ABSTRACT As modelling of nonlinear oscillator systems plays an important part in science and engineering fields, a double loop Radial Basis Function Neural Network (RBFNN) with adaptive radius and lattices is proposed for handling this issue. In this design, a large enough lattice arranged to cover all of the trajectories is taken as the mapping center of the RBFNN at the initial condition. The number of lattices will be dynamically adjusted, and those lattices far from the trajectories will be removed. Applying Taylor expansion in local space, the activated radius factor can be separated from the Gaussian function. In order to guarantee that the modelling scheme has the characteristic of fast convergence, the error power function is utilized to minimize the gain parameter of the error differential equation. In the double loop structure, the updated equation of weights and activated radius can be determined by the Lyapunov function, which can guarantee that the weights and the activated radius will converge to the neighborhood of their true value and the tracking error of state trajectories will converge to the neighborhood of zero. In order to show the effectiveness and superiority of the double loop RBFNN proposed in this paper, Helmholtz–Duffing and Vanderpol–Duffing are used as the testing objects of the nonlinear oscillator system while comparing with Deterministic Learning.

Language Models (LMs) have achieved impressive success in various question answering (QA) tasks but have shown limited performance on structured reasoning. Recent research suggests that Knowledge Graph (KG) can augment text data by providing a structured background to enhance reasoning capabilities of LMs. Therefore, how to integrate and reason over KG representations and language context remains an open question. In this work, we propose MJR, a novel model to integrate encoded representations of LMs and graph neural network through multiple layers of feature interaction operations. Subsequently, the fused feature representations in two modalities are fed into a multi-head representation fusion module to comprehensively capture semantic and graph structure information, thereby enhancing language understanding and reasoning capabilities. In addition, we investigate the performance and applicability of different types of large language models as text encoder in the question-answering task. We evaluate our model on three common dataset: CommonsenseQA, OpenBookQA, and MedQA-USMLE datasets. The results demonstrate the advancements of MJR over existing LMs, LM+KG and LLMs models in reasoning for question answering.

3-D shape reconstruction is essential in the navigation of minimally invasive and auto robot-guided surgeries whose operating environments are indirect and narrow, and there have been some works that focused on reconstructing the 3-D shape of the surgical organ through limited 2-D information available. However, the lack and incompleteness of such information caused by intraoperative emergencies (such as bleeding) and risk control conditions have not been considered. In this article, a novel hierarchical shape-perception network (HSPN) is proposed to reconstruct the 3-D point clouds (PCs) of specific brains from one single incomplete image with low latency. A branching predictor and several hierarchical attention pipelines are constructed to generate PCs that accurately describe the incomplete images and then complete these PCs with high quality. Meanwhile, attention gate blocks (AGBs) are designed to efficiently aggregate geometric local features of incomplete PCs transmitted by hierarchical attention pipelines and internal features of reconstructing PCs. With the proposed HSPN, 3-D shape perception and completion can be achieved spontaneously. Comprehensive results measured by Chamfer distance (CD) and PC-to-PC error demonstrate that the performance of the proposed HSPN outperforms other competitive methods in terms of qualitative displays, quantitative experiment, and classification evaluation.

Global internet usage is growing dramatically in recent years, making it difficult for operators to allocate resources rationally and maintain network security. If the network traffic can be accurately and timely forecasted, it can help operators relieve this pressure. However, due to the non-linearity and non-stationarity of network traffic, it's difficult to capture dynamic characteristics to obtain excellent prediction results for large number of traditional methods. To address this issue, we propose a novel long short-term memory (LSTM) neural network combined with variational mode decomposition (VMD) for network traffic forecasting. It firstly applies VMD to decompose the time series according to frequency information, extracting nonlinear and non-stationary features in the sequence, and then LSTM exploited to capture the long-term dependencies of network traffic data. We combine the decomposition algorithm with LSTM to establish a mapping between historical network traffic data and future ones. Experimental results based on real-world datasets demonstrate that the prediction accuracy of our model overperforms the existing state-of-the-art algorithms,

Credit transactions are vital financial activities that yield substantial economic benefits. To further improve lending decisions, stakeholders require accurate and interpretable credit scoring methods. While the majority of previous studies have focused on the relationship between individual features and credit risk, only a few have investigated cross-features. Notably, cross-features can not only represent structured data effectively but also provide richer semantic information than individual features. Nevertheless, most previous methods for learning cross-feature effects from credit data have been implicit and unexplainable. This paper proposes a new credit scoring model based on contrastive augmentation and tree-enhanced embedding mechanisms, termed CATE. The proposed model automatically constructs explainable cross-features by using tree-based models to learn decision rules from the data. Moreover, the importance of each local cross-feature is then derived through an attention mechanism . Finally, the credit score of a user is evaluated using embedding vectors. Experimental results on 4 public datasets demonstrated the interpretability of our proposed method and outperformed 13 state-of-the-art benchmark methods in terms of performance.

