Adaptive Learning Algorithm Research Articles

Adaptive critic learning for approximate optimal event-triggered tracking control of nonlinear systems with prescribed performances

This paper studies the approximate optimal event-triggered tracking control problem for a class of second-order nonlinear systems with prescribed performances. By employing a prescribed performances function, restrictions on tracking errors are removed for prescribed performances including convergence rate, maximum overshoot, and steady-state error. On this basis, a complex optimal tracking control problem can be transformed into an optimal stabilising control problem by constructing an augmented system. Subsequently, an event-triggered mechanism is introduced to reduce communication resources. Then, an adaptive critic learning algorithm is used to solve the Hamilton-Jacobi-Bellman equation, where the weights in the critic networks are tuned through the gradient descent approach and experience replay technology. Through the developed learning method, the persistence of excitation condition is released, and the data efficiency is improved. The Lyapunov stability theory is introduced to verify the stability of the close-loop system and the tracking errors are uniformly ultimately bounded. Finally, simulation results are presented to show the effectiveness of the proposed approach.

GAL: Graph-Induced Adaptive Learning for Weakly Supervised 3D Object Detection

Weakly Supervised 3D Object Detection (WS3DOD) aims to perform 3D object detection with little reliance on 3D labels, which greatly reduces the cost of 3D annotations. In recent literature, the pseudo-label-based approach brings impressive performance, which generates 3D pseudo-labels from 2D bounding boxes. Despite their success, two key issues remain unresolved that reduce the quality of 3D pseudo-labels: 1) the existing local object locating algorithm can not capture complete clusters of points globally, and 2) the existing algorithm can not capture sparse points caused by the unevenly distributed points obtained by LiDAR cameras. Hence, we propose GAL, a Graph-induced Adaptive Learning algorithm, to generate 3D pseudo-labels. First, we propose the Cluster Locating algorithm based on the Minimum Spanning Tree (MST) to globally locate the objects, which can leverage the characteristic that points inside an object are compact while points between objects are discrete. Second, we propose a density-guided adaptive learning algorithm to optimise the Cluster Locating algorithm, named Cuboid Drift. Cuboid Drift considers the inhomogeneous distribution of reflected points on different reflective surfaces of LiDAR imaging. Finally, 3D pseudo-labels generated by GAL are leveraged to train 3D detectors. Extensive experiments on the challenging <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">KITTI </i> and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">DAIR-V2X-V</i> dataset demonstrate that GAL without 3D labels can be comparable with strongly supervised approaches and outperforms the previous state-of-the-art WS3DOD methods. Moreover, our method saves 88% of the time spent on pseudo-label generation.

Examination of the performance of ASEL and MPViT algorithms for classifying mangrove species of multiple natural reserves of Beibu Gulf, south China

Mangrove is the highest productive ecosystems in the global coastal zone, which has high blue carbon sink function and carbon neutrality potential. Fine species classification is essential for mangrove conservation and sustainable development, and has attracted much attention in recent years using ensemble learning and multi-dimensional data. However, the current mangrove species classification based on traditional stacking ensemble learning still faces challenges due to the correlation between base classifiers, differences in meta-classifier capabilities, the subjectivity of parameter tuning, and data redundancy. To address these issues, this paper utilized unmanned aerial vehicle (UAV) multispectral images of three mangrove nature reserves in Beibu Gulf, south China, to examine the classification and generalization ability of our proposed Adaptive Stacking Ensemble Learning (ASEL) algorithm for different mangrove species. We also aim to verify the feasibility of the Multi-Path Vision Transformer for Dense Prediction (MPViT) algorithm for mangrove species mapping, and compare its performance with the ASEL algorithm for mangrove species classification. Finally, we used the SHapley Additive Explanations (SHAP) method to measure the contribution of feature variables to the model, exploring the sensitivity of different image features to mangrove species mapping. This study highlights that: (1) The two ASEL algorithms achieved high accuracy classification of mangrove species with the overall classification accuracy ranging from 79.8% to 96.2%. The ACE-Stacking and AOM-Stacking algorithms performed better classification ability than the traditional stacking algorithm, with the mean overall classification accuracy increasing from 0.9% to 3.3%. The McNemar test further indicated that the differences in classification results derived from three algorithms were significant at the 95% confidence interval, demonstrating the better classification and generalization ability of the ASEL algorithm. (2) The MPViT algorithm achieved better classification accuracy in the three reserves, with an overall accuracy of 95.5%-97.3%, which was 0.6%-5.4% higher than the two ASEL algorithms. The average accuracy of identifying mangrove species was over 95.3% in all reserves, demonstrating the desirable mangrove classification performance of MPViT algorithm. (3) The identification accuracies (F1 scores) of mangrove species in different reserves were ranged from 0.848 to 0.984. Cyperus malaccensi had the highest identification accuracy. (4) The SHAP method interpreted the great contribution of digital surface model and variable atmospherically resistant index features on mangrove species classification. The red band and ratio vegetation index was sensitive to Aegiceras corniculatum and to Cyperus malaccensi.

