Search

This paper deals with a cost sensitive extension of the standard Support Vector Machine (SVM) using an ordered weighted sum of the deviations of misclassified individuals with respect to their corresponding supporting hyperplanes. In contrast with previous heuristic approaches, an exact method that applies the ordered weighted average operator in the classical SVM model is proposed. Specifically, when weights are sorted in non-decreasing order, a quadratic continuous formulation is developed. For general weights, a mixed integer quadratic formulation is proposed. In addition, our results prove that nonlinear kernel functions can be also applied to these new models extending its applicability beyond the linear case. Extensive computational results reported in the paper show that the predictive performance provided by the proposed exact solution approaches are better than the ones provided by the classical models (linear and nonlinear kernel) and similar or better than the previous ones provided by the heuristic solution by Maldonado et al. (2018). © 2021 The Authors, Alfredo Marín has been supported by Spanish Ministry of Science and Innovation under project PID2019-110886RB-I00 . Part of this research was conducted while he was on sabbatical at Universidad de Sevilla, Spain . Luisa I. Martínez-Merino, Antonio M. Rodríguez-Chía and Justo Puerto thank the Agencia Estatal de Investigación (AEI) and the European Regional Development’s funds (ERDF) : project PID2020-114594GB-C21/C22 ; Regional Government of Andalusia : projects FEDER-UCA18-106895 , FEDER-US-1256951 and P18-FR-1422 ; and Fundación BBVA : project NetmeetData (Ayudas Fundación BBVA a equipos de investigación científica 2019). Luisa I. Martínez-Merino has been supported by Subprograma Juan de la Cierva Formación 2019 ( FJC2019-039023-I ) and PAIDI 2020 postdoctoral fellowship financed by European Social Fund and Junta de Andalucía

The traditional literature analysis mainly focuses on the literature metadata such as topics, authors, keywords, references, and rarely pays attention to the main content of papers. However, in many scientific domains such as science, computing, engineering, the methods and datasets involved in the papers published carry important information and are quite useful for domain analysis and recommendation. Method and dataset entities have various forms, which are more difficult to recognize than common entities. In this paper, we propose a novel Method and Dataset Entity Recognition model (MDER), which is able to effectively extract the method and dataset entities from the main textual content of scientific papers. The model is the first to combine rule embedding, a parallel structure of Convolutional Neural Network (CNN) and a two-layer Bi-directional Long Short-Term Memory (BiLSTM) with the self-attention mechanism. We evaluate the proposed model on datasets constructed from the published papers of different research areas in computer science. Our model performs well in multiple areas and features a good capacity for cross-area learning and recognition. Ablation study indicates that building different modules collectively contributes to the good entity recognition performance as a whole. The data augmentation positively contributes to model training, making our model much more robust. We finally apply the proposed model on PAKDD papers published from 2009–2019 to mine insightful results over a long time span. 1

The recently introduced Marine Predators Algorithm (MPA) exhibits competitive performance in solving optimization problems. However, it often prematurely converges due to an imbalance between its exploration and exploitation capabilities. Therefore, in this paper, an improved MPA variant using a proposed Local Escaping Operator (LEO) is introduced. In the MPA, solution candidates are replaced by better candidates from the previous iteration, mimicking a memory of already visited prey-abundant areas. This indicates a weak inter-dependence between the candidates in the population and possible acceptance of new solution candidates created outside the algorithm without damaging the optimization process. Consequently, in the proposed approach, the worst candidates are replaced with solutions created by the LEO. The LEO is based on representative solutions, taking into account the relationship between predators and the characteristics of their population. The approach is experimentally compared with the state-of-the-art meta-heuristics on 82 test functions, including IEEE Congress on Evolutionary Computation test suites (CEC’14 and CEC’17) and three engineering problems. The results show the superiority of the LEO-MPA over the MPA and recent algorithms. Furthermore, in this paper, the suitability of the hybridization of meta-heuristics with the LEO is discussed and successful attempts with the best algorithms are shown.

The digitalization of the supply chain (SC) enables companies to address customers’ new requirements, the challenges of managing the SC and the expectations, and efficiency improvement. Although digital supply chain (DSC) is a buzzword, very few organizations and decision-makers understand the challenges of transforming from a traditional supply chain to a DSC. Therefore, the purpose of this paper is to develop and implement an integrated knowledge-based system (KBS) and evaluate the overall DSC readiness value (score) of an organization. DSC readiness factors are identified from an extensive literature review and validated by experts from academia and industry. The proposed KBS is based on Fuzzy-AHP​ that establishes a link between DSC readiness factors and their impact on performance and evaluates overall DSC readiness value (score). The proposed KBS has been validated in an Indian manufacturing company, and results show that the overall DSC readiness value (score) of the case company is 0.267. This paper will help organizations evaluate their current position in transforming from traditional supply chain to DSC and make a plan to achieve digitalization completely in their SC operations. Managers, practitioners, and decision-makers who are involved in digitalization can use our study’s findings as a starting point for aiding the transformation.

Gradient boosted trees (GBTs) are ubiquitous models used by researchers, machine learning (ML) practitioners, and data scientists because of their robust performance, interpretable behavior, and ease-of-use. One critical challenge in training GBTs is the tuning of their hyperparameters. In practice, selecting these hyperparameters is often done manually. Recently, the ML community has advocated for tuning hyperparameters through black-box optimization and developed state-of-the-art systems to do so. However, applying such systems to tune GBTs suffers from two drawbacks. First, these systems are not \textit{model-aware}, rather they are designed to apply to a \textit{generic} model; this leaves significant optimization performance on the table. Second, using these systems requires \textit{domain knowledge} such as the choice of hyperparameter search space, which is an antithesis to the automatic experimentation that black-box optimization aims to provide. In this paper, we present SigOpt Mulch, a model-aware hyperparameter tuning system specifically designed for automated tuning of GBTs that provides two improvements over existing systems. First, Mulch leverages powerful techniques in metalearning and multifidelity optimization to perform model-aware hyperparameter optimization. Second, it automates the process of learning performant hyperparameters by making intelligent decisions about the optimization search space, thus reducing the need for user domain knowledge. These innovations allow Mulch to identify good GBT hyperparameters far more efficiently -- and in a more seamless and user-friendly way -- than existing black-box hyperparameter tuning systems.

