A5070390129- Quantum Computing Algorithms and Architecture
- Neural Networks and Applications
- Fuzzy Logic and Control Systems
- Quantum Information and Cryptography
- Evolutionary Algorithms and Applications
- Neural Networks and Reservoir Computing
- Reinforcement Learning in Robotics
- Quantum and electron transport phenomena
- Metaheuristic Optimization Algorithms Research
- Modular Robots and Swarm Intelligence
- Evolutionary Psychology and Human Behavior
- Quantum many-body systems
- AI-based Problem Solving and Planning
- Face recognition and analysis
- Face Recognition and Perception
- Robot Manipulation and Learning
- Advanced Vision and Imaging
- Logic, Reasoning, and Knowledge
- Fault Detection and Control Systems
- 3D Shape Modeling and Analysis
- Model Reduction and Neural Networks
- Quantum-Dot Cellular Automata
- Robotics and Sensor-Based Localization
- Physics of Superconductivity and Magnetism
- Teleoperation and Haptic Systems
North Carolina State University
2002-2024
University College London
2018-2022
Rahko (United Kingdom)
2018-2021
University of British Columbia
2017
Radboud University Nijmegen
2015-2016
University of Southern California
2007-2010
Southern California University for Professional Studies
2007
University of Strathclyde
1991-2005
Turing Institute
1989-2003
Parametrized quantum circuits initialized with random initial parameter values are characterized by barren plateaus where the gradient becomes exponentially small in number of qubits. In this technical note we theoretically motivate and empirically validate an initialization strategy which can resolve plateau problem for practical applications. The technique involves randomly selecting some values, then choosing remaining so that circuit is a sequence shallow blocks each evaluates to...
Quantum circuits with hierarchical structure have been used to perform binary classification of classical data encoded in a quantum state. We demonstrate that more expressive the same family achieve better accuracy and can be classify highly entangled states, for which there is no known efficient method. compare performance several different parameterizations on two machine learning datasets, Iris MNIST, synthetic dataset states. Finally, we robust noise deploy an classifier ibmqx4 computer.
Quantum systems interacting with an unknown environment are notoriously difficult to model, especially in presence of non-Markovian and non-perturbative effects. Here we introduce a neural network based approach, which has the mathematical simplicity Gorini–Kossakowski–Sudarshan–Lindblad master equation, but is able model effects different regimes. This achieved by using recurrent networks (RNNs) for defining Lindblad operators that can keep track memory Building upon this framework, also...
Studying general quantum many-body systems is one of the major challenges in modern physics because it requires an amount computational resources that scales exponentially with size system.Simulating evolution a state, or even storing its description, rapidly becomes intractable for exact classical algorithms. Recently, machine learning techniques, form restricted Boltzmann machines, have been proposed as way to efficiently represent certain states applications state tomography and ground...
We propose an efficient method for simultaneously optimizing both the structure and parameter values of quantum circuits with only a small computational overhead. Shallow that use optimization perform significantly better than updates alone, making this particularly suitable noisy intermediate-scale computers. demonstrate variational eigensolver finding ground states Lithium Hydride Heisenberg model in simulation, state Hydrogen gas on IBM Melbourne computer.
We investigate fully self-consistent multiscale quantum-classical algorithms on current generation superconducting quantum computers, in a unified approach to tackle the correlated electronic structure of large systems both chemistry and condensed matter physics. In these contexts, strongly region extended system is isolated self-consistently coupled its environment via sampling reduced density matrices. analyze viability devices provide required fidelity objects for robust efficient...
The developments of quantum computing algorithms and experiments for atomic scale simulations have largely focused on chemistry molecules, while their application in condensed matter systems is scarcely explored. Here we present a algorithm to perform dynamical mean field theory (DMFT) calculations currently available computers, demonstrate it two hardware platforms. DMFT required properly describe the large class materials with strongly correlated electrons. computationally challenging part...