Effective LSTMs with seasonal-trend decomposition and adaptive learning and niching-based backtracking search algorithm for time series forecasting

Long short-term memory faces challenges in information mining and parameter selection due to inherent uncertainty and randomness. In this study, we propose a novel hybrid model called ‘STL-ALN_BSA-LSTM’ that aims to tackle this issue. In the proposed model, data is first prepossessed using seasonal-trend decomposition to extract informative features. Next, we propose an adaptive learning and niching-based backtracking search algorithm to enhance the robustness and learning ability of the original backtracking search algorithm, which affords a balance between exploration and exploitation. Subsequently, the weights and thresholds within each LSTM are optimized by our optimization algorithm automatically, leading to improved interpretability and accuracy of the model. Finally, the proposed optimization algorithm was evaluated using five benchmark functions, and the hybrid model was compared with sixteen baselines on two real-world datasets. For rainfall forecasting, the hybrid model achieved the root mean square error of 7.3772, mean absolute error of 2.6187, and R2 of 0.8388. For power generation forecasting, our model outperformed the second-best model by 29.6%, 34.4%, and 16 ‰, respectively, as in the above metrics. These results demonstrate that the proposed hybrid model can attain higher accuracy in real-life applications than previous models, which contributes in-depth insights into parameter optimization for time series forecasting, and provides valuable suggestions in resource planning, risk reduction, and decision-making. The core code of this work can be found from https://github.com/zjuml/STL-ALNBSA-LSTM.

A reduced order model based on adaptive proper orthogonal decomposition incorporated with modal coefficient learning for digital twin in process industry

The digital twin (DT) technology provides a viable and promising direction for improving the level of the production status monitoring and the overall product quality in various fields. However, the accuracy of working condition identification, the timeliness of process adjustment, and the stability of product quality are put forward higher requirements in the process industry, which is characterized by nonlinear, large-scale, and dynamic complex systems. Therefore, it still remains a tricky challenge to construct and maintain an effective and accurate DT model in the process industry. A reduced order model (ROM) with the adaptive updating ability is proposed. The adaptive proper orthogonal decomposition (APOD) is adopted to achieve the continuous iteration and the adaptive optimization of the reduced basis set. Correspondingly, an adaptive learning algorithm based on the least squares support vector regression (LS-SVR) is developed to quickly obtain the modal coefficients and effectively circumvent the prohibitively high computational cost. In this way, the physical field of interest is expressed in a low-dimensional approximation with a high accuracy. The effectiveness of the method is verified by a case study in the process industry. Results show that the proposed model displays a high-precision fitting and a significant time saving for the full order model (FOM).

Strengthening School Management with Digital Education Technology to Improve the Quality of Educational Output

Digital educational technologies enhance teaching methodologies through multimedia resources, interactive simulations and real-time assessment tools. Teachers can leverage adaptive learning algorithms to adapt learning to the individual needs of students, driving engagement, understanding, and academic success. Moreover, data-driven decision-making is facilitated by digital educational technologies. Educators gain valuable insights into student performance through sophisticated analytics, enabling them to identify areas for improvement and implement targeted interventions. This approach supports early identification of at-risk students and ensures that a comprehensive support system is in place. In addition, digital education technologies promote collaboration and communication between various stakeholders, including teachers, students, parents and administrators. This enables real-time feedback, collaboration between parents and teachers, and remote learning opportunities, creating a participatory and dynamic educational environment. In conclusion, embracing digital education technology has great potential to strengthen school management and meet the demands of 21st century quality education. By automating administrative tasks, improving teaching methodologies, enabling data-driven decision-making and promoting collaboration, educational institutions can adapt to evolving challenges and deliver high-quality educational outcomes. Embracing these technological advances is important for preparing students for the demands of the digital age and ensuring a successful future.