Although traditional symbolic reasoning methods are highly interpretable, their application in knowledge graph link prediction is limited due to their low computational efficiency. In this paper, we propose a new neural symbolic reasoning method: RNNCTPs, which improves computational efficiency by re-filtering the knowledge selection of Conditional Theorem Provers (CTPs), and is less sensitive to the embedding size parameter. RNNCTPs are divided into relation selectors and predictors. The relation selectors are trained efficiently and interpretably, so that the whole model can dynamically generate knowledge for the inference of the predictor. In all four datasets, the method shows competitive performance against traditional methods on the link prediction task, and can have higher applicability to the selection of datasets relative to CTPs., Comment: 16 pages, 8 figures,5 tables

Emotional Support Conversation (ESConv) aims to reduce help-seekers'emotional distress with the supportive strategy and response. It is essential for the supporter to select an appropriate strategy with the feedback of the help-seeker (e.g., emotion change during dialog turns, etc) in ESConv. However, previous methods mainly focus on the dialog history to select the strategy and ignore the help-seeker's feedback, leading to the wrong and user-irrelevant strategy prediction. In addition, these approaches only model the context-to-strategy flow and pay less attention to the strategy-to-context flow that can focus on the strategy-related context for generating the strategy-constrain response. In this paper, we propose a Feedback-Aware Double COntrolling Network (FADO) to make a strategy schedule and generate the supportive response. The core module in FADO consists of a dual-level feedback strategy selector and a double control reader. Specifically, the dual-level feedback strategy selector leverages the turn-level and conversation-level feedback to encourage or penalize strategies. The double control reader constructs the novel strategy-to-context flow for generating the strategy-constrain response. Furthermore, a strategy dictionary is designed to enrich the semantic information of the strategy and improve the quality of strategy-constrain response. Experimental results on ESConv show that the proposed FADO has achieved the state-of-the-art performance in terms of both strategy selection and response generation. Our code is available at https://github.com/Thedatababbler/FADO., Accepted on Knowl. Based Syst. (SCI I)

Financiado para publicar en acceso aberto. Universidade da Coruña/CISUG [Abstract]: In recent years, mainly due to the application of smartphones in this area, research in human activity recognition (HAR) has shown a continuous and steady growth. Thanks to its wide range of sensors, its size, its ease of use, its low price and its applicability in many other fields, it is a highly attractive option for researchers. However, the vast majority of studies carried out so far focus on laboratory settings, outside of a real-life environment. In this work, unlike in other papers, progress was sought on the latter point. To do so, a dataset already published for this purpose was used. This dataset was collected using the sensors of the smartphones of different individuals in their daily life, with almost total freedom. To exploit these data, numerous experiments were carried out with various machine learning techniques and each of them with different hyperparameters. These experiments proved that, in this case, tree-based models, such as Random Forest, outperform the rest. The final result shows an enormous improvement in the accuracy of the best model found to date for this purpose, from 74.39% to 92.97%. Xunta de Galicia; ED431G 2019/01 Xunta de Galicia; ED481A 2020/003 Xunta de Galicia; ED431C 2022/46 Xunta de Galicia; ED431C 2018/49 Xunta de Galicia; ED431C 2021/53 This research was partially funded by MCIN/AEI/10.13039/ 501100011033, NextGenerationEU/PRTR, FLATCITY-POC , Spain [grant number P DC2021-121239-C31]; MCIN/AEI/10.13039/ 501100011033 MAGIST, Spain [grant number P ID2019-105221RB-C41]; Xunta de Galicia/FEDER-UE, Spain [grant numbers ED431G 2019/01 , ED481A 2020/003 , ED431C 2022/46 , ED431C 2018/49 and ED431C 2021/53 ]. Funding for open access charge: Universidade da Coruña/CISUG.

Data availability: The code used in this paper is available at: https://github.com/mxt0607/Two_Stage_RUL_Prediction. This article puts forward a two-stage integrated method to predict the remaining useful life (RUL) of lithium-ion batteries (LIBs). At the first stage, a convolutional neural network (CNN) is employed to preliminarily estimate the cycle life of each testing LIB, where the network structure of the CNN is carefully designed to extract the discharge capacity features. By analyzing the cycle lives, an LIB which has the most similar degradation mode to each testing LIB is chosen from the training dataset. The capacities of the selected LIB are identified based on a double exponential model (DEM). At the second stage, the identified DEM is utilized as the initial mean function of the Gaussian process regression (GPR) algorithm. The GPR algorithm is then applied to early RUL prediction of each testing LIB in a personalized manner. To verify the efficacy of the proposed method, four LIBs with long-term cycle lives are selected as the testing dataset. Experimental results show the superior performance of the proposed method over the standard CNN-based RUL prediction method and the standard GPR-based RUL prediction method. This work was supported in part the National Natural Science Foundation of China under Grants 62273264, 51905197 and 61933007.

This paper investigates Graph Neural Networks (GNNs) application for self-supervised network intrusion and anomaly detection. GNNs are a deep learning approach for graph-based data that incorporate graph structures into learning to generalise graph representations and output embeddings. As network flows are naturally graph-based, GNNs are a suitable fit for analysing and learning network behaviour. The majority of current implementations of GNN-based Network Intrusion Detection Systems (NIDSs) rely heavily on labelled network traffic which can not only restrict the amount and structure of input traffic, but also the NIDSs potential to adapt to unseen attacks. To overcome these restrictions, we present Anomal-E, a GNN approach to intrusion and anomaly detection that leverages edge features and graph topological structure in a self-supervised process. This approach is, to the best our knowledge, the first successful and practical approach to network intrusion detection that utilises network flows in a self-supervised, edge leveraging GNN. Experimental results on two modern benchmark NIDS datasets not only clearly display the improvement of using Anomal-E embeddings rather than raw features, but also the potential Anomal-E has for detection on wild network traffic., Comment: Accepted by Knowledge-Based Systems