Image captioning is still a challenging task aiming at describing the contents of image by words. Current image caption methods usually assume the object relation to be important if the semantic and spatial geometric relationships between objects are close and large, but the relations meeting this assumption are not necessarily important to describe the contents of image in a fine-grained way. That is, the importance of fine-grained object relations is not properly taken into account. Besides, current Transformer based image caption models also fail to consider the importance of fine-grained objects, since they generate all the words of a sentence at one time, which cannot Figure out which objects are more important and vice versa. In this paper, we propose a novel Fine-grained Adaptive Object Transformer (FineFormer) network, which can jointly discover the importance of fine-grained objects and object relations for image captioning. Specifically, a new concept of adaptive soft-foreground attention is proposed to highlight the fine-grained objects dominating the descriptive contents. To characterize and calculate the important relations between fine-grained objects, we also propose an adaptive object relation attention to refine the object relation from the generation process of relation. As such, FineFormer can describe the contents of image more accurately, by reducing the interference of unimportant objects in the background. Extensive experiments on the highly-competitive MS-COCO dataset demonstrated the superiority of our FineFormer.

In this paper, robust hybrid beamforming designs are developed for multi-user millimeter-wave multiple-input multiple-output relay systems in the presence of correlated channel state information errors. Analog beamforming matrices are designed to maximize the equivalent channel gains. Baseband beamforming matrices are solved by optimizing the upper bound of the averaged achievable sum rate. Simulation results show that our algorithm achieves better performance compared with previous existing hybrid beamforming designs.

Group recommendation has recently drawn a lot of attention to the recommender system community. Currently, several deep learning-based approaches are leveraged to learn preferences of groups for items and predict next items in which groups may be interested. Yet, their recommendation performance is still unsatisfactory due to sparse group–item interactions. To address this challenge, this study presents a novel model, called group recommendation model with two-stage deep learning (GRMTDL), which encompasses two sequential stages: 1) group representation learning (GRL) and 2) group preference learning (GPL). In GRL, we first construct an undirected tripartite graph over group–user–item interactions, and then employ it to accurately learn group semantic features through a spatial-based variational graph autoencoder network. While in GPL, we first introduce a dual PL-network that contains two structure-sharing subnetworks: 1) group PL-network employed for GPL and 2) user PL-network utilized for user preference learning. Then, we design a novel layered transfer learning (LTL) method to learn group preferences by alternately optimizing these two subnetworks. In particular, it can effectively absorb knowledge of user preferences into the process of GPL. Furthermore, extensive experiments on four real-world datasets demonstrate that the proposed GRMTDL model outperforms the state-of-the-art baselines for group recommendation.

The rapid geographic spread of COVID-19, to which various factors may have contributed, has caused a global health crisis. Recently, the analysis and forecast of the COVID-19 pandemic have attracted worldwide attention. In this work, a large COVID-19 data set consisting of COVID-19 pandemic, COVID-19 testing capacity, economic level, demographic information, and geographic location data in 184 countries and 1241 areas from December 18, 2019, to September 30, 2020, were developed from public reports released by national health authorities and bureau of statistics. We proposed a machine learning model for COVID-19 prediction based on the broad learning system (BLS). Here, we leveraged random forest (RF) to screen out the key features. Then, we combine the bagging strategy and BLS to develop a random-forest-bagging BLS (RF-Bagging-BLS) approach to forecast the trend of the COVID-19 pandemic. In addition, we compared the forecasting results with linear regression (LR) model, [Formula: see text]-nearest neighbors (KNN), decision tree (DT), adaptive boosting (Ada), RF, gradient boosting DT (GBDT), support vector regression (SVR), extra trees (ETs) regressor, CatBoost (CAT), LightGBM (LGB), XGBoost (XGB), and BLS.The RF-Bagging BLS model showed better forecasting performance in terms of relative mean-square error (RMSE), coefficient of determination ([Formula: see text]), adjusted coefficient of determination ([Formula: see text]), median absolute error (MAD), and mean absolute percentage error (MAPE) than other models. Hence, the proposed model demonstrates superior predictive power over other benchmark models.

