摘要 :
In this work we present an algorithm to perform algorithmic differentiation in the context of quantum computing. We present two versions of the algorithm, one which is fully quantum and one which employees a classical step (hybrid...
展开
In this work we present an algorithm to perform algorithmic differentiation in the context of quantum computing. We present two versions of the algorithm, one which is fully quantum and one which employees a classical step (hybrid approach). Since the implementation of elementary functions is already possible on quantum computers, the scheme that we propose can be easily applied. Moreover, since some steps (such as the CNOT operator) can (or will be) faster on a quantum computer than on a classical one, our procedure may ultimately demonstrate that quantum algorithmic differentiation has an advantage relative to its classical counterpart.
收起
摘要 :
The key point of introducing quantum genetic algorithm to a quantum backpropagation neural network model is to overcome local stagnation problem which used to be Achilles' heel. In this paper, we propose a new quantum backpropagat...
展开
The key point of introducing quantum genetic algorithm to a quantum backpropagation neural network model is to overcome local stagnation problem which used to be Achilles' heel. In this paper, we propose a new quantum backpropagation (QBP) model based on the quantum genetic algorithm (QGA) and make simulations with this model to see whether QGA can really upgrade QBP and, in addition, to ensure that both quantum neural networks are better than classical backpropagation (CBP) neural networks from many points of view. Numerical experiments have been built to illustrate the efficiency of the new QBP algorithm over CBP and the original QBP algorithm. However, the proposed model has shown superior results to the rest of models in terms of correction rate and training time. That is to say quantum genetic algorithm-based quantum backpropagation neural network converges earlier than the other two models and that's why we can reduce the time needed to train.
收起
摘要 :
Abstract The aim of this study is to present a novel application of Levenberg–Marquardt backpropagation (LMB) to investigate numerically the solution of functional differential equations (FDE) arising in quantum calculus models (...
展开
Abstract The aim of this study is to present a novel application of Levenberg–Marquardt backpropagation (LMB) to investigate numerically the solution of functional differential equations (FDE) arising in quantum calculus models (QCMs). The various types of discrete versions of FDM in QCMs are always found to be stiff to solve due to involvement of delay and to overcome the said difficulty, we proposed intelligent computing platform via LMB networks. In order to generate dataset for LMB networks, firstly, the FDEs in QCMs are converted into recurrence relations, then these recurrence systems are solved numerically on a specific input grids in case of both types of FDEs with q-exponential function as well as stable with decreasing behavior characteristics. The training, testing and validation samples based processes are employed to construct LMB networks by exploiting approximation theory on mean square error sense for obtaining the solutions of both types of FDEs. The exhaustive conducted simulation studies for solving FDEs in QCMs via absolute error and mean squared error endorse the accuracy, potential, convergence, stability and worth of proposed technique, which further certified through viable training state parameters, outcomes of error histograms, values of regression/correlation indices.
收起
摘要 :
An energy-efficient adiabatic learning neuro cell is proposed. The cell can be used for on-chip learning of adiabatic superconducting artificial neural networks. The static and dynamic characteristics of the proposed learning cell...
展开
An energy-efficient adiabatic learning neuro cell is proposed. The cell can be used for on-chip learning of adiabatic superconducting artificial neural networks. The static and dynamic characteristics of the proposed learning cell have been investigated. Optimization of the learning cell parameters was performed within simulations of the multi-layer neural network supervised learning with the resilient propagation method.
收起
摘要 :
Entanglement of a quantum system depends upon the relative phase in complicated ways, which no single measurement can reflect. Because of this, “entanglement witnesses” (measures that estimate entanglement) are necessarily limit...
展开
Entanglement of a quantum system depends upon the relative phase in complicated ways, which no single measurement can reflect. Because of this, “entanglement witnesses” (measures that estimate entanglement) are necessarily limited in applicability and/or utility. We propose here a solution to the problem using quantum neural networks. A quantum system contains the information of its entanglement; thus, if we are clever, we can extract that information efficiently. As proof of concept, we show how this can be done for the case of pure states of a two-qubit system, using an entanglement indicator corrected for the anomalous phase oscillation. Both the entanglement indicator and the phase correction are calculated by the quantum system itself acting as a neural network.
收起
摘要 :
This paper describes how the fault tolerance of the backpropagation algorithm can be used to accommodate the realistic (nonideal) transfer characteristics of the optical communication links used, between neural layers, in optoelec...
展开
This paper describes how the fault tolerance of the backpropagation algorithm can be used to accommodate the realistic (nonideal) transfer characteristics of the optical communication links used, between neural layers, in optoelectronic neural networks. In particular the authors demonstrate that networks, utilizing MSM (metal-semiconductor-metal) photodiodes (PDs) and either LED (light emitting diode) or MQW (multiple quantum well) laser transmitters within these intraneural links, are capable of performing satisfactorily even in the presence of such nonideal device phenomena as: 60% optical crosstalk, 50% optoelectronic device variation, or a thresholded (I/sub th/=0.5*I/sub max/) laser output characteristic. Subsequent to this, the authors then show how it is possible to use this fault tolerance to simplify the neuron architecture, to the extent that it consists only of MSM PDs a current amplifier, and an MQW laser. The overall neuron transfer function is then a first-order approximation to the original sigmoidal function.
收起
摘要 :
In the Present work, Parallel Backpropagation Artificial Networks are employed to optimize and predict the various system parameters of a semiconductor nanodevice. Segregation of the total training range can be obtained by employi...
展开
In the Present work, Parallel Backpropagation Artificial Networks are employed to optimize and predict the various system parameters of a semiconductor nanodevice. Segregation of the total training range can be obtained by employing parallel backpropagation networks operating in different ranges, thereby reducing, significantly, the total training duration. For numerical calculations, electron temperature model has been utilized and relevant scattering mechanisms are incorporated in the calculations. The present model yields a better prediction of system parameters and can be utilized in future Nanodevice fabrication.
收起
摘要 :
This paper presents the systematic characterization of the
molecular beam epitaxy (MBE) process to quantitatively model the effects
of process conditions on film qualities. A five-layer, undoped AlGaAs
and InGaAs single quantum we...
展开
This paper presents the systematic characterization of the
molecular beam epitaxy (MBE) process to quantitatively model the effects
of process conditions on film qualities. A five-layer, undoped AlGaAs
and InGaAs single quantum well structure grown on a GaAs substrate is
designed and fabricated. Six input factors (time and temperature for
oxide removal, substrate temperatures for AlGaAs and InGaAs layer
growth, beam equivalent pressure of the As source and quantum well
interrupt time) are examined by means of a fractional factorial
experiment. Defect density, X-ray diffraction, and photoluminescence are
characterized by a static response model developed by training
back-propagation neural networks. In addition, two novel approaches for
characterized reflection high-energy electron diffraction (RHEED)
signals used in the real-time monitoring of MBE are developed. In the
first technique, principal component analysis is used to reduce the
dimensionality of the RHEED data set, and the reduced RHEED data set is
used to train neural nets to model the process responses. A second
technique uses neural nets to model RHEED intensity signals as time
series, and matches specific RHEED patterns to ambient process
conditions. In each case, the neural process models exhibit good
agreement with experimental results
收起
原文获取