Gene expression data have become increasingly important in machine learning and computational biology over the past few years. In the field of gene expression analysis, several matrix factorization-based dimensionality reduction methods have been developed. However, such methods can still be improved in terms of efficiency and reliability. In this paper, an innovative approach to feature selection, called Dual Regularized Unsupervised Feature Selection Based on Matrix Factorization and Minimum Redundancy (DR-FS-MFMR), is introduced. The major focus of DR-FS-MFMR is to discard redundant features from the set of original features. In order to reach this target, the primary feature selection problem is defined in terms of two aspects: (1) the matrix factorization of data matrix in terms of the feature weight matrix and the representation matrix, and (2) the correlation information related to the selected features set. Then, the objective function is enriched by employing two data representation characteristics along with an inner product regularization criterion to perform both the redundancy minimization process and the sparsity task more precisely. To demonstrate the proficiency of the DR-FS-MFMR method, a large number of experimental studies are conducted on nine gene expression datasets. The obtained computational results indicate the efficiency and productivity of DR-FS-MFMR for the gene selection task.

Human motion prediction, which plays a key role in computer vision, generally requires a past motion sequence as input. However, in real applications, a complete and correct past motion sequence can be too expensive to achieve. In this paper, we propose a novel approach to predicting future human motions from a much weaker condition, i.e., a single image, with mixture density networks (MDN) modeling. Contrary to most existing deep human motion prediction approaches, the multimodal nature of MDN enables the generation of diverse future motion hypotheses, which well compensates for the strong stochastic ambiguity aggregated by the single input and human motion uncertainty. In designing the loss function, we further introduce the energy-based formulation to flexibly impose prior losses over the learnable parameters of MDN to maintain motion coherence as well as improve the prediction accuracy by customizing the energy functions. Our trained model directly takes an image as input and generates multiple plausible motions that satisfy the given condition. Extensive experiments on two standard benchmark datasets demonstrate the effectiveness of our method in terms of prediction diversity and accuracy.

Few-shot point cloud semantic segmentation learns to segment novel classes with scarce labeled samples. Within an episode, a novel target class is defined by a few support samples with corresponding binary masks, where only the points of this class are labeled as foreground and others are regarded as background. In the tasks involving multiple target classes, since the meanings of background are diverse for different target classes, background ambiguities appear: Some points labeled as background in one support sample may be of other target classes. It will result in incorrect guidance and damage model's segmentation performance. However, previous methods in the literature do not consider this problem. In this paper, we propose a simple yet effective approach to tackle background ambiguities, which adopts the entropy of predictions on query samples to the training objective function as an additional regularization. Besides, we design a feature transformation operation to reduce the feature differences between support and query samples. With our proposed approach, fine-tuning, a weak baseline method for few-shot segmentation, gains significant performance improvement (e.g., 7.48% and 7.04% in 2-way-1-shot and 3-way-1-shot tasks of S3DIS, respectively) and outperforms current state-of-the-art methods in all the task settings of S3DIS and ScanNet benchmark datasets. Ministry of Education (MOE) National Research Foundation (NRF) This work was supported by National Natural Science Foundation of China (NSFC) 61876208, Tip-top Scientific and Technical Innovative Youth Talents of Guangdong Special Support Program (2019TQ05X200), 2022 Tencent Wechat Rhino-Bird Focused Research Program Research (Tencent WeChat RBFR2022008) and the National Research Foundation, Singapore under its AI Singapore Programme (AISG Award No: AISG-RP-2018-003), the Ministry of Education, Singapore, under its Academic Research Fund Tier 2 (MOE-T2EP20220-0007) and Tier 1 (RG95/20).

This paper presents a batch-wise density-based clustering approach for local outlier detection in massive-scale datasets. Unlike the well-known traditional algorithms, which assume that all the data is memory-resident, our proposed method is scalable and processes the input data chunk-by-chunk within the confines of a limited memory buffer. A temporary clustering model is built at the first phase; then, it is gradually updated by analyzing consecutive memory loads of points. Subsequently, at the end of scalable clustering, the approximate structure of the original clusters is obtained. Finally, by another scan of the entire dataset and using a suitable criterion, an outlying score is assigned to each object called SDCOR (Scalable Density-based Clustering Outlierness Ratio). Evaluations on real-life and synthetic datasets demonstrate that the proposed method has a low linear time complexity and is more effective and efficient compared to best-known conventional density-based methods, which need to load all data into the memory; and also, to some fast distance-based methods, which can perform on data resident in the disk.

Automatic code review (ACR), which can relieve the costs of manual inspection, is an indispensable and essential task in software engineering. To deal with ACR, existing work is to serialize the abstract syntax tree (AST). However, making sense of the whole AST with sequence encoding approach is a daunting task, mostly due to some redundant nodes in AST hinder the transmission of node information. Not to mention that the serialized representation is inadequate to grasp the information of tree structure in AST. In this paper, we first present a new large-scale Apache Automatic Code Review (AACR) dataset for ACR task since there is still no publicly available dataset in this task. The release of this dataset would push forward the research in this field. Based on it, we propose a novel Simplified AST based Graph Convolutional Network (SimAST-GCN) to deal with ACR task. Concretely, to improve the efficiency of node information dissemination, we first simplify the AST of code by deleting the redundant nodes that do not contain connection attributes, and thus deriving a Simplified AST. Then, we construct a relation graph for each code based on the Simplified AST to properly embody the relations among code fragments of the tree structure into the graph. Subsequently, in the light of the merit of graph structure, we explore a graph convolution networks architecture that follows an attention mechanism to leverage the crucial implications of code fragments to derive code representations. Finally, we exploit a simple but effective subtraction operation in the representations between the original and revised code, enabling the revised difference to be preferably learned for deciding the results of ACR. Experimental results on the AACR dataset illustrate that our proposed model outperforms the state-of-the-art methods., 15 pages

