Search

Multi-view clustering (MVC) algorithms are prevalent in machine learning tasks due to their superiority in mining complementary information from multi-view data. Despite their widespread use, existing MVC approaches ignore the presence of inconsistencies in the multi-view features during multi-view data fusion, which leads to suboptimal clustering performance. To address this challenge, the paper presents a novel approach called multi-view clustering via latent consistency multi-graph fusion. Specifically, a novel latent representation learning model is devised to learn a unified representation from multi-view features, simultaneously capturing the high-order relationships within the multi-view data. Additionally, to alleviate the adverse effects caused by the inconsistency multi-view feature, a consistent multi-graph fusion module is proposed to provide consistent constraints for latent representations. To be specific, the graphs that reflect the manifold structure for each view are fused to guide a consistent latent representation learning. After that, an alternating optimization algorithm is designed to update variables iteratively and alternatively, among which the fusion weights are assigned adaptively. Finally, comprehensive experiments conducted on seven widely recognized benchmark multi-view datasets affirm the efficacy of the proposed method, showcasing its superiority over state-of-the-art approaches. [ABSTRACT FROM AUTHOR]

Unsupervised hashing has been extensively applied in large-scale multi-modal retrieval by mapping original data from heterogeneous modalities into unified binary codes. However, there still remain challenges especially how to balance the individual modality-specific representations and common representation preserving intrinsic linkages among heterogeneous modalities. In this paper, we propose a novel fast U nsupervised M ulti-modal H ashing based on P iecewise L earning, denoted as UMHPL , to deal with the mentioned issue. Initially, we formulate the problem as matrix factorization to derive the individual modality-specific latent representations and common latent representation with consensus matrices in a brief time. To maintain the integrality of multi-modal data, we integrate them by adaptive weight factors and nuclear norm minimization. Subsequently, we establish a connection between the individual modality-specific latent representations and common latent representation based on the piecewise hash learning framework to reinforce the discriminative competency of model, which leads the hash codes more compact. Finally, an effective discrete optimization algorithm in mathematical logic and functional analysis is proposed. Comprehensive experiments on Wiki, MIRFlirck, NUS-WIDE, and MSCOCO datasets demonstrate the superior performance of UMHPL to state-of-the-art hashing methods. [ABSTRACT FROM AUTHOR]

Federated learning has become increasingly popular in recent years and is widely applied to various fields of machine learning. However, the localized execution of federated learning lacks visibility of the training process to a third party (e.g., the model user or auditor), which raises the need to verify the training process. Privacy concerns over the public release of sensitive data for verification purposes are a prominent issue. Therefore, enabling a public verification of the training process without revealing sensitive data is a challenge. In this paper, we focus on verifiability and privacy in federated learning and propose a verifiable and privacy-preserving federated learning scheme (VPFL). We first employ zero-knowledge proofs to allow a third party to publicly verify the training process, which enhances the transparency of the training process and the reliability of the model. Then, to further protect the privacy of sensitive data, