Training of quantum Neural Networks can be affected by barren plateaus-flat areas in the landscape of the cost function, which impede the model optimisation. While there exist methods of dealing with barren plateaus, ...
详细信息
Training of quantum Neural Networks can be affected by barren plateaus-flat areas in the landscape of the cost function, which impede the model optimisation. While there exist methods of dealing with barren plateaus, they could reduce the model's effective dimension-the measure of its capacity to learn. This paper therefore reports an investigation of four barren plateaus countermeasures, i.e. restricting the model's circuit depth and relying on the local cost function;layer-by-layer circuit pre-training;relying on the circuit block structure to support its initialisation;as well as, model creation without any constraints. Several experiments were conducted to analyse the impact of each countermeasure on the model training, its subsequent ability to generalise and its effective dimension. The results reveal which of the approaches enhances or impedes the quantum model's capacity to learn, which gives more predictable learning outcomes, and which is more sensitive to training data. Finally, the paper provides some recommendations on how to utilise the effective dimension measurements to assist quantum model development.
Recently, proposed algorithms for quantum computing and generated quantum computer technologies continue to evolve. On the other hand, machine learning has become an essential method for solving many problems such as ...
详细信息
Recently, proposed algorithms for quantum computing and generated quantum computer technologies continue to evolve. On the other hand, machine learning has become an essential method for solving many problems such as computer vision, natural language processing, prediction and classification. quantum machine learning is a new field devel-oped by combining the advantages of these two primary methods. As a hybrid approach to quantum and classical computing, variational quantum circuits are a form of machine learning that allows predicting an output value against input variables. In this study, the effects of superposition and entanglement on weather forecasting, were investigated using a variational quantum circuit model when the dataset size is small. The use of the entanglement layer between the variational layers has made significant improvements on the circuit performance. The use of the superposition layer before the data encoding layer resulted in the use of less variational layers.
We use an artificial neural network (ANN) model to identify the entanglement class of an experimentally generated three-qubit pure state drawn from one of the six inequivalent classes under stochastic local operations...
详细信息
We use an artificial neural network (ANN) model to identify the entanglement class of an experimentally generated three-qubit pure state drawn from one of the six inequivalent classes under stochastic local operations and classical communication (SLOCC). The ANN model is also able to detect the presence of genuinely multipartite entanglement (GME) in the state. We apply data science techniques to reduce the dimensionality of the problem, which corresponds to a reduction in the number of required density matrix elements to be computed. The ANN model is first trained on a simulated dataset containing randomly generated states and is later tested and validated on noisy experimental three-qubit states cast in the canonical form and generated on a nuclear magnetic resonance (NMR) quantum processor. We benchmark the ANN model via support vector machines (SVMs) and K-nearest neighbor (KNN) algorithms and compare the results of our ANN-based entanglement classification with existing three-qubit SLOCC entanglement classification schemes such as 3-tangle and correlation tensors. Our results demonstrate that the ANN model can perform GME detection and SLOCC class identification with high accuracy, using a priori knowledge of only a few density matrix elements as inputs. Since the ANN model works well with a reduced input dataset, it is an attractive method for entanglement classification in real-life situations with limited experimental data sets.
Sparse coding provides a versatile framework for efficiently capturing and representing crucial data (information) concisely, which plays an essential role in various computer science fields, including data compressio...
详细信息
Sparse coding provides a versatile framework for efficiently capturing and representing crucial data (information) concisely, which plays an essential role in various computer science fields, including data compression, feature extraction, and general signal processing. Here, we propose symmetric quantum neural networks for realizing sparse coding and decoding algorithms. Our networks consist of multi-layer, two-level unitary transformations that are naturally suited for optical circuits. Specifically, the two networks we propose can be efficiently trained together or separately via a quantum natural gradient descent algorithm. Utilizing the trained model, we achieve coding and decoding of sparse data including sparse classical data of binary and grayscale images, as well as sparse quantum data that are quantum states in a certain smaller subspace. The results demonstrate an accuracy of 98.77% for image reconstruction and a fidelity of 97.68% for quantum state revivification. Our quantum sparse coding and decoding model offers improved generalization and robustness compared to the classical model, giving insights to further research on quantum advantages in artificial neural networks.
Since the first reports of quantum dot (QD) based quasi-molecules formed by near-field-coupled QDs, how to design and apply such systems in the field of quantuminformationprocessing are still in learning process. Ex...
详细信息
Since the first reports of quantum dot (QD) based quasi-molecules formed by near-field-coupled QDs, how to design and apply such systems in the field of quantuminformationprocessing are still in learning process. Experimentally, single-particle spectroscopy is the method of choice for studying kinetic processes in such structures. However, current efforts focus on detecting all the photons emitted by the quasi-molecule, making it difficult to understand the precise changes occurring in each of the coupled QDs. In this study, a different approach is implemented by adapting the kinetic Monte Carlo algorithms to gain comprehensive insights into the temporal evolution of each coupled QD. This encompasses not only photon emission events, but also the full range of kinetic processes, including exciton transfer between the QDs. It is, therefore, important to gain insight into the adaptation pathways caused by exciton exchange between near-field coupled QDs. The attention is strict to a system of two interacting QDs in different coupling regimes. The results point to new mechanisms of sequential/cascade excitonic relaxation, which are not typical for a QD in an isolated state. For this reason, the methodology presented here seems to be significant and complementary to the experimental approach.
quantum error correction techniques are important for implementing fault-tolerant quantum computation, and topological quantum error correcting codes provide feasibility for implementing large-scale fault-tolerant qua...