Federated learning (FL) is a privacy-preserving paradigm that collaboratively train machine learning models with distributed data stored in different silos without exposing sensitive information. Different from most existing FL approaches requiring data from different parties share either the same feature space or sample ID space, federated transfer learning (FTL), which is a recently proposed FL concept, is designed for situations where data from different parties differ not only in samples but also in feature space. However, like most traditional FL approaches, FTL methods also suffer from issues caused by insufficiency of overlapping data. In this paper, we propose a novel FTL framework referred to as Semi-Supervised Federated Heterogeneous Transfer Learning (SFHTL) to leverage on the unlabeled non-overlapping samples to reduce model overfitting as a result of insufficient overlapping training samples in FL scenarios. Unlike existing FTL approaches, SFHTL makes use of non-overlapping samples from all parties to expand the training set for each party to improve local model performance. Through extensive experimental evaluation based on real-world datasets, we demonstrate significant advantages of SFHTL over state-of-the-art approaches. Nanyang Technological University National Research Foundation (NRF) This research is supported, in part, by the National Natural Science Foundation of China under Grant NSFC 62106167; Project Funded by the Priority Academic Program Development of Jiangsu Higher Education Institutions; the National Research Foundation, Singapore under its AI Singapore Programme (AISG Award No: AISG2-RP-2020-019); the RIE 2020 Advanced Manufacturing and Engineering (AME) Programmatic Fund (No. A20G8b0102), Singapore; Nanyang Assistant Professorship (NAP); the Joint NTUWeBank Research Centre on Fintech (Award No: NWJ-2020-008), Nanyang Technological University, Singapore; and Joint SDU-NTU Centre for Artificial Intelligence Research (C-FAIR) (NSC-2019- 011). Qiang Yang is supported in part by China National Key Research and Development Program of China under Grant No. 2018AAA0101100.

Nonnegative matrix factorization (NMF) has been widely studied in recent years due to its effectiveness in representing nonnegative data with parts-based representations. For NMF, a sparser solution implies better parts-based representation.However, current NMF methods do not always generate sparse solutions.In this paper, we propose a new NMF method with log-norm imposed on the factor matrices to enhance the sparseness.Moreover, we propose a novel column-wisely sparse norm, named $\ell_{2,\log}$-(pseudo) norm to enhance the robustness of the proposed method.The $\ell_{2,\log}$-(pseudo) norm is invariant, continuous, and differentiable.For the $\ell_{2,\log}$ regularized shrinkage problem, we derive a closed-form solution, which can be used for other general problems.Efficient multiplicative updating rules are developed for the optimization, which theoretically guarantees the convergence of the objective value sequence.Extensive experimental results confirm the effectiveness of the proposed method, as well as the enhanced sparseness and robustness.

Bayesian transfer learning (BTL) is defined in this paper as the task of conditioning a target probability distribution on a transferred source distribution. The target globally models the interaction between the source and target, and conditions on a probabilistic data predictor made available by an independent local source modeller. Fully probabilistic design is adopted to solve this optimal decision-making problem in the target. By successfully transferring higher moments of the source, the target can reject unreliable source knowledge (i.e. it achieves robust transfer). This dual-modeller framework means that the source's local processing of raw data into a transferred predictive distribution -- with compressive possibilities -- is enriched by (the possible expertise of) the local source model. In addition, the introduction of the global target modeller allows correlation between the source and target tasks -- if known to the target -- to be accounted for. Important consequences emerge. Firstly, the new scheme attains the performance of fully modelled (i.e. conventional) multitask learning schemes in (those rare) cases where target model misspecification is avoided. Secondly, and more importantly, the new dual-modeller framework is robust to the model misspecification that undermines conventional multitask learning. We thoroughly explore these issues in the key context of interacting Gaussian process regression tasks. Experimental evidence from both synthetic and real data settings validates our technical findings: that the proposed BTL framework enjoys robustness in transfer while also being robust to model misspecification.

In this paper, a novel unsupervised low-rank representation model, i.e., Auto-weighted Low-Rank Representation (ALRR), is proposed to construct a more favorable similarity graph (SG) for clustering. In particular, ALRR enhances the discriminability of SG by capturing the multi-subspace structure and extracting the salient features simultaneously. Specifically, an auto-weighted penalty is introduced to learn a similarity graph by highlighting the effective features, and meanwhile, overshadowing the disturbed features. Consequently, ALRR obtains a similarity graph that can preserve the intrinsic geometrical structures within the data by enforcing a smaller similarity on two dissimilar samples. Moreover, we employ a block-diagonal regularizer to guarantee the learned graph contains $k$ diagonal blocks. This can facilitate a more discriminative representation learning for clustering tasks. Extensive experimental results on synthetic and real databases demonstrate the superiority of ALRR over other state-of-the-art methods with a margin of 1.8\%$\sim$10.8\%.

Yoga is a traditional Indian exercise. It specifies various body postures called asanas, practicing them is beneficial for the physical, mental, and spiritual well-being. To support the yoga practitioners, there is a need of an expert yoga asanas recognition system that can automatically analyze practitioner’s postures and could provide suitable posture correction instructions. This\ud paper proposes YogNet, a multi-person yoga expert system for 20 asanas using a two-stream deep\ud spatiotemporal neural network architecture. The first stream utilizes a keypoint detection approach\ud to detect the practitioner’s pose, followed by the formation of bounding boxes across the subject. The model then applies time distributed convolutional neural networks (CNNs) to extract framewise postural features, followed by regularized long short-term memory (LSTM) networks to give temporal predictions. The second stream utilizes 3D-CNNs for spatiotemporal feature extraction from RGB videos. Finally, the scores of two streams are fused using multiple fusion techniques.\ud A yoga asana recognition database (YAR) containing 1206 videos is collected using a single 2D web camera for 367 minutes with the help of 16 participants and contains four view variations i.e. front, back, left, and right sides. The proposed system is novel as this is the earliest two-stream deep learning-based system that can perform multi-person yoga asanas recognition and correction\ud in realtime. Simulation result reveals that YogNet system achieved 77.29%, 89.29%, and 96.31% accuracies using pose stream, RGB stream, and via fusion of both streams, respectively. These results are impressive and sufficiently high for recommendation towards general adaption of the system.

