Parameter Update Research Articles

Research on Federated Learning Data Sharing Scheme Based on燚ifferential燩rivacy

To realize data sharing, and to fully use the data value, breaking the data island between institutions to realize data collaboration has become a new sharing mode. This paper proposed a distributed data security sharing scheme based on C/S communication mode, and constructed a federated learning architecture that uses differential privacy technology to protect training parameters. Clients do not need to share local data, and they only need to upload the trained model parameters to achieve data sharing. In the process of training, a distributed parameter update mechanism is introduced. The server is mainly responsible for issuing training commands and parameters, and aggregating the local model parameters uploaded by the clients. The client mainly uses the stochastic gradient descent algorithm for gradient trimming, updates, and transmits the trained model parameters back to the server after differential processing. To test the performance of the scheme, in the application scenario where many medical institutions jointly train the disease detection system, the model is tested from multiple perspectives by taking medical data as an example. From the testing results, we can know that for this specific test dataset, when the parameters are properly configured, the lowest prediction accuracy rate is 90.261% and the highest accuracy rate is up to 94.352. It shows that the performance of the model is good. The results also show that this scheme realizes data sharing while protecting data privacy, completes accurate prediction of diseases, and has a good effect.

Cloud-Edge Collaboration-Based Local Voltage Control for DGs With Privacy Preservation

The increased distributed generators (DGs) have exacerbated voltage violations in active distribution networks (ADNs). Local reactive power control of DG inverters can realize a fast response to frequent voltage fluctuations. However, commonly used model-based voltage control depends upon accurate network parameters and entire ADN data, which may cause the sensitive information leakage of ADN and DG behaviors in practical operation. In this article, a cloud-edge collaboration-based local voltage control strategy for DGs is proposed with privacy preservation. First, a local voltage control framework is established based on cloud-edge collaboration, in which a surrogate model is built based on the graph convolutional neural networks to estimate the ADN voltages. By transferring the surrogate model, the edge side can obtain the exact voltage estimation in the local curve tuning process without the authority of the whole ADN data, preserving the network parameters of ADN. Then, the interarea coordination based on federated learning is proposed to realize the parameter updating of DG control curves, which can achieve better voltage control performance. By updating surrogate submodels based on private data distributed across multiple edge devices, federated learning can effectively preserve DG behaviors. Finally, the effectiveness and adaptability of the proposed control strategy are validated using the modified IEEE 33-node system. The proposed local DG control strategy can effectively cope with voltage problems and enhance the adaptability to variations in practical operation states while considering privacy preservation.

Construction method of hybrid quantum long-short term memory neural network for image classification

Long-short term memory (LSTM) neural network solves the problems of long-term dependence, gradient disappearance and gradient explosion by introducing memory units, and is widely used in time series analysis and prediction. Combining quantum computing with LSTM neural network will help to improve its computational efficiency and reduce the number of model parameters, thus significantly improving the performance of traditional LSTM neural network. This paper proposes a hybrid quantum LSTM (hybrid quantum long-short term memory, HQLSTM) network model that can be used to realize the image classification. It uses variable quantum circuits to replace the nerve cells in the classical LSTM network to realize the memory function of the quantum network. At the same time, it introduces Choquet integral operator to enhance the degree of aggregation between data. The memory cells in the HQLSTM network are composed of multiple variation quantum circuits (VQC) that can realize different functions. Each VQC consists of three parts: the coding layer, which uses angle coding to reduce the complexity of network model design; the variation layer, which is designed with quantum natural gradient optimization algorithm, so that the gradient descent direction does not target specific parameters, thereby optimizing the parameter update process and improving the generalization and convergence speed of the network model; the measurement layer, which uses the Pauli Z gate to measure, and the expected value of the measurement result is input to the next layer to extract useful information from the quantum circuit. The experimental results on the MNIST, FASHION-MNIST and CIFAR datasets show that the HQLSTM model achieves higher image classification accuracy and lower loss value than the classical LSTM model and quantum LSTM model. At the same time, the network space complexity of HQLSTM and quantum LSTM are significantly reduced compared with the classical LSTM network.