The Most Interesting Anomalies Discovered in ZTF DR17 from the SNAD-VI Workshop

The SNAD team has developed an adaptive learning algorithm, named Pine Forest (PF), to enhance anomaly detection in astronomical data. Recognizing the essential role of human engagement in the discovery process, PF presents outliers to a human expert for review, and filters out trees which disagree with the feedback provided. During the sixth annual SNAD workshop (https://snad.space/2023/), held in 2023 July, we applied PF to the Zwicky Transient Facility’s DR17 data. Interesting discoveries include long-duration objects such as supernovae, along with fast transients like red dwarf flares and one microlensing event. As a result, new variable stars were identified and labeled in the SNAD knowledge database.

A reference free non-negative adaptive learning system for health care monitoring and adaptive physiological artifact elimination in brain waves

Electroencephalogram (EEG), also referred to as brain wave (BW), is a physiological phenomenon that depicts how the human brain functions. Brain wave analysis is fundamental in applications like brain-computer interference (BCI), beamforming, sleep analysis, epilepsy detection, and emotion recognition. In real-time applications, the brain wave encounters many physiological and non-physiological artifacts during acquisition. Due to this phenomenon, the analysis method is complicated and obscures the brain wave’s tiny features. This study proposes an intelligent signal enhancement unit (SEU) for processing EEG signals to enable decision-making under certainty. The proposed SEU enables healthcare professionals to analyze high-resolution brain wave components for various applications. A new singular spectrum decomposition (SSD) based on score reconstruction (score RC) is used in the first phase of the SEU to generate the artifact nature, which is then used as a reference signal in the adaptive artifact cancellation (AAC) method. The SSD performs the embedding, decomposition, grouping, and reconstruction procedures to provide the reference signal. A modified Logarithmic Non-Negative Adaptive Learning Algorithm (MLNNAL) is employed in the second stage of the SEU to improve the EEG signal. With the help of this proposed adaptive learning, a system with lower computing complexity which is stable and has non-negative weights can be realized. The adaptive learning algorithm’s weight recursion continually reweights the weight vector for each iteration to eliminate artifacts from contaminated brain waves. Excess mean square error (EMSE), signal to noise ratio improvement (SNRI) and computational cost of the adaptive learning algorithm are used to evaluate how the proposed SEU performs.

Second and Foreign Language Vocabulary Learning through Digital Reading: A Meta-Analysis

Digital reading facilitates L2 development by allowing anytime-anywhere learning with various digital resources. Although there has been increasing research exploring the role of digital reading on L2 vocabulary learning, synthesized evidence regarding the overall facilitating power of digital reading is still lacking. This meta-analysis aggregates 21 important empirical studies published within the last 20 years and indexed in scholarly recognized databases, so as to provide a comprehensive panoramic assessment of how effectively digital reading has enhanced second and foreign language vocabulary acquisition with diversified learner backgrounds and learning environments. A total of 77 effect sizes were yielded across different studies, and random-effect modeling was employed for analyzing the study-level heterogeneity and sub-group variability. Results demonstrate that digital reading had an overall significant effect (dimmediate = 1.45, p < .01; ddeleayed = 2.98, p < .01) on L2 vocabulary learning of between-subject studies. For within-subject studies, digital reading was found to have an upper-medium (d = 1.39, p < .01) and medium (d = .86, p < .01) effect on immediate and delayed L2 vocabulary post-tests. Moderating factor analysis results show that L2 proficiency, vocabulary test formats, and digital resources could robustly explain the variance of effect sizes. The findings have strong pedagogical implications on the effective design for digital reading tasks, including the development of adaptive learning algorithms and personalized lexical glosses. Recommendations for future research in the field are provided by pinpointing where to improve in terms of experimental design and the focus of the learner group.

Hyperbolic function based hybrid consistent adaptive sequential DL algorithms for PIR sensor based intruder detection

This study demonstrates that a novel dictionary learning (DL) algorithm-based approach can be successfully used in conjunction with an indigenously developed hardware module, using four pyroelectric infrared (PIR) sensors, to enhance the performance of an intruder detection system. In this work, initially, a hyperbolic function based Consistent Adaptive Sequential Dictionary Learning (CAS-DL) algorithm has been proposed (HCAS-DL). Then another dictionary learning algorithm, called label consistency based K-SVD (LCK-SVD) is considered, where the label information of each dictionary atom has been introduced to enforce discriminability in sparse code. The objective function in LCK-SVD is formuated simultaneously considering the concept of label consistency (LC) constraints, reconstruction errors, and classification errors. Next, the newly proposed HCAS-DL is hybridized with LCK-SVD to form a novel version of hybrid consistent adaptive sequential dictionary learning approach, named here as the LC-HCAS-DL algorithm. Extensive experiments have been performed to establish the suitability of our proposed approach for the problem under consideration. The performance evaluation clearly establishes that the LC-HCAS-DL algorithm can achieve superior performance compared to recently proposed LC-MCAS-DL algorithm and other state-of-the-art algorithms for the intruder detection problem under consideration.

Adaptive First-Crossing Approach for Life-Cycle Reliability Analysis

Abstract Life-cycle reliability analysis can effectively estimate and present the changes in the state of safety for structures under dynamic uncertainties during their lifecycle. The first-crossing approach is an efficient way to evaluate time-variant reliability-based on the probabilistic characteristics of the first-crossing time point (FCTP). However, the FCTP model has a number of critical challenges, such as computational accuracy. This paper proposes an adaptive first-crossing approach for the time-varying reliability of structures over their whole lifecycle, which can provide a tool for cycle-life reliability analysis and design. The response surface of FCTP regarding input variables is first estimated by performing support vector regression. Furthermore, the adaptive learning algorithm for training support vector regression is developed by integrating the uniform design and the central moments of the surrogate model. Then, the convergence condition, which combines the raw moments and entropy of the first-crossing probability distribution function (PDF), is constructed to build the optimal first-crossing surrogate model. Finally, the first-crossing PDF is solved using the adaptive kernel density estimation to obtain the time-variant reliability trend during the whole lifecycle. Examples are demonstrated to specify the proposed method in applications.

Time-series prediction using a regularized self-organizing long short-term memory neural network

Industrial process data are naturally in the form of complex time-series with high nonlinearities and dynamics. Long short-term memory (LSTM) networks are suitable for developing prediction models to handle nonlinear and dynamic process. However, LSTM neural networks have typically large and predefined structures, which may result in overfitting, and an optimal hidden neurons for a given problem cannot be automatically obtained. For this reason, a regularized self-organizing LSTM (RSO-LSTM) is proposed to optimize both the structure and the parameters of the network. First, an adaptive learning algorithm based on l2-norm regularization is introduced for parameter adjustment. Thereafter, both the prediction accuracy and weight dispersion are considered to avoid overfitting. Second, a growing strategy is designed based on hidden neuronal sensitivity. The structure of the LSTM can then be determined automatically with improved compactness. Finally, a convergence analysis is performed to ensure the feasibility of RSO-LSTM. To demonstrate the merits of the proposed RSO-LSTM for time-series​ prediction, its results for three benchmark experiments and real industrial data of a municipal solid waste incineration process were examined and compared with those of other methods. The results indicated the superiority and potential of RSO-LSTM for industrial applications.

Integration of sEMG-Based Learning and Adaptive Fuzzy Sliding Mode Control for an Exoskeleton Assist-as-Needed Support System

This paper presents an adaptive Fuzzy Sliding Mode Control approach for an Assist-as-Needed (AAN) strategy to achieve effective human–exoskeleton synergy. The proposed strategy employs an adaptive instance-based learning algorithm to estimate muscle effort, based on surface Electromyography (sEMG) signals. To determine and control the inverse dynamics of a highly nonlinear 4-degrees-of-freedom exoskeleton designed for upper-limb therapeutic exercises, a modified Recursive Newton-Euler Algorithm (RNEA) with Sliding Mode Control (SMC) was used. The exoskeleton position error and raw sEMG signal from the bicep’s brachii muscle were used as inputs for a fuzzy inference system to produce an output to adjust the sliding mode control law parameters. The proposed robust control law was simulated using MATLAB-Simulink, and the results showed that it could instantly adjust the necessary support, based on the combined motion of the human–exoskeleton system’s muscle engagement, while keeping the state trajectory errors and input torque bounded within ±5×10−2 rads and ±5 N.m, respectively.

Measuring direction of arrival in cognitive radio systems using variable step adaptive learning for spectrum sensing

In wireless communications cognitive radios are rapidly growing technology and it is used in medical telemetry applications to provide treatment to remote located patients on time. Generally, interferences are occurred due to a shortage of frequency range and spectrums. To avoid these interferences new technologies are developed with cognitive radios. In the received signal there are some noisy signals are occurred due to signal estimation is not done properly. For estimating input signal primary user and secondary user signals direction of arrival is taken into consideration. To improve further output and to remove noise signals of received signal proposed an adaptive learning methodology. In this paper a regularization based variable step adaptive learning algorithm (VSALA) is developed for avoiding minimum disturbance constraint functions in cost function. Further regularization factor and step size combination results a faster convergence when compared to conventional normalized adaptive algorithm (NALA). The proposed learning methodology is analysed for different threshold values like 0.1, 0.5 and 1.0 for estimating the direction of arrival (DOA) and their corresponding beam forming patterns are shown.

An Adaptive Deep Learning Neural Network Model to Enhance Machine-Learning-Based Classifiers for Intrusion Detection in Smart Grids

Modern smart grids are built based on top of advanced computing and networking technologies, where condition monitoring relies on secure cyberphysical connectivity. Over the network infrastructure, transported data containing confidential information, must be protected as smart grids are vulnerable and subject to various cyberattacks. Various machine learning based classifiers were proposed for intrusion detection in smart grids. However, each of them has respective advantage and disadvantages. Aiming to improve the performance of existing machine learning based classifiers, this paper proposes an adaptive deep learning algorithm with a data pre-processing module, a neural network pre-training module and a classifier module, which work together classify intrusion data types using their high-dimensional data features. The proposed Adaptive Deep Learning (ADL) algorithm obtains the number of layers and the number of neurons per layer by determining the characteristic dimension of the network traffic. With transfer learning, the proposed ADL algorithm can extract the original data dimensions and obtain new abstract features. By combining deep learning models with traditional machine learning-based classification models, the performance of classification of network traffic data is significantly improved. By using the Network Security Laboratory-Knowledge Discovery in Databases (NSL-KDD) dataset, experimental results show that the proposed ADL algorithm improves the effectiveness of existing intrusion detection methods and reduces the training time, indicating a promising candidate to enhance network security in smart grids.

Adaptive transfer learning for multimode process monitoring and unsupervised anomaly detection in steam turbines

Through condition-based maintenance strategy, engineers can monitor the health states of equipment and take actions based on the sensor data. Limited by the low failure frequency and high monitoring costs, it is difficult to obtain sufficient historical data of all fault types for condition monitoring (CM). In the steam turbine operation, environmental factors, varying power consumption and manual adjustments can lead to a multimode process, which consists of multiple normal and abnormal conditions. This paper proposes a framework for online unsupervised CM and anomaly detection, not relying on expert knowledge or labeled historical data. Since there are often few monitoring data at the beginning of a new incoming operating mode, an adaptive self-transfer learning algorithm based on Gaussian processes is developed to model the monitoring data with uncertainty information, and to capture the cross-correlations between the different normal modes. A two-hierarchical identification criterion based on the predicted posterior intervals is introduced to first identify the change-points in the observations, and second to decide whether it is an anomaly or a transition between normal modes. The proposed framework is tested on a real steam turbine. The results illustrate its high effectiveness.

