Bulletin of the American Physical Society
APS March Meeting 2021
Volume 66, Number 1
Monday–Friday, March 15–19, 2021; Virtual; Time Zone: Central Daylight Time, USA
Session V32: Quantum Machine Learning IILive
|
Hide Abstracts |
Sponsoring Units: DQI GDS Chair: Alba Cervera-Lierta, Univ of Toronto |
Thursday, March 18, 2021 3:00PM - 3:12PM Live |
V32.00001: A Quantum Reservoir Computing Approach to Image Classification Fangjun Hu, Saeed Khan, Gerasimos Angelatos, Hakan E Tureci Recent proposals of Quantum Neural Networks (QNNs) and their implementations in near-term quantum hardware has highlighted the severe limitations imposed by the associated resource requirements. The practical and useful implementation of quantum neural networks to scale has the address the question of the optimal approach to encoding the information to be processed, and the subsequent extraction of the processed information from a large quantum system. In this work, we consider and analyze the efficacy of a reservoir computing approach to address these issues. We propose a superconducting quantum circuit network as a physical reservoir processor and present a unified description of the device operation, from classical information input to computational output via quantum measurement. We model a small experimentally-realizable network, and compare its performance on a pattern recognition task to that of recent QNN approaches [1]. |
Thursday, March 18, 2021 3:12PM - 3:24PM Live |
V32.00002: Neuromorphic computing with single-element quantum reservoirs Luke Govia, William D Kalfus, Guilhem Ribeill, Graham E Rowlands, Hari Krovi, Thomas A Ohki We study the noise-resilient neuromorphic computing scheme of reservoir computing with a quantum system as a reservoir. We consider quantum reservoirs formed by a single physical element, such as can be implemented in near-term, NISQ-era devices by a quantum nonlinear oscillator. By studying the performance of our single-element reservoirs on signal processing and memory capacity benchmarks, we demonstrate computational capability expanding with Hilbert space dimension, and quantum advantage arising from the intrinsic nonlinearity of quantum measurement. Beyond quantum reservoir computing, the latter may have impact across quantum machine learning. We study the impact of realistic experimental conditions such as noise and parameter fluctuations, and discuss near-term implementations. Our results show that single-element quantum reservoir computing is an attractive modality for quantum information processing on near-term hardware. |
Thursday, March 18, 2021 3:24PM - 3:36PM Live |
V32.00003: Quantum Learning at High Temperatures in a Dissipative Electronic System John Miller, Martha Villagran, Jarek Wosik, Ayo Kolapo Applications of quantum machine learning, and more generally quantum information processing, would expand enormously if dissipative quantum devices could be developed to show enhanced properties even at high temperatures. The charge density wave (CDW) is an electron condensate robust at temperatures that can exceed that of the human body – even the boiling point of water – in some systems. Several CDW systems show evidence for innate quantum learning in a highly dissipative environment. This includes a pulse-duration memory effect showing rapid learning – 1-3 training pulses vs. 100’s to 1000’s needed in classical simulations. Additional quantum behaviors include transport behavior indicating time-correlated, coherent tunneling of CDW electrons, and quantum interference in CDW rings and crystals with columnar defects. We discuss proposed concepts that exploit such phenomena, including a CDW quantum reservoir computing concept and quantum devices based on patterned ion implantation of CDW materials. |
Thursday, March 18, 2021 3:36PM - 3:48PM Live |
V32.00004: Storage properties of a quantum perceptron Aikaterini Gratsea, Valentin Kasper, Maciej Lewenstein Artificial neural networks are central for machine learning algorithms and artificial intelligence. The generalization to quantum artificial networks has inspired a lot of research in recent years. Different architectures for quantum perceptrons have been proposed, but the abilities of such a quantum machine remain debated. In this work, we explore the storage capacity of a specific quantum perceptron architecture. We use techniques of statistical mechanics and connect the storage capacity of the quantum perceptron with the theory of spin glasses. As a result, if the activation threshold is close to zero, the storage capacity of the quantum perceptron scales exponentially with the number of physical spins. If the activation threshold is close to one, the storage capacity decreases rapidly. We validate our result numerically and present a phase transition between the spin glass and para-magnetic phase. Finally, the here presented work inspires further studies of quantum neural networks using techniques of statistical physics. |
Thursday, March 18, 2021 3:48PM - 4:00PM Live |
V32.00005: Kerr Network Reservoir Computing for Quantum State Measurement Gerasimos Angelatos, Saeed Khan, Hakan E Tureci Quantum measurement protocols combining fast high fidelity readout with rapid calibration are increasingly important as device sizes grow in the NISQ era. Here we propose reservoir processing as a hardware-based solution to superconducting qubit readout. We consider a small network of Kerr oscillators, implementable with minimal hardware overhead, and theoretically analyze its ability to operate as a reservoir computer and classify stochastic time-dependent signals subject to quantum statistical features. We then apply this Kerr network reservoir computer to multi-qubit readout in a regime of multiplexing that presents maximal cross-talk between the readout channels. We demonstrate rapid multinomial classification of these measurement trajectories with a fidelity exceeding that of conventional filtering approaches. This reservoir computing framework avoids computationally expensive training standard for neural-networks, and requires orders-of-magnitude less training data than an optimal matched filter for the same task. Our results indicate that an unoptimized Kerr network can operate as a low latency analog processor at the computational edge and provide rapid and robust processing of quantum state measurement. |