Optimization of Mobile Network Radio Coverage by Automating Radio Parameter Updates Using Parsing

Optimization of Mobile Network Radio Coverage by Automating Radio Parameter Updates Using Parsing

H∞ Control for Oscillator Systems With Event-Triggering Signal Transmission of Internet of Things

This article proposes to design a distributed <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$H_{\infty} $ </tex-math></inline-formula> optimal control algorithm for Van der Pol oscillators with unknown internal dynamics, input constraints and external disturbances, via event-triggering signal transmission of the Internet of Things (IoT). First, the graph theory for the IoT is introduced. Second, the dynamics of Van der Pol oscillators are transformed into the tracking dynamics which cooperate via the IoT network. Third, unlike the existing online optimal control algorithms using adaptive dynamic programming, we design an <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$H_{\infty} $ </tex-math></inline-formula> optimal control algorithm employing an event-triggering signal transmission mechanism to reduce the burden of communication resource and computation bandwidth of the IoT network. As the triggering condition and approximation parameter update policies are appropriately designed, the algorithm guarantees that the Zeno phenomenon is free, the consensus errors are uniformly ultimately bounded, and the external disturbance is compensated. Finally, numerical simulation results with comparison to the time-triggering algorithms confirm the effectiveness of the proposed algorithm.

On the Adaptive Tracking Problem of Time-Varying Systems using the Congelation for Variables Method

This paper discusses the use of the so-called congelation of variables method in the reference tracking problem for a class of nonlinear systems with matched, time-varying, parametric uncertainty. It is shown that the tracking problem using the proposed framework can be decomposed into a zero-regulation problem and a disturbance rejection problem. To reduce the size of the disturbance term, a dynamical system with known input is incorporated into the parameter update law to generate a time-varying nominal parameter. To reject the remaining disturbance term, a sign-like function is exploited, which achieves an adjustable tracking performance in an average sense, while guaranteeing boundedness of all trajectories of the closed-loop system. The theory is illustrated by solving a path-tracking problem and it is shown that the proposed control scheme outperforms the classical scheme in the presence of time-varying parameters.

Efficient Parameterized Precoder Tracking in Slowly Time-Varying MIMO Channels

Precoding in MIMO wireless communication systems is essential for achieving high data rates. However, feeding back precoders places a feedback burden. Exploiting temporal correlation and use of adaptive feedback can significantly reduce the feedback requirement, but even this does not scale well for larger MIMO systems. We propose an approach to partition precoder update parameters, and selectively feed back only those parameters that vary significantly. Specifically, we express precoder variations as zero-diagonal complex matrices whose elements have nearly constant magnitude, thus requiring only phase information. Simulations reveal that this method achieves high rates at 50% lower feedback than prior approaches.

Transient Improvement of Model-Reference Adaptive Control and Parameter Convergence Rate Analysis

The parameter convergence analysis of a modified model-reference adaptive control (M-MRAC) algorithm is presented. The plants considered are continuous with relative degree one. It is verified that the M-MRAC scheme approaches the MRAC scheme when a control design gain increases. Moreover, before this gain reaches high values, the M-MRAC improves the update parameters’ convergence. This analysis is performed by noting a clear connection between the M-MRAC and the conventional MRAC by adding a control output injection term. The methodology developed allows for estimating the parameter convergence rate, and simulation results illustrate the improvement in the transient behavior and the parameter convergence attained with the analyzed adaptive schemes.

Finite element model updating of İskenderpaşa mosque and minaret based on experimental measurements

Historical masonry structures have an important place in cultural heritage. Therefore, maintenance of these structures should be made periodically to prevent natural and manmade hazards. The maintenance of these structures is divided into two parts, namely experimental and numerical applications. Experimental, especially non-destructive methods are crucial for increasing the knowledge level of historical structures. Also, experimental tests provide information data that is used in the modal updating process of numerical models. Thus, simulations that are impossible to make experimentally can be easily carried out with updated numerical models. In this paper, a historical masonry mosque called İskenderpaşa and its minaret is investigated by nondestructive experimental tests. Also, finite element models of both structures are created by ANSYS software. To obtain reliable numerical models, finite element model updating processes are made with the aid of experimental data. At the end of the study, the differences between natural frequencies are reduced from 18.6% to 4.9% for the mosque and from 15.2% to 4.8% for the minaret. According to the study, the modulus of elasticity is the most effective update parameter for both structures.