Not identified as being exposed or infected, the group of asymptomatic and presymptomatic patients has become the key source of infectious hosts for the COVID-19 pandemic, triggering the re-emergence of outbreaks. Acknowledging the impacts of movement of unidentified patients and the limited testing capacity on understanding the spread of the virus, an augmented Susceptible-Exposed-Infectious-Confirmed-Recovered (SEICR) model integrating intercity migration data and testing capacity is developed to probe into the number of unidentified COVID-19 infected patients. This model allows evaluation of the effectiveness of active interventions, and more accurate prediction of the pandemic progression in a country, region or city. A pseudo-coevolutionary algorithm is adopted in the model fitting to provide an effective estimation of high-dimensional unknown parameter sets using a limited amount of historical data. The model is applied to 175 regions in Australia, Canada, Italy, Japan, Spain, the UK and USA to estimate the number of unconfirmed cases using limited historical data. Results showed that the actual number of infected cases could be 4.309 times as many as the official confirmed number. By implementing mass COVID-19 testing, the number of infected cases could be reduced by about 50%.

The Coronavirus Disease 2019 (COVID-19) began to outbreak since December 2019 and widely spread over the world. How to accurately predict the spread of COVID-19 is one of the essential issues for controlling the pandemic. This study establishes a general model that can predict the trend of COVID-19 in a country based on historical COVID-19 data in 184 countries. First, Savitzky-Golay (S-G) filter is utilized to detect multiple waves of COVID-19 in a country. Then, a PSO-SIR (particle swarm optimization susceptible-infected-recovery) model is provided for data augmentation. Finally, a novel PSO-BLS (particle swarm optimization broad learning system) is proposed for predicting the trend of COVID-19. Experimental results show that compared with the deep learning models (ANN, CNN, LSTM, and GRU), the PSO-BLS algorithm has higher accuracy and stability in predicting the number of active infected cases and removed cases.

PM2.5 (particular matter with a diameter of 2.5μm or less) is one of the most important indicators of air pollution. In the field of environmental science, how to forecast PM2.5 is an important topic. We construct a previous 24-hour indicator before the predicted point to construct an enhanced dataset for PM2.5 concentration prediction. However, with a large scale of features, the performances of fundamental neural networks are not stable or accurate enough. As a result, an ensemble GRU (Gate Recurrent Unit) neural network is proposed for short-term PM2.5 prediction. This approach can improve accuracy while maintaining stability by combining the outputs after varying training. In this study, a dataset, which recording 6 indicators (PM2.5, PM10, CO, NO2, O3, SO2) for more than 20,000 hours in Shenzhen, is adopted to evaluate the proposed approach. Experimental results indicate that the proposed ensemble GRU model provides the lowest scores in MSE, RMSE criteria, and the best average-results in R2, MSE, RMSE scores.

The housing market is increasing huge, predicting housing prices is not only important for a business issue, but also for people. However, housing price fluctuations have a lot of influencing factors. Also, there is a non-linear relationship between housing prices and housing factors. Most econometric or statistical models cannot capture non-linear relationships yet. Therefore, we propose housing price prediction models based on deep learning methods, which can capture non-linear relationships. In this work, we construct a dataset, including the housing attributes data and macroeconomic data in Taiwan from January 2013 to December 2018. The housing attributes data includes two types of housing transactions, which are “land + building” (Type1) and “land + building + park” (Type2). Macroeconomic data includes housing investment demand ratio, owner-occupier housing ratio, housing price to income ratio, housing loan burden ratio, and housing bargaining space ratio. Then, this dataset is utilized to evaluate the prediction methods based on deep learning algorithms BPNN and CNN to predict housing prices. Experimental results show that CNN with housing features has the best prediction effect. This study can be used to develop targetted interventions aimed at the housing market.

The public's purchase incentive increases dramatically during the holiday season and subsequently returns to normal levels. This seasonality is common in various scenarios and highlights the following questions: how does the public's purchase incentive fluctuate over the course of a year? Which factors are conducive to this seasonal behavior and how can they be modeled? In this paper, we propose a model that explicitly integrates temporal point process theory with the construction of a networked community, to describe the dynamics of collective action propagation with seasonal fluctuation. Furthermore, a database is constructed of sales records for 21 video game consoles and 13 237 video games in France, Germany, Japan, the U.K., the USA, and worldwide from 1989 to 2018. Experimental results suggest that peak desire always appears in the holiday season about one week before Christmas and is about four times higher than consumption desire in a normal period in all areas.