Thursday, March 18, 2021 4:00PM - 4:12PM Live |
V32.00006: Quantum Thermodynamics of Quantum Boltzmann Machines Sangchul Oh, Sabre Kais Quantum Boltzmann machines are the quantum extension of classical Boltzmann machines. The classical Boltzmann machine is a generative probabilistic model to find a Boltzmann distribution, given classical data. In Ref. [1], the learning process of the classical restricted Boltzmann machine was analyzed in the context of statistical mechanics. The entropy, free energy, work, and Jarzynski equality were investigated. Here we study quantum thermodynamics of the training process of the quantum restricted Boltzmann machine with the hidden and the visible units for data given by density matrices. Similar to the thermodynamic analysis of the classical restricted Boltzmann machine, we calculate how internal energy, free energy and entropy, and entanglement between visible and hidden layers of the quantum restricted Boltzmann machine change during the learning. Quantum Jarzynski equality is examined. Also, we discuss how counter-diabatic driving could accelerate the learning of classical and quantum Boltzmann machines. |
Thursday, March 18, 2021 4:12PM - 4:24PM Live |
V32.00007: Quantum-assisted GAN networks for particle shower simulation Andrea Delgado This work explores the applicability of quantum machine learning (QML) methods, especially quantum-classical associative adversarial networks, to tune the training complexity for generative models for particle shower simulation. These QML methods will allow for fast and reliable simulations of particle showers in calorimeters in subatomic physics experiments if successful. Our approach will train generative layers that implement quantum versions of generative adversarial network (GAN) models. These implementations have been developed, but implementing these models on near-term hardware will suffer from the input-output size problem since quantum computing hardware is currently too small to handle the input data from HEP datasets directly. For this reason, we consider an approach for incorporating near-term quantum hardware into deep learning models in which a quantum model is trained and deployed on quantum hardware and used to implement a portion (e.g., a layer of a deep neural network) of the overall deep learning model. |
Thursday, March 18, 2021 4:24PM - 4:36PM Live |
V32.00008: Quantum generative adversarial networks with provable convergence Murphy Yuezhen Niu, Michael Broughton, Alexander Zlokapa, Masoud Mohseni, Vadim Smelyanskiy, Hartmut Neven Generative adversarial networks (GAN) are an important architecture in unsupervised machine learning, enabling the generation of new data produced by a desirable physical model by learning purely from an existing dataset without accessing the physical model itself. Since quantum states are the most general form of any physical data, realizing a GAN architecture in the quantum domain promises an even wider application of GANs in scientific discovery. In this work, we prove that the iterative training of a discriminator circuit against a generator circuit of previously proposed quantum GANs does not converge for certain initializations, but instead exhibits periodic oscillation between two configurations. We propose a new type of architecture for quantum generative adversarial networks (Q-GAN) to overcome such limitations by harnessing the entangling power of a quantum circuit and allowing the discriminator circuit to take both generator output and true quantum data as input. By adversarially learning efficient representations of quantum states, we prepare an approximate quantum random access memory (QRAM) and demonstrate its use in applications including the training of quantum neural networks. |
Thursday, March 18, 2021 4:36PM - 4:48PM Live |
V32.00009: Quantum Long Short-Term Memory Samuel Yen-Chi Chen, Shinjae Yoo, Yao-Lung L. Fang Recurrent neural networks (RNN) have been used to model data with sequential and temporal dependency. One of its variants, Long Short-Term Memory (LSTM), has been successfully applied to a wide spectrum of such tasks. In this talk, we propose a model of LSTM based on the hybrid quantum-classical paradigm, which we call QLSTM. The proposed architecture is successful in several testing cases with temporal or sequential dependencies. In particular, we show that for certain scenarios, our quantum version of LSTM learns faster or reaches the optimal accuracies faster than its classical counterpart. In addition, with variational quantum circuits as the building blocks, the proposed architecture provides potential applications in the Noisy Intermediate-Scale Quantum (NISQ) devices. |
Thursday, March 18, 2021 4:48PM - 5:00PM Live |