Complex modified projective synchronization of fractional-order complex-valued dynamic network with time-varying coupling and parameters estimation

The problem of Complex Modified Projective Synchronization (CMPS) for fractional-order complex-valued dynamic networks (FOCDNs) is investigated in this paper. Compared with real-valued dynamic networks, complex-valued dynamic networks have expansive application space. First, coupling weight has been generally considered as static constant in most situations, we consider time-varying coupling weight and the unknown parameters in the FOCDNs. Meanwhile, several parameter update laws are designed to estimate unknown parameters and an adaptive coupling weight is defined to adjust coupling nodes in order to solve CMPS problem of the network. Second, some effective criteria for CMPS of FOCDNs are derived by fractional calculus and Lyapunov stability theory. According to the theoretical analysis, FOCDNs with time-varying delay can achieve CMPS via adaptive controller. Finally, the effectiveness of the modified projective strategy for fractional-order complex networks is illustrated via two numerical examples in the complex-valued space.

Modelling managed forest ecosystems in Sweden: An evaluation from the stand to the regional scale

Incorporation of a forest management module in the dynamic vegetation model LPJ-GUESS has allowed the study and predictions of management treatment effects on the carbon cycle and on forest ecosystem structure. In this study, LPJ-GUESS is evaluated at the regional scale against observational data from the Swedish National Forest Inventory. Simulated standing volume is compared against observations for the four most common forest types in the country. Furthermore, eddy-covariance flux measurements from the Integrated Carbon Observation System (ICOS) are used to evaluate model predictions of net ecosystem exchange (NEE), gross primary productivity (GPP) and ecosystem respiration (Reco) at the site scale. The model results suggest an adequate representation of standing volume in monocultures of Norway spruce and Scots pine for regional simulations in southern and central Sweden, after an updated parameterization of the species. For northern Sweden, the standing volume in Norway spruce monocultures was overestimated with the updated parameter values. At the stand scale, the model produced mixed results for carbon fluxes when evaluated against eddy-covariance data for two sites, one in central and one in southern Sweden. The interannual variation of GPP was well captured for the central Swedish site, but the modelled average GPP for the period 2015–2019 was overestimated by 9%. For the southern Swedish site, GPP was underestimated by 15% for the corresponding period and the simulated interannual variation was half of the observed. The seasonal estimates of modelled net ecosystem exchange (NEE) deviated from observations and the simulated standing volume was underestimated by 25% for both sites. The results highlight further potential to perform species-specific calibration to capture latitudinal gradients in key ecosystem properties, and to incorporate additional characteristics of site quality which could benefit model accuracy at the scale of individual forest stands, both regarding simulated carbon fluxes and forest stand variables.

Recursive Estimation in the Moving Window: Efficient Detection of the Distortions in the Grids with Desired Accuracy

The development of fast convergent and computationally efficient algorithms for monitoring waveform distortions and harmonic emissions will be an important problem in future electrical networks due to the high penetration level of renewable energy systems, smart loads, new types of power electronics, and many others. Estimating the signal quantities in the moving window is the most accurate way of monitoring these distortions. Such estimation is usually associated with significant computational loads, which can be reduced by utilizing the recursion and information matrix properties. Rank two update representation of the information matrix allows the derivation of a new computationally efficient recursive form of the inverse of this matrix and recursive parameter update law. Newton-Schulz and Richardson correction algorithms are introduced in this paper to prevent error propagation and for accuracy maintenance. Extensive comparative analysis is performed on real data for proposed recursive algorithms and the Richardson algorithm with an optimally chosen preconditioner. Recursive algorithms show the best results in estimation with ill-conditioned information matrices.

Analysis of Surrogate-Assisted Information-Geometric Optimization Algorithms

Surrogate functions are often employed to reduce the number of objective function evaluations in a continuous optimization. However, their effects have seldom been investigated theoretically. This paper analyzes the effect of a surrogate function in the information-geometric optimization (IGO) framework, which includes as an algorithm instance a variant of the covariance matrix adaptation evolution strategy—a widely used solver for black-box continuous optimization. We derive a sufficient condition on the surrogate function for the parameter update in the IGO algorithms to point to a descent direction of the objective function expected over the search distribution. The condition is expressed in terms of three measures of correlation between the objective function and the surrogate function. Our result constitutes a partial justification for the use of a surrogate function in IGO algorithms.