Imbalanced dataset is a common issue in many applications. The one-class Support Vector Machine (SVM) is found to be an effective algorithm to construct classification models over the underlying imbalanced dataset. In some cases, feature extraction is hard and one would prefer using pre-defined kernels to train the model. In traditional practice, a valid kernel has to satisfy the Mercer's condition, which may restrict the design of kernel functions or matrices. In this paper, an indefinite kernel extension is applied to the one-class SVM model in order to relieve such limitation. To illustrate its performance, the algorithm is applied to perform virtual screening of drugs.

Recently, technological advances have made possible the measure of daily (even hourly) user(sales) growth of a product or service. Here, questions comes: how users grow and how the promotion influence the user growth? Here, we develop a model that can describe the growth of the user population of a newly launched product or service, which can answer this question. To develop this model, we consider a network of interacting individuals, whose actions or transitions are determined by the states (behaviour) of their neighbours as well as their own personal decisions. This model leads to a simple growth equation connecting the growth of user population with the total number of prospective users, and the effects of peer influence and personal choice. Several real-life datasets of a variety of products and services have been analyzed. Results suggest that they all follow the proposed growth equation. The numerical procedure for finding the model parameters thus ensure the relative effectiveness, market size and promotional efforts to be estimated from the available historical growth data.

Precipitation or rainfall prediction is an important practical problem in meteorology research. However, a reliable forecast is difficult to achieve due to the complexity of the climate system and the huge set of available climatic data at nearby sites. Thus, in most existing precipitation prediction systems or methods, inevitable partial (selective) consideration of the types of climatic data has often led to less satisfactory prediction results. Thanks to the advent of modern sensor technology, various types of climatic data in time-series forms can be collected at observation sites or satellites. In this work, we reconstruct a climate dataset including more than 30 climatic variables measured by more than 103,473 observation sites covering the world surface from 1800 to 2017. Then, we apply state-of-the-art machine learning methods, including deep learning (CNN, RNN, and LSTM networks) and ensemble learning (Adaboost, GBDT, and XGBoost), to develop a short-term precipitation system. Experiments on real-world data show that incorporating multiple climate variables into a prediction system improves the prediction results. The best performance of the proposed method reaches an accuracy of more than 80%.

By exploiting the advanced information and communication technologies, the current community healthcare systems provide digital healthcare services. However, the current healthcare framework for senior citizen in outdoor environment faces new challenges. The traditional healthcare systems are not efficient, and they do not comprise user‐friendly devices and interfaces suitable for the elderly in outdoor environment. Hence, in this work, we develop an outdoor healthcare system for community senior citizens based on unmanned aerial vehicle (UAV) and Internet of things (IoT). Our system includes physical devices, wireless and wired networks, cloud and data center, and smart terminals. Further, the analysis of the proposed healthcare architecture is presented from the perspective of different layers, and an algorithm on UAV for creating high‐speed communication channel and delivering medicine is provided. In addition, the healthcare UAV, related devices, and friendly APPs are designed. The proposed framework is evaluated by comparison with the current healthcare architecture. The emulating and experimental results show that our proposed system can provide a high‐quality wireless communication link even at large communication distances, and in real testbed our proposal can reduce the response time by about 20% in comparison with the current methods.

In our analysis, we consider daily rainfall records y(n) from 3825 areas covering entire China. The daily rainfall records is 50 years long from 1961 to 2011. We study the statistical properties of the daily rainfall data and return intervals Tqbetween two consecutive rainfall records above some threshold q. Using the detrended fluctuation analysis (DFA) method to analyze the long-term correlation properties of the rainfall record y(n) and return intervals Tq, we find that the long-term correlation is not exist in rainfall record, while a weak long-term correlation is found in return intervals, pointing to a random regulation of rainfall dynamics. Then, we further analyze the probability distribution function (PDF) of the rainfall record and return intervals of all the 3825 areas, and find that all the PDF obey a power-law distribution.

During the past decade, graph-based semi-supervised learning has become one of the most important research areas in machine learning and artificial intelligence community. In this paper, we propose a Capped l2,1-Norm Regularized Dictionary Learning to construct the graph for semi-supervised learning (SSL). The new sparse coding is robust to the outliers and insensitive to noise. Then, the developed graph construction can better characterize the geometrical structure of data mainfold and be utilized for efficient graph based SSL. the proposed SSL methods can also be easily extended to deal with out-of-sample data. Simulation results show that the proposed method can achieve better performance compared with other state-of-the-art graph based SSL methods.