详细信息
quantum error correction techniques are important for implementing fault-tolerant quantum computation, and topological quantum error correcting codes provide feasibility for implementing large-scale fault-tolerant quantum computation. Here, we propose a deep reinforcement learning framework for implementing quantum error correction algorithms for errors on heavy hexagonal codes. Specifically, we construct the double deep Q learning model with policy reuse method, so that the decoding agent does not have to explore the learning from scratch when dealing with new error syndrome, but instead reuses past policies, which can reduce the computational complexity. And the double deep Q network can avoid the problem of threshold being overestimated and get the true decoding threshold. Our experimental results show that the error correction accuracy of our decoder can reach 91.86%. Different thresholds are obtained according to the code distance, which is 0.0058 when the code distance is 3, 5, 7, and 0.0065 when the code distance is 5, 7, 9, both higher than that of the classical minimum weight perfect matching decoder. Compared to the threshold of the MWPM decoder under the depolarizing noise model, the threshold of our decoder is improved by 32.63%, which enables better fault-tolerant quantum computation.
Machine learning techniques have achieved impressive results in recent years and the possibility of harnessing the power of quantum physics opens new promising avenues to speed up classical learning methods. Rather th...
详细信息
Machine learning techniques have achieved impressive results in recent years and the possibility of harnessing the power of quantum physics opens new promising avenues to speed up classical learning methods. Rather than viewing classical and quantum approaches as exclusive alternatives, their integration into hybrid designs has gathered increasing interest, as seen in variational quantumalgorithms, quantum circuit learning, and kernel methods. Here we introduce deep hybrid classical-quantum reservoir computing for temporal processing of quantum states where information about, for instance, the entanglement or the purity of past input states can be extracted via a single-step measurement. We find that the hybrid setup cascading two reservoirs not only inherits the strengths of both of its constituents but is even more than just the sum of its parts, outperforming comparable non-hybrid alternatives. The quantum layer is within reach of state-of-the-art multimode quantum optical platforms while the classical layer can be implemented in silico.
There are quantumalgorithms for finding a function f satisfying a set of conditions, such as solving partial differential equations, and these achieve exponential quantum speedup compared to existing classical method...
详细信息
There are quantumalgorithms for finding a function f satisfying a set of conditions, such as solving partial differential equations, and these achieve exponential quantum speedup compared to existing classical methods, especially when the number d of the variables of f is large. In general, however, these algorithms output the quantum state which encodes f in the amplitudes, and reading out the values of f as classical data from such a state can be so time-consuming that the quantum speedup is ruined. In this study, we propose a general method for this function readout task. Based on the function approximation by a combination of tensor network and orthogonal function expansion, we present a quantum circuit and its optimization procedure to obtain an approximating function of f that has a polynomial number of degrees of freedom with respect to d and is efficiently evaluable on a classical computer. We also conducted a numerical experiment to approximate a finance-motivated function to demonstrate that our method works.
quantumalgorithms typically comprise classical pre- and post-processing tasks, making quantum applications inherently hybrid. To facilitate the integration and orchestration challenges arising when building hybrid ap...
详细信息
ISBN:
(纸本)9783031681646;9783031681653
quantumalgorithms typically comprise classical pre- and post-processing tasks, making quantum applications inherently hybrid. To facilitate the integration and orchestration challenges arising when building hybrid applications, a quantum-specific workflow modeling extension was introduced. However, it lacks support for state-of-the-art techniques that have emerged in recent years, e.g., circuit cutting and warm-starting. As the usage of these complex techniques further complicates integrating quantum circuit executions with classical applications, we introduce new activity types and data objects to facilitate their integration. Moreover, we formalize the quantum workflow modeling extension by presenting a metamodel as well as transformation algorithms, ensuring its compatibility with existing workflow languages. Furthermore, we showcase the practical feasibility of our approach by presenting a system architecture, a prototypical implementation, and a case study.
In the noisy intermediate-scale quantum (NISQ) era, the computing power displayed by quantum computing hardware may be more advantageous than classical computers, but the emergence of the barren plateau (BP) has hinde...
详细信息
In the noisy intermediate-scale quantum (NISQ) era, the computing power displayed by quantum computing hardware may be more advantageous than classical computers, but the emergence of the barren plateau (BP) has hindered quantum computing power and cannot solve large-scale problems. This summary analyzes the phenomenon of the BP in the quantum neural network that is rapidly developing in the NISQ era. This article will review the research status of the BP problem in the quantum neural network (QNN) in the past five years from the analysis of the source of the BP, the current stage solution, and the future research direction. First of all, the source of the BP was briefly explained and then classified the BP solution from different perspectives, including quantum embedding in QNN, ansatz parameter selection and structural design, and optimization algorithms. Finally, the BP problem in the QNN is summarized, and the research direction for solving problems in the future is made.
暂无评论