Blind image deblurring is an important yet very challenging problem in low-level vision. Traditional optimization based methods generally formulate this task as a maximum-a-posteriori estimation or variational inference problem, whose performance highly relies on the handcraft priors for both the latent image and the blur kernel. In contrast, recent deep learning methods generally learn, from a large collection of training images, deep neural networks (DNNs) directly mapping the blurry image to the clean one or to the blur kernel, paying less attention to the physical degradation process of the blurry image. In this paper, we present a deep variational Bayesian framework for blind image deblurring. Under this framework, the posterior of the latent clean image and blur kernel can be jointly estimated in an amortized inference fashion with DNNs, and the involved inference DNNs can be trained by fully considering the physical blur model, together with the supervision of data driven priors for the clean image and blur kernel, which is naturally led to by the evidence lower bound objective. Comprehensive experiments are conducted to substantiate the effectiveness of the proposed framework. The results show that it can not only achieve a promising performance with relatively simple networks, but also enhance the performance of existing DNNs for deblurring.

With the growing demand for hand hygiene and convenience of use, palmprint recognition with touchless manner made a great development recently, providing an effective solution for person identification. Despite many efforts that have been devoted to this area, it is still uncertain about the discriminative ability of the contactless palmprint, especially for large-scale datasets. To tackle the problem, in this paper, we build a large-scale touchless palmprint dataset containing 2334 palms from 1167 individuals. To our best knowledge, it is the largest contactless palmprint image benchmark ever collected with regard to the number of individuals and palms. Besides, we propose a novel deep learning framework for touchless palmprint recognition named 3DCPN (3D Convolution Palmprint recognition Network) which leverages 3D convolution to dynamically integrate multiple Gabor features. In 3DCPN, a novel variant of Gabor filter is embedded into the first layer for enhancement of curve feature extraction. With a well-designed ensemble scheme,low-level 3D features are then convolved to extract high-level features. Finally on the top, we set a region-based loss function to strengthen the discriminative ability of both global and local descriptors. To demonstrate the superiority of our method, extensive experiments are conducted on our dataset and other popular databases TongJi and IITD, where the results show the proposed 3DCPN achieves state-of-the-art or comparable performances.

Knowledge graph (KG) embedding encodes the entities and relations from a KG into low-dimensional vector spaces to support various applications such as KG completion, question answering, and recommender systems. In real world, knowledge graphs (KGs) are dynamic and evolve over time with addition or deletion of triples. However, most existing models focus on embedding static KGs while neglecting dynamics. To adapt to the changes in a KG, these models need to be retrained on the whole KG with a high time cost. In this paper, to tackle the aforementioned problem, we propose a new context-aware Dynamic Knowledge Graph Embedding (DKGE) method which supports the embedding learning in an online fashion. DKGE introduces two different representations (i.e., knowledge embedding and contextual element embedding) for each entity and each relation, in the joint modeling of entities and relations as well as their contexts, by employing two attentive graph convolutional networks, a gate strategy, and translation operations. This effectively helps limit the impacts of a KG update in certain regions, not in the entire graph, so that DKGE can rapidly acquire the updated KG embedding by a proposed online learning algorithm. Furthermore, DKGE can also learn KG embedding from scratch. Experiments on the tasks of link prediction and question answering in a dynamic environment demonstrate the effectiveness and efficiency of DKGE., Comment: 46 pages

Meta-learning has gained wide popularity as a training framework that is more data-efficient than traditional machine learning methods. However, its generalization ability in complex task distributions, such as multimodal tasks, has not been thoroughly studied. Recently, some studies on multimodality-based meta-learning have emerged. This survey provides a comprehensive overview of the multimodality-based meta-learning landscape in terms of the methodologies and applications. We first formalize the definition of meta-learning in multimodality, along with the research challenges in this growing field, such as how to enrich the input in few-shot learning (FSL) or zero-shot learning (ZSL) in multimodal scenarios and how to generalize the models to new tasks. We then propose a new taxonomy to discuss typical meta-learning algorithms in multimodal tasks systematically. We investigate the contributions of related papers and summarize them by our taxonomy. Finally, we propose potential research directions for this promising field., Comment: Accepted by Knowledge-Based Systems; 21 pages

Many important real-world applications involve time-series data with skewed distribution. Compared to conventional imbalance learning problems, the classification of imbalanced time-series data is more challenging due to high dimensionality and high inter-variable correlation. This paper proposes a structure preserving Oversampling method to combat the High-dimensional Imbalanced Time-series classification (OHIT). OHIT first leverages a density-ratio based shared nearest neighbor clustering algorithm to capture the modes of minority class in high-dimensional space. It then for each mode applies the shrinkage technique of large-dimensional covariance matrix to obtain accurate and reliable covariance structure. Finally, OHIT generates the structure-preserving synthetic samples based on multivariate Gaussian distribution by using the estimated covariance matrices. Experimental results on several publicly available time-series datasets (including unimodal and multimodal) demonstrate the superiority of OHIT against the state-of-the-art oversampling algorithms in terms of F1, G-mean, and AUC.

The last two decades have seen a tremendous increase in the adoption of Semantic Web technologies as a result of the availability of big data, the growth in computational power and the advancement of artificial intelligence (AI) technologies. Cutting-edge semantic techniques are now able to capture sentiments more accurately in various practical applications, including economic and financial forecasting. In particular, the extraction of sentiment from news text, social media and blogs for the prediction of economic and financial variables has attracted attention in recent years. Despite many successful applications of sentiment analysis (SA) in these domains, the range of semantic techniques employed is still limited and mostly focused on the detection of sentiment at a coarse-grained level, that is, whether the sentiment expressed by the entire text of a sentence is either positive or negative. This paper proposes a novel methodology for Fine-Grained Aspect-based Sentiment (FiGAS) analysis. The aim of the approach is to identify the sentiment associated to specific topics of interest in each sentence of a document and assigning real-valued polarity scores between -1 and +1 to those topics. The proposed approach is completely unsupervised and customized to the economic and financial domains by using a specialized lexicon make available along with the source code of FiGAS. Our lexicon-based SA approach relies on a detailed set of semantic polarity rules that allow understanding the origin of sentiment, in the spirit of the recent trend on \textit{Interpretable AI}. We provide an in-depth comparison of the performance of the FiGAS algorithm relative to other popular lexicon-based SA approaches in predicting a humanly annotated data set in the economic and financial domains. Our results indicate that FiGAS statistically outperforms the other methods by providing a sentiment score that is closer to one of the human annotators.