In this paper, we develop a stochastic model from the perspective of complex networks to investigate the effects of cyber coupling on cascading failures in coupled power systems. The failure spreading in the coupled system is described by state transition and modeled as a Markov process. We simulate the dynamic profile of the cascading failures caused by the attack of cyber malwares, considering the effects of power overloading, contagion and interdependence between power grids and cyber networks. We study the coupled system created by coupling the UIUC 150 Bus System with synthesized scale-free cyber network. Simulation results present that the dynamic profile of the cascading failures in a coupled system displays a “staircase-like” pattern which can be interpreted as a combined feature of the typical step propagation profile triggered repeatedly by cyber attacks due to cyber network coupling. Results also show that cyber coupling can intensify both the extent and rapidity of power blackouts. Moreover, adopting assortative coupling patten accelerates the failures propagation, in particular under high-degree cyber node targeted attack.

In this paper, we investigate the dynamic profiles of the cascading failure propagations in power systems. We use a circuit-based power flow model and combine it with a stochastic model to describe the uncertain failure time instants. The sequence of failures is determined by power stresses of individual elements which are governed by deterministic circuit equations, while the time durations between failures are described by stochastic processes. The use of stochastic processes here addresses the uncertainties in individual components' physical failure mechanisms which may depend on manufacturing quality and environmental factors. In this model, the element failure rate is related to the extent of overloading. A network-based stochastic model is developed to study the failure propagation dynamics of the entire power network. Simulation results show that our model generates dynamic profiles of cascading failures that contains all salient features displayed in historical blackout data. The proposed model thus offers predictive information about occurrences of large-scale blackouts.

The insufficiency of labeled samples is a major problem in automatic image annotation. However, unlabeled samples are readily available and abundant. Hence, semi-supervised learning methods, which utilize partly labeled samples and a large amount of unlabeled samples, have attracted increased attention in the field of image classification. During the past decade, graph-based semi-supervised learning became one of the most important research areas in semi-supervised learning. In this letter, we propose a novel and effective graph based semi-supervised learning method for image classification. The new method is based on local and global regression regularization. The local regression regularization adopts a set of local classification functions to preserve both local discriminative and geometrical information; while the global regression regularization preserves the global discriminative information and calculates the projection matrix for out-of-sample extrapolation. Extensive simulations based on synthetic and real-world datasets verify the effectiveness of the proposed method.

S-system is a commonly used model for dynamic biological system. However, due to technical limitations, biological reaction networks are often only partially observable, which indicates that not all individuals incorporated in the model can be measured directly. Given a limited amount and quality of experimental data, it cannot be assured that the model structure and parameters can be estimated unambiguously. Hence, it is important to known in what situation the S-system is identifiable and which are not. This knowledge is essential on the construction of model and, indeed, the analysis of the underlying biological system. In this work we shall perform a theoretic discussion on the identifiability of S-system. We shall show that the S-system is identifiable under a set of assumptions. To illustrate the identification performance of S-system, an application on yeast fermentation pathway is conducted.

The inverse problem of identifying unknown parameters of known structure dynamical biological systems, which are modelled by ordinary differential equations or delay differential equations, from experimental data is treated in this paper. A two stage approach is adopted: first, combine spline theory and Nonlinear Programming (NLP), the parameter estimation problem is formulated as an optimization problem with only algebraic constraints; then, a new differential evolution (DE) algorithm is proposed to find a feasible solution. The approach is designed to handle problem of realistic size with noisy observation data. Three cases are studied to evaluate the performance of the proposed algorithm: two are based on benchmark models with priori-determined structure and parameters; the other one is a particular biological system with unknown model structure. In the last case, only a set of observation data available and in this case a nominal model is adopted for the identification. All the test systems were successfully identified by using a reasonable amount of experimental data within an acceptable computation time. Experimental evaluation reveals that the proposed method is capable of fast estimation on the unknown parameters with good precision.

Knowledge of interactions between proteins is an importance piece of information on the understanding of cellular mechanism. One of the challenges is to quantify how similar the two protein-protein interaction (PPI) networks. To provide a candidate solution of this problem, a distance measure for PPI networks is proposed. This distance measure involves solving an optimisation problem with which the objective function being a weighted sum of topological and biological measure of an alignment respectively. To solve this problem, the projected subgradient method is employed. To illustrate the performance and usage of this distance measure, it is applied to the PPI networks of herpesvirus family. Five herpesviruses: EpsteinBarr virus (EBV), Herpes simplex virus (HSV), Mouse Cytomegalovirus (mCMV), Kaposi's sarcoma-associated herpesvirus (KSHV) and Varicella zoster virus (VZV) are considered in this paper. PPI network distances of the five her-pesviruses are computed and visualized using multidimensional scaling (MDS). Results show that the distance measure can reflect the dissimilarity among organisms. In addition our algorithm can also separate herpesvirus subfamilies given only topological information of the PPI network.