Adversarial learning is one of the most successful approaches to modeling high-dimensional probability distributions from data. The quantum computing community has recently begun generalize this idea and look for potential applications. In work, we derive an adversarial algorithm problem approximating unknown pure state. Although could be done on universal computers, formulation enables us execute near-term computers. Two parametrized circuits are optimized in tandem: tries approximate...
Solving for molecular excited states remains one of the key challenges modern quantum chemistry. Traditional methods are constrained by existing computational capabilities, limiting complexity molecules that can be studied or accuracy results obtained. Several computing have been suggested to address this limitation. However, these typically hardware requirements which may not achieved in near term. We propose a variational machine learning based method determine aiming at being as resilient...
Machine learning is seen as a promising application of quantum computation. For near-term noisy intermediate-scale (NISQ) devices, parametrized circuits (PQCs) have been proposed machine models due to their robustness and ease implementation. However, the cost function normally calculated classically from repeated measurement outcomes, such that it no longer encoded in state. This prevents value being directly manipulated by computer. To solve this problem, we give routine embed for into...
Autonomous navigation controllers were developed for fixed wing unmanned aerial vehicle (UAV) applications using incremental evolution with multi-objective genetic programming (GP). We designed four fitness functions derived from flight simulations and used GP to evolve able locate a radar source, navigate the UAV source efficiently on-board sensor measurements, circle closely around emitter. selected realistic parameters inputs aid in transference of evolved physical UAVs. both direct...
It is shown how artificial neural nets can be used to solve a difficult learning problem. The task balance pole that hinged movable cart by applying either left or right force the cart. control process consists of developing pattern formations give required motor drive control. latter implemented with connectionist net Rumelhart semilinear feedforward type. At each instant in time, values training set system's state variables are processed into single which turn applied input layer net....
Abstract— Medical‐grade monochrome monitors typically display 8 bits of data. This study determined if 11‐bit displays could improve observer performance and decrease use window/level. 8‐ from three manufacturers were used at sites. Six radiologists each site viewed 100 DR chest images (half with a pulmonary nodule) on both displays. Decisions, confidence, nodule location, viewing time, window/level recorded. There was no significant difference in ROC Az as function bit depth. The average...
In this paper we show how a self-organizing Kohonen neural network can use hyperellipsoid clustering (HEC) to build maps from actual sonar data. Since the HEC algorithm uses Mahalanobis distance, elongated shapes (typical of data) be learned. The distance metric also gives stochastic measurement data point's association with node. Hence, used topographical and recognize its own cites for self-localization. number nodes regulated in manner by using Kolmogorov-Smirnov (KS) test cluster...
This paper demonstrates a method for tensorizing neural networks based upon an efficient way of approximating scale invariant quantum states, the Multi-scale Entanglement Renormalization Ansatz (MERA). We employ MERA as replacement fully connected layers in convolutional network and test this implementation on CIFAR-10 CIFAR-100 datasets. The proposed outperforms factorization using tensor trains, providing greater compression same level accuracy compression. demonstrate with 14000 times...
An algorithm that learns to partition the state-space for a machine learned control application is presented, and idea of competitive learning, form unsupervised introduced. A theoretical framework partitioning based on neural network learning model T. kohonen's feature maps (1982, 1984) developed. This aimed at BOXES algorithm. The goal was enhance functionality capability by testing strategies. modified did show an improved performance when compared but needs be tested against other known...
Judgments about personality based on facial appearance are strong effectors in social decision making, and known to have impact areas from presidential elections jury decisions. Recent work has shown that it is possible predict perception of memorability, trustworthiness, intelligence other attributes human face images. The most successful these approaches require images expertly annotated with key landmarks. We demonstrate a Convolutional Neural Network (CNN) model able perform the same...
A control surface can be learned and represented by a neural network through the adoption of reinforcement learning scheme. The authors use to learn mapping between dynamic system's state space possible actions. system is incrementally defined, an appropriate action assigned each part from binary vector input. One problem this type partitioning itself, i.e., whether automatically partition into number situations. If so, achieved faster in optimal way. unsupervised algorithm for adaptive...