In this paper we propose a deep learning based approach for image retrieval using EEG. Our approach makes use of a multi-modal deep neural network based on metric learning, where the EEG signal from a user observing an image is mapped together with visual information extracted from the image. The inspiration behind this work is the vision of a system which allows the user to navigate their image catalogue just by thinking about the image they want to see. Thanks to our metric learning approach, the system is scalable in that it can operate with new images that have never been used in training, resulting in a zero-shot image retrieval system. This framework is tested in two different standard EEG image-viewing datasets, where we demonstrate state-of-the-art results in this complex scenario.

This paper considers the problem of Bayesian transfer learning-based knowledge fusion between linear state-space processes driven by uniform state and observation noise processes. The target task conditions on probabilistic state predictor(s) supplied by the source filtering task(s) to improve its own state estimate. A joint model of the target and source(s) is not required and is not elicited. The resulting decision-making problem for choosing the optimal conditional target filtering distribution under incomplete modelling is solved via fully probabilistic design (FPD), i.e. via appropriate minimization of Kullback-Leibler divergence (KLD). The resulting FPD-optimal target learner is robust, in the sense that it can reject poor-quality source knowledge. In addition, the fact that this Bayesian transfer learning (BTL) scheme does not depend on a model of interaction between the source and target tasks ensures robustness to the misspecification of such a model. The latter is a problem that affects conventional transfer learning methods. The properties of the proposed BTL scheme are demonstrated via extensive simulations, and in comparison with two contemporary alternatives., Comment: 39 pages

Deep learning plays a vital role in classifying different arrhythmias using electrocardiography (ECG) data. Nevertheless, training deep learning models normally requires a large amount of data and can lead to privacy concerns. Unfortunately, a large amount of healthcare data cannot be easily collected from a single silo. Additionally, deep learning models are like black-box, with no explainability of the predicted results, which is often required in clinical healthcare. This limits the application of deep learning in real-world health systems. In this paper, to address the above-mentioned challenges, we design a novel end-to-end framework in a federated setting for ECG-based healthcare using explainable artificial intelligence (XAI) and deep convolutional neural networks (CNN). The federated setting is used to solve challenges such as data availability and privacy concerns. Furthermore, the proposed framework effectively classifies different arrhythmias using an autoencoder and a classifier, both based on a CNN. Additionally, we propose an XAI-based module on top of the proposed classifier for interpretability of the classification results, which helps clinical practitioners to interpret the predictions of the classifier and to make quick and reliable decisions. The proposed framework was trained and tested using the baseline Massachusetts Institute of Technology - Boston’s Beth Israel Hospital (MIT-BIH) Arrhythmia database. The trained classifier outperformed existing work by achieving accuracy up to 94.5% and 98.9% for arrhythmia detection using noisy and clean data, respectively, with five-fold cross-validation. We also propose a new communication cost reduction method to reduce the communication costs and to enhance the privacy of users’ data in the federated setting. While the proposed framework was tested and validated for ECG classification, it is general enough to be extended to many other healthcare applications.

As an economical, low-carbon and convenient travel model, bike-sharing has become common in many cities around the world. However, the daily usage of shared bikes results in the dispatching problem, i.e., dispatching bikes to the specific destinations to satisfy riding demands. The bike-sharing platform can hire riders as workers and pay to incentivize them to accomplish the dispatching tasks. However, there exist multiple workers competing for the dispatching tasks, and they may strategically report their task accomplishing costs (which are usually private information only known by themselves) in order to make more profits, which may result in inefficient task dispatching results. In this paper, we first design a dispatching algorithm named GDY-MAX to allocate tasks to workers, which can achieve good performance. However, it cannot prevent workers strategically misreporting their task accomplishing costs. Regarding this issue, we further design a strategy proof mechanism under the budget constraint, which consists of a task dispatching algorithm and a worker pricing algorithm. We theoretically prove that our mechanism can satisfy incentive compatibility, individual rationality, budget constraint and a constant approximation ratio. Furthermore, we run extensive experiments to evaluate our mechanism based on a Mobike dataset. The results show that the performance of the proposed strategy proof mechanism and GDY-MAX is similar to the optimal algorithm in terms of the coverage ratio of accomplished task regions and the sum of task region value, and our mechanism has better performance than the uniform algorithm in terms of the total payment and the unit cost value.

Recently, many evolutionary computation methods have been developed to solve the feature selection problem. However, the studies focused mainly on small-scale issues, resulting in stagnation issues in local optima and numerical instability when dealing with large-scale feature selection dilemmas. To address these challenges, this paper proposes a novel weighted differential evolution algorithm based on self-adaptive mechanism, named SaWDE, to solve large-scale feature selection. First, a multi-population mechanism is adopted to enhance the diversity of the population. Then, we propose a new self-adaptive mechanism that selects several strategies from a strategy pool to capture the diverse characteristics of the datasets from the historical information. Finally, a weighted model is designed to identify the important features, which enables our model to generate the most suitable feature-selection solution. We demonstrate the effectiveness of our algorithm on twelve large-scale datasets. The performance of SaWDE is superior compared to six non-EC algorithms and six other EC algorithms, on both training and test datasets and on subset size, indicating that our algorithm is a favorable tool to solve the large-scale feature selection problem. Moreover, we have experimented SaWDE with six EC algorithms on twelve higher-dimensional data, which demonstrates that SaWDE is more robust and efficient compared to those state-of-the-art methods. SaWDE source code is available on Github at https://github.com/wangxb96/SaWDE .