Methodology for the Redesign of Compressor Blades Undergoing Nonlinear Structural Interactions: Application to Blade-Tip/Casing Contacts

Abstract Over the past decade, the drive towards more efficient aircraft engines has pushed the boundaries of operating ranges far beyond a linear structural context. Nonlinear interfaces, such as blade-tip/casing contacts, are to be expected in nominal operating conditions. However, current blade design methodologies still rely on empirical structural considerations, often linear, which may lead to costly redesign operations. This work aims at proposing a methodology for the redesign of blades undergoing nonlinear structural interactions. A three-step redesign process is considered: (1) parameterization of an existing blade, (2) update of blade parameters with respect to a surrogate performance criterion, and (3) performance check of the optimized blade. An original two-way parameterization method is proposed to parameterize existing blades and generate models from blade parameters. As a proof-of-concept, the redesign of the NASA compressor blade rotor 37 and fan blade rotor 67 with respect to blade-tip/casing contacts is considered. High-fidelity parameterized models of the initial blades are obtained and their dynamic response to contact interactions are analyzed. Geometries are updated with respect to their clearance consumption, as its minimization has shown beneficial effects on the considered contact interactions. The proposed methodology allows us to better assess the relevance of this performance criterion in the context of blade-tip/casing contacts.

Parameter-efficient feature-based transfer for paraphrase identification

Abstract There are many types of approaches for Paraphrase Identification (PI), an NLP task of determining whether a sentence pair has equivalent semantics. Traditional approaches mainly consist of unsupervised learning and feature engineering, which are computationally inexpensive. However, their task performance is moderate nowadays. To seek a method that can preserve the low computational costs of traditional approaches but yield better task performance, we take an investigation into neural network-based transfer learning approaches. We discover that by improving the usage of parameters efficiently for feature-based transfer, our research goal can be accomplished. Regarding the improvement, we propose a pre-trained task-specific architecture. The fixed parameters of the pre-trained architecture can be shared by multiple classifiers with small additional parameters. As a result, the computational cost left involving parameter update is only generated from classifier-tuning: the features output from the architecture combined with lexical overlap features are fed into a single classifier for tuning. Furthermore, the pre-trained task-specific architecture can be applied to natural language inference and semantic textual similarity tasks as well. Such technical novelty leads to slight consumption of computational and memory resources for each task and is also conducive to power-efficient continual learning. The experimental results show that our proposed method is competitive with adapter-BERT (a parameter-efficient fine-tuning approach) over some tasks while consuming only 16% trainable parameters and saving 69-96% time for parameter update.

Composite adaptive finite‐time fuzzy control for switched nonlinear systems with preassigned performance

SummaryThis article investigates the composite adaptive fuzzy finite‐time prescribed performance control issue of switched nonlinear systems subject to the unknown external disturbance and performance requirement. First, by utilizing the compensation and prediction errors, the piecewise switched composite parameter update law is employed to improve the approximation accuracy of the unknown nonlinearity. Then, the improved fractional‐order filter and error compensation signal are introduced to cope with the influences caused by the explosive calculation and filter error, respectively. Meanwhile, the effect of the compound disturbances consisting of the unknown disturbances and approximation errors is reduced appropriately by designing the piecewise switched nonlinear disturbance observer. Moreover, stability analysis results prove that the proposed preassigned performance control scheme not only ensures that all states of the closed‐loop system are practical finite‐time bounded, but also that the tracking error converges to a preassigned area with a finite time. Ultimately, the simulation examples are given to demonstrate the effectiveness of the proposed control strategy.