Self-Organizing Interval Type-2 Fuzzy Neural Network With Adaptive Discriminative Strategy

Covariate shift is a critical issue of interval type-2 fuzzy neural networks (IT2FNNs) due to the distribution discrepancy between training and testing samples. In this situation, IT2FNNs usually struggle to identify potential features from samples with explicit inductive biases. To address this problem, a self-organizing IT2FNN with an adaptive discriminative strategy (ADS-SOIT2FNN) is developed to maintain the identification performance in the presence of covariate shift. First, a granularity-based metric (GM), using higher-order statistics of local samples, is designed to distinguish the distribution discrepancy caused by covariate shift. The multiple kernels incorporated into GM are able to cover the sample features of the whole Hilbert space. Second, a self-organizing strategy, associated with GM-based discriminative information, is presented to alleviate the structural bias by growing and pruning fuzzy rules. Then, a compact structure of ADS-SOIT2FNN is achieved to adapt to the covariate shift of samples and further strengthen its inductive ability. Third, an adaptive risk mitigation learning algorithm (RMLA) is introduced to update the parameters of ADS-SOIT2FNN. RMLA can regulate the derivatives of parameters with arbitrary distribution samples, which is beneficial for maintaining the global accuracy by relieving the risk of parameter biases. Finally, the effectiveness of ADS-SOIT2FNN is verified by some experiments for identifying nonlinear systems with covariate shift.

Research on RBF neural network adaptive control of three-point contactless measuring device for CNC roller grinder

Abstract In order to improve the CNC roll grinder production efficiency, a novel three-point non-contact measurement device is proposed. An adaptive back-stepping control system that combines with a radial basis function neural network (RBFNN) was designed to control the measurement device working position and also to track the periodic reference movement trajectory. In the proposed control system, the RBFNN approximation and tracking ability are used to identify the unknown measuring device dynamic information. Then, the Lyapunov stability theorem is used to derive the adaptive online learning algorithm. All control algorithms are placed in the control chip based on the TMS320F28335 archive. The simulation results show that the proposed control system has a good control effect when applied to the three-point non-contact measuring device in CNC roll grinders.

Dynamic Event-Triggered Reinforcement Learning-Based Consensus Tracking of Nonlinear Multi-Agent Systems

In this paper, we present a novel approach to address the event-triggered optimized consensus tracking control problem in a class of uncertain nonlinear multi-agent systems (MASs). To optimize control performance, we employ an adaptive reinforcement learning (RL) algorithm based on the actor-critic architecture and utilize the backstepping method. The proposed RL-based optimized controller employs a novel event-triggered strategy, dynamically adjusting sampling errors online to reduce communication resource usage and computational complexity through the intermittent transmission of state signals. We establish the boundedness of all signals in the closed-loop MAS through stability analysis using the Lyapunov method, and demonstrate the prevention of Zeno behavior. Numerical simulations of a practical multi-electromechanical system are provided to validate the effectiveness of the proposed scheme.

Classification of tracheal stenosis with asymmetric misclassification errors from EMG signals using an adaptive cost-sensitive learning method

Upper airway obstruction is characterized by loss of normal airway architecture resulting from various disorders such as infections and asthma. Early detection of airway obstruction is essential to prevent medical deterioration. The objective of this study was to non-invasively identify early-stage tracheal stenosis, using electromyography (EMG) signals of inspiratory muscles. The identification of tracheal stenosis has been defined as an asymmetric misclassification cost problem. Specifically, the EMG signals are used as input to a ResNet-like architecture for tabular data with an Adaptive Cost-Sensitive Learning (AdaCSL) algorithm. The electrical activity of the external intercostal muscle of four healthy individuals was recorded while they breathed through two different tubes, one simulating a narrowed airway and the other simulating a normal airway. Two experiment settings were designed. The first setting aimed to classify tracheal stenosis in a specific subject by training the model on data from other subjects, reflecting the case of diagnosing a new subject. To overcome multi-subject variations, the second setting aimed to classify tracheal stenosis by mixing all subjects’ training and test data. The ResNet-like architecture with an AdaCSL algorithm was significantly better in the first experiment setting with costs that were 43%, 48%, and 59% lower than the cost of the second-best alternative for three different misclassification cost values. It also achieved a lower cost in the second experiment setting over other classifiers. The experiments emphasize the capability of using inspiratory muscle EMG signals to diagnose respiratory disease and demonstrate the usefulness of the AdaCSL algorithm for personalized monitoring.