Representation learning on graphs has emerged as a powerful mechanism to automate feature vector generation for downstream machine learning tasks. The advances in representation on graphs have centered on both homogeneous and heterogeneous graphs, where the latter presenting the challenges associated with multi-typed nodes and/or edges. In this paper, we consider the additional challenge of evolving graphs. We ask the question of whether the advances in representation learning for static graphs can be leveraged for dynamic graphs and how? It is important to be able to incorporate those advances to maximize the utility and generalization of methods. To that end, we propose the Framework for Incremental Learning of Dynamic Networks Embedding (FILDNE), which can utilize any existing static representation learning method for learning node embeddings while keeping the computational costs low. FILDNE integrates the feature vectors computed using the standard methods over different timesteps into a single representation by developing a convex combination function and alignment mechanism. Experimental results on several downstream tasks, over seven real-world data sets, show that FILDNE is able to reduce memory (up to 6x) and computational time (up to 50x) costs while providing competitive quality measure gains (e.g., improvements up to 19 pp AUC on link prediction and up to 33 pp mAP on graph reconstruction) with respect to the contemporary methods for representation learning on dynamic graphs.

Gender violence is a problem that affects millions of people worldwide. Among its many manifestations Intimate Partner Violence (IPV) is one of the most common. In Spain, a police monitoring protocol has been developed to minimize recidivism in IPV cases. This protocol is complemented by VioGen, an Intimate Partner Violence Risk Assessment Tool (IPVRAT) created by the Spanish State Secretariat for Security of the Ministry of Interior (SES) for risk prediction. VioGen’s goal is to help the authorities determine what security and safety measures are most suitable. This paper improves on the current version of VioGen by introducing a model based on machine learning and data science and by studying the predictive value of exogenous and historical variables. The model is fitted on an anonymized database provided by SES and extracted from VioGen. This database includes the 2-year evolution of 46,047 new cases of IPV violence reported between October 2016 and December 2017, making it the largest database analyzed in the field. Obtained results show a clear improvement in the predictive capabilities of the new model against the original system, where it would have corrected more than 25% of the infra-protected cases, while improving the overall accuracy at the same time. Finally, lessons learned from the performed study and experiments are reported to aid in the design of future IPVRAT. In particular, insights show that IPVRAT should not treat cases statically as the incorporation of information regarding their evolution improves significantly the model’s performance.

Recently, distillation approaches for extracting general knowledge from a teacher network to guide a student network have been suggested. Most existing methods transfer knowledge from the teacher to the student network by feeding a sequence of random minibatches sampled uniformly from the data. We argue that, instead, a compact student network should be guided gradually using samples ordered in a meaningful sequence. Thus, the gap in feature representation between the teacher and student network can be bridged step by step. In this paper, we provide a curriculum learning knowledge distillation framework via instance-level sequence learning. It employs the student network of the early epoch as a snapshot to create a curriculum for the student network’s next training phase. We performed extensive experiments using the CIFAR-10, CIFAR-100, SVHN, and CINIC-10 datasets. When compared with several state-of-the-art methods, our framework achieved the best performance with fewer iterations.

Due to the development of graph neural networks, graph-based representation learning methods have made great progress in recommender systems. However, data sparsity is still a challenging problem that most graph-based recommendation methods are confronted with. Recent works try to address this problem by utilizing side information. In this paper, we exploit the relevant and easily accessible textual information by advanced natural language processing (NLP) models and propose a light RGCN-based (RGCN, relational graph convolutional network) collaborative filtering method on heterogeneous graphs. Specifically, to incorporate rich textual knowledge, we utilize a pre-trained NLP model to initialize the embeddings of text nodes. Afterward, by performing a simplified RGCN-based node information propagation on the constructed heterogeneous graph, the embeddings of users and items can be adjusted with textual knowledge, which effectively alleviates the negative effects of data sparsity. Moreover, the matching function used by most graph-based representation learning methods is the inner product, which is not appropriate for the obtained embeddings that contain complex semantics. We design a predictive network that combines graph-based representation learning with neural matching function learning, and demonstrate that this architecture can bring a significant performance improvement. Extensive experiments are conducted on three publicly available datasets, and the results verify the superior performance of our method over several baselines., Comment: Accepted by Knowledge-Based Systems

In this paper, the twin-systems approach is reviewed, implemented, and competitively tested as a post-hoc explanation-by-example solution to the eXplainable Artificial Intelligence (XAI) problem. In twin-systems, an opaque artificial neural network (ANN) is explained by “twinning” it with a more interpretable case-based reasoning (CBR) system, by mapping the feature weights from the former to the latter. Extensive comparative tests are performed, over four experiments, to determine the optimal feature-weighting method for such twin-systems. Twin-systems for traditional multilayer perceptron (MLP) networks (MLP–CBR twins), convolutional neural networks (CNNs; CNN–CBR twins), and transformers for NLP (BERT–CBR twins) are examined. In addition, Feature Activation Maps (FAMs) are explored to enhance explainability by providing an additional layer of explanatory insight. The wider implications of this research on XAI is discussed, and a code library is provided to ease replicability.

The file attached to this record is the author's final peer reviewed version. The Publisher's final version can be found by following the DOI link. Achieving balance between convergence and diversity is a challenge in many-objective optimization problems (MaOPs). Many-objective evolutionary algorithms (MaOEAs) based on dominance and decomposition have been developed successfully for solving partial MaOPs. However, when the optimization problem has a complicated Pareto front (PF), these algorithms show poor versatility in MaOPs. To address this challenge, this paper proposes a co-guided evolutionary algorithm by combining the merits of dominance and decomposition. An elitism mechanism based on cascading sort is exploited to balance the convergence and diversity of the evolutionary process. At the same time, a reference point adaptation method is designed to adapt to different PFs. The performance of our proposed method is validated and compared with seven state-of-the-art algorithms on 200 instances of 27 widely employed benchmark problems. Experimental results fully demonstrate the superiority and versatility of our proposed method on MaOPs with regular and irregular PFs.

Subgraph matching is one fundamental issue in the research area of graph analysis, which has a wild range of applications, including question answering, semantic search and community detection. Recent studies have designed some near-optimal matching orders and data indexes to reduce the unpromising redundant calculations. However, the influences of recalculation on some positive candidate vertices were ignored in the iterative process of subgraph matching. In this paper, the novel concepts of an anchored node pair and anchored relationship are proposed as the theoretical basis to solve recalculation on subgraph matching. Thus, two key aspects are considered to reduce the redundant and recalculated node pairs. The first aspect involves using the dominating relationship of the anchored node and its follower to prune the negative node pairs, and an index of matching-driven flow graph is built to minimize the positive candidate vertices by using a heuristic algorithm. The second aspect involves exploring the anchored relationship to analyze the recalculated region of a matching stream, and two novel strategies are designed to manipulate the intermediate results of the partial subgraph isomorphism to avoid revalidation in the subgraph matching process. Extensive empirical studies on real and synthetic datasets demonstrate that our techniques outperform the state-of-the-art algorithms.

