a promising path toward near-term quantum utility in combinatorial optimization.
The design of deep neural network (DNN) architectures is a complex and computationally intensive task. Traditional methods often rely on manual tuning or random search, which are inefficient in exploring the exponentially large architectural space. This paper proposes a novel approach to DNN architecture optimization leveraging quantum-inspired algorithms. Specifically, we explore the application of algorithms inspired by quantum annealing and quantum Monte Carlo methods to efficiently search for optimal DNN configurations. Our framework aims to minimize the trade-off between accuracy, latency, and energy consumption – key considerations in real-world DNN deployments. We present a theoretical framework and outline a methodology for integrating these quantum-inspired techniques. The core idea is to map the architectural design problem onto a suitable quantum system, allowing for the parallel exploration of potential solutions. This research provides a new direction for automated DNN design, potentially leading to significant improvements in performance and efficiency.
This paper investigates the potential of quantum computing to significantly enhance the efficiency of Monte Carlo Tree Search (MCTS) algorithms. MCTS, a cornerstone of modern artificial intelligence, particularly in game playing, relies heavily on computationally intensive simulations to evaluate game states and guide the search process. We propose a hybrid quantum-classical algorithm that leverages quantum superposition and entanglement to achieve a parallel exploration of the vast search space inherent in MCTS. The core mechanism involves utilizing quantum computers for specific tasks within the MCTS workflow, primarily the evaluation of game states. We formulate the problem as a quantum optimization task, mapping the state evaluation to a quantum objective function. The algorithm employs a classical controller to manage the quantum computation, orchestrating the simulations and extracting information. Preliminary theoretical analysis suggests a potential exponential speedup in state evaluation compared to classical simulations, contingent on the quality of the quantum hardware and the effectiveness of the hybrid approach. This research contributes to the broader field of quantum-accelerated AI and provides a framework for future investigations into the practical application of quantum computing within complex decision-making algorithms.
The recent advances in the world of quantum technologies have prompted the development of various quantum-based algorithms, some of which are suitable to run on available quantum devices. Two leading candidates in this area are Quantum Annealing (QA) and hybrid quantum-classical variational algorithms, such as the Quantum Approximate Optimization Algorithm (QAOA). In the first part of the Thesis, we re-examine the issue of whether Simulated Quantum Annealing (SQA) with Path Integral Monte Carlo has anything to do with the coherent QA Schrödinger dynamics. We do that by studying the random quantum Ising Chain in a transverse field. In the second part of the Thesis we address the issue of schedule optimization in both {em digitized}-QA and QAOA. Traditional schedule optimization in QA requires spectral information on the problem (e.g., location of the minimum gaps), which is in most cases inaccessible. On the other hand, in QAOA, the issue of schedule optimization --- alias, the optimization of the variational parameters --- is often an expensive task that constitutes a bottleneck for the algorithm. We show that, by combining the framework of QAOA and digitized-QA, optimal digitized-QA protocols can be constructed efficiently. When using this approach, the computational cost of schedule optimization is also significantly reduced, leading to a computational advantage both over a linear-schedule QA, and over an unstructured QAOA search. While studying these issues, we also developed rigorous variational bounds that provide insight on the best possible performance that one can hope to obtain from a digital QAOA circuit.