V32.00010: Implementation of quantum machine learning for electronic structure calculations of periodic systems on NISQ devices Shree Hari Sureshbabu, Rongxin Xia, Sabre Kais Recent progress in the development of quantum machine learning algorithms has attracted a lot of attention especially for the purpose of electronic structure calculations. These algorithms demonstrate accurate electronic structure calculations of lattice models, molecular systems, and has also been extended to periodic systems. Among these, a hybrid approach using Restricted Boltzmann Machine (RBM) and a quantum algorithm to optimize the objective function is a promising method due to its efficiency and ease of implementation. However, implementing these algorithms based on the RBM approach on an actual quantum computer requires a modification since only one ancilla qubit is not sufficient. We present the modified approach that can be implemented on Noisy Intermediate-Scale Quantum (NISQ) devices along with the results of implementing this method on IBM-Q for the computation of the electronic structure of graphene. |
Thursday, March 18, 2021 5:00PM - 5:12PM Live |
V32.00011: Analysis of a Quantum Kernel-Based Classifier Using a Tunable Trapped Ion Noisy Simulator Keith Kenemer, Michael Cubeddu, Ian MacCormack, Conor Delaney, Nidhi Aggarwal, Prineha Narang In this work, we develop a tunable trapped-ion noisy simulator to analyze the noise-sensitivity of a relevant quantum machine learning (QML) algorithm with respect to various noise metrics specific to existing and near-term trapped-ion hardware. Investigating the effects of trapped-ion noise on the classification performance of a quantum-enhanced kernel-based classifier is insightful for the future use of these devices for larger-scale machine learning tasks. We explore the noise-sensitivity trade-offs associated with model training in simulated environments with varying amounts of noise. As trapped-ion quantum computers may offer several advantages over superconducting devices in the realm of QML, such as all-to-all connectivity, stable higher energy atomic levels for constructing qudits, and accessible many-qubit entangling gates, it is important that we analyze and explore strategies to mitigate the effects that noise can have on QML algorithms running on these near-term trapped-ion quantum processors. |
Thursday, March 18, 2021 5:12PM - 5:24PM Live |
V32.00012: RL-QAOA: A Reinforcement Learning Approach to Many-Body Ground State Preparation Jiahao Yao, Lin Lin, Marin Bukov We proposed a reinforcement learning (RL) approach to preparing the ground state of many-body quantum systems. This class of method formulates a Markovian decision process for the underlined quantum control problems and utilizes the policy gradient algorithm to find optimal variational parameters. The algorithm focuses mainly on Quantum Approximate Optimization Algorithm (QAOA) and proves efficient in preparing the ground state, especially with the presence of noise. Some variants of the algorithms take a model-based approach, which further improves the sample efficiency of the algorithms; others generalize the QAOA ansatz to a versatile one. This work sheds light on reinforcement-learning-aided quantum control algorithms. |
Thursday, March 18, 2021 5:24PM - 5:36PM Not Participating |
V32.00013: Reinforcement learning for semi-autonomous approximate quantum eigensolver Francisco Albarran-Arriagada, Juan Carlos Retamal, Lucas Lamata, Enrique Solano The characterization of an operator by its eigenvectors and eigenvalues allows us to know its action over any quantum state. Here, we propose a protocol to obtain an approximation of the eigenvectors of an arbitrary Hermitian quantum operator. This protocol is based on measurement and feedback processes, which characterize a reinforcement learning protocol. With this proposal, we can obtain an approximation of the eigenvectors of a random qubit operator with average fidelity over 90% in less than 10 iterations, and surpass 98% in less than 300 iterations. Moreover, for the two-qubit cases, the four eigenvectors are obtained with fidelities above 89% in 8000 iterations for a random operator, and fidelities of 99% for an operator with the Bell states as eigenvectors. |
Thursday, March 18, 2021 5:36PM - 5:48PM On Demand |
V32.00014: Differentiable Quantum Architecture Search Shixin Zhang, Chang-Yu Hsieh, Shengyu Zhang, Hong Yao Quantum architecture search (QAS) is the process of automating architecture engineering of quantum circuits. It has been desired to construct a powerful and general QAS platform which can significantly accelerate current efforts to identify quantum advantages of error-prone and depth-limited quantum circuits in the NISQ era. Hereby, we propose a general framework of differentiable quantum architecture search (DQAS), which enables automated designs of quantum circuits in an end-to-end differentiable fashion. We present several examples of circuit design problems to demonstrate the power of DQAS. For instance, unitary operations are decomposed into quantum gates, noisy circuits are re-designed to improve accuracy, and circuit layouts for quantum approximation optimization algorithm are automatically discovered and upgraded for combinatorial optimization problems. These results not only manifest the vast potential of DQAS being an essential tool for the NISQ application developments, but also present an interesting research topic from the theoretical perspective as it draws inspirations from the newly emerging interdisciplinary paradigms of differentiable programming, probabilistic programming, and quantum programming. |
Thursday, March 18, 2021 5:48PM - 6:00PM On Demand |
V32.00015: Quantum computation on defective circuits Mohammad Ansari Using quantum neural network, I'll show techniques that hels one to take a better use of defective quantum processor, those are sufferring from a few unexpected qubit-qubit disconnections and/or malfunctioning qubits. I present how to update ideal gate operations to get the expected output state from such faulty circuit and present examples performed on real quantum processors as well as simulation machines. |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700