Person re-identification (ReID) suffers from camera view variants. Existing works, which typically learn a feature for each image, share a limitation that the learned features are single-view: each feature only contains information in one camera view. Thus, view bias occurs when matching pedestrians across camera views. In this paper, we seek to mitigate the view bias by generating multi-view features (fusion of features from a fixed number of cameras). To this end, we define the complementary-view features (complementary features to generate multi-view features with single-view features) and perform in-depth analysis. Based on this insight, we alleviate the view bias in testing and training, respectively. In testing, we present Multi-view Message Passing (MVMP), which generates multi-view features by aggregating single-view features from the neighborhood. In training, we propose Multi-view Feature Fusion Network (MFFN), which involves the single-view feature extractor and the complementary-view feature aggregator. MFFN makes the network sensitive to view-specific cues by adding constraints on multi-view features rather than single-view features. In addition, MVMP and MFFN have two key advantages: (1) They are parameter-free. (2) They can be applied to any Convolutional Neural Networks (CNNs) readily without extra supervision. Extensive experiments are conducted to validate the superiority of our method for person ReID over state-of-the-art methods on four benchmark datasets (Market-1501, DukeMTMC-reID, CUHK03, and MSMT17). The code is available at https://github.com/Yinsongxu/MVMP_MFFN .

Knowledge graphs are typical large-scale multi-relational structures and useful for many artificial intelligence tasks. However, knowledge graphs often have missing facts, which limits the development of downstream tasks. To refine the knowledge graphs, knowledge graph embedding models have been developed. Knowledge graph embedding models aim to learn distributed representations for entities and relations and predict unknown triplets by scoring candidate triplets. Nevertheless, state-of-the-art works either aim to capture different relation patterns, or to model the multi-fold relations, and yet fail to consider these two aspects simultaneously. To fill this gap, in this paper, we propose a novel knowledge graph embedding model, MRotatE. It exploits triplet features from the perspective of relational and entity rotations, which can model and infer various relation patterns and handle with multi-fold relations at the same time. The experimental results demonstrate that MRotatE outperforms existing approaches and attains the state-of-the-art performance.

Electromagnetic simulation tools have been playing an increasing role in the design of contemporary antenna structures. The employment of electromagnetic analysis ensures reliability of evaluating antenna characteristics but also incurs considerable computational expenses whenever massive simulations are involved (e.g., parametric optimization, uncertainty quantification). This high cost is the most serious bottleneck of simulation-driven design procedures, and may be troublesome even for local tuning of geometry parameters, let alone global optimization. On the one hand, globalized search is often necessary because the design problem might be multimodal (i.e., the objective function features multiple local optima) or a reasonably good initial design may not be available. On the other hand, the computational efficiency of popular algorithmic approaches, primarily, nature-inspired population-based algorithms, is generally poor. Combining metaheuristics procedures with surrogate modelling techniques and sequential sampling methods alleviates the problem to a certain extent but modelling of nonlinear antenna responses over broad frequency ranges is extremely challenging, and the aforementioned solutions are normally limited to rather simple structures described by a few parameters. This paper proposes a novel approach to global optimization of multi-band antennas. The major component of the presented framework is the knowledge-based inverse surrogate constructed at the level of response features (e.g., frequency and level locations of the antenna resonances). The surrogate facilitates decision-making process of inexpensive identification of the most promising regions of the parameter space, and a rendition of the good-quality initial design for further local tuning. Our methodology is validated using three examples of dual- and triple-band antennas. The average optimization cost is only 150 full-wave antenna analyses while ensuring precise allocation of the antenna resonances at the target frequencies. This performance is demonstrated superior over both local optimizers and population-based metaheuristics.

Predictive clustering trees (PCTs) are a well established generalization of standard decision trees, which can be used to solve a variety of predictive modeling tasks, including structured output prediction. Combining them into ensembles yields state-of-the-art performance. Furthermore, the ensembles of PCTs can be interpreted by calculating feature importance scores from the learned models. However, their learning time scales poorly with the dimensionality of the output space. This is often problematic, especially in (hierarchical) multi-label classification, where the output can consist of hundreds of potential labels. Also, learning of PCTs can not exploit the sparsity of data to improve the computational efficiency, which is common in both input (molecular fingerprints, bag of words representations) and output spaces (in multi-label classification, examples are often labeled with only a fraction of possible labels). In this paper, we propose oblique predictive clustering trees, capable of addressing these limitations. We design and implement two methods for learning oblique splits that contain linear combinations of features in the tests, hence a split corresponds to an arbitrary hyperplane in the input space. The methods are efficient for high dimensional data and capable of exploiting sparse data. We experimentally evaluate the proposed methods on 60 benchmark datasets for 6 predictive modeling tasks. The results of the experiments show that oblique predictive clustering trees achieve performance on-par with state-of-the-art methods and are orders of magnitude faster than standard PCTs. We also show that meaningful feature importance scores can be extracted from the models learned with the proposed methods., 26 pages, 11 figures. Submitted to Knowledge-based systems. Update: fixed typos, general grammar improvements, updated 2 references

One Class Slab Support Vector Machines (OCSSVM) have turned out to be better in terms of accuracy in certain classes of classification problems than the traditional SVMs and One Class SVMs or even other One class classifiers. This paper proposes fast training method for One Class Slab SVMs using an updated Sequential Minimal Optimization (SMO) which divides the multi variable optimization problem to smaller sub problems of size two that can then be solved analytically. The results indicate that this training method scales better to large sets of training data than other Quadratic Programming (QP) solvers., This version of the manuscript has been updated and is being reviewed by https://www.journals.elsevier.com/knowledge-based-systems