A method for the lifetime sensorless estimation of surface and core temperature in lithium-ion batteries via online updating of electrical parameters

Online temperature estimates are essential to the thermal monitoring and control of battery cells for battery management systems (BMSs). Due to hardware limitations, there has been a surge in interest in sensorless approaches for both surface and core temperatures. On this account, several methods have been proposed in the literature for the coestimation of state of charge (SOC) and temperature via RC-based electrical and thermal models and Extended Kalman Filters (EKFs). However, the stability and reliability of these schemes over the complete cell lifetime, when the effects of battery aging become apparent, have not been addressed thoroughly. In this article, a dual state-parameter estimation is carried out on an enhanced equivalent circuit model to coestimate the SOC and SOH on a commercial nickel-rich, silicon–graphite cell throughout its entire lifetime. A thermal model has been characterized based on the previous electrical model for the estimation of surface and core temperature of the cell. The continuous updating and correction of electrical parameters prove to be critical for temperature estimations to remain accurate in the long run, yielding a root mean square error (RMSE) in surface temperature below 1.2 °C for as long as 800 cycles.

An end-to-end deep generative approach with meta-learning optimization for zero-shot object classification

Zero-shot object classification aims to recognize the object of unseen classes whose supervised data are unavailable in the training stage. Recent zero-shot learning (ZSL) methods usually propose to generate new supervised data for unseen classes by designing various deep generative networks. In this paper, we propose an end-to-end deep generative ZSL approach that trains the data generation module and object classification module jointly, rather than separately as in the majority of existing generation-based ZSL methods. Due to the ZSL assumption that unseen data are unavailable in the training stage, the distribution of generated unseen data will shift to the distribution of seen data, and subsequently causes the projection domain shift problem. Therefore, we further design a novel meta-learning optimization model to improve the proposed generation-based ZSL approach, where the parameters initialization and the parameters update algorithm are meta-learned to assist model convergence. We evaluate the proposed approach on five standard ZSL datasets. The average accuracy increased by the proposed jointly training strategy is 2.7% and 23.0% for the standard ZSL task and generalized ZSL task respectively, and the meta-learning optimization further improves the accuracy by 5.0% and 2.1% on two ZSL tasks respectively. Experimental results demonstrate that the proposed approach has significant superiority in various ZSL tasks.

Multi-Motor Cooperative Control Strategy for Speed Synchronous Control of Construction Platform

Barrel-type structures are ubiquitous in industry and life. A novel construction platform of barrel-type structures is introduced, and it involves a multi-motor system. The synchronous control of multiple motors has great importance. To guarantee the high precision, high stability, and fast response for the collaborative control of multiple motors, this paper proposes a multi-motor cooperative control strategy. Firstly, a speed synchronization control structure between multiple permanent magnet synchronous motors (PMSM) is designed by way of the mean coupling control structure. Then, the conventional PID controller is improved by machine learning. Moreover, a radial basis function neural (RBF) network is introduced to the conventional PID algorithm for system identification processing, and the gradient descent algorithm is used for parameter updating. An improved variable speed integral term is introduced into the integral term of the conventional PID algorithm to eliminate the error as soon as possible. Finally, it is verified via numerical simulation experiments that show the multi-motor cooperative control strategy has high anti-interference ability and robustness.

Event-trigger rolling horizon optimization for congestion management considering peer-to-peer energy trading among microgrids

This paper designs an event-trigger rolling horizon optimization framework to alleviate electricity congestion caused by peer-to-peer (P2P) energy transactions among microgrids (MGs) under uncertainty. This framework incorporating abounded time dimensions includes three processes: first, a discrete-time linear programming model for the P2P energy trading among MGs with real-time data updating for reaction to underling uncertainties, are formulated to derive initial schemes in prediction time windows; second, a modified optimal power flow model is used to verify congestion in control time windows, with congestion occurrence defined as an event; third, the power rescheduling of MGs is triggered by this event for reducing periodic calculation burden and executed to generate new trading schemes for congestion alleviation in control time windows. We develop an online distributed optimization algorithm to realize independent decisions of MGs and guarantee rolling updates of system parameters through embedding alternating direction method of multipliers into every time window. Finally, simulations in different cases are implemented to illustrate the reasonability and effectiveness of the proposed optimization framework. Results show that congestion is reasonably managed while ensuring the diversity of P2P energy transactions. Compared with the period-driven rolling horizon optimization framework, the computational time is significantly reduced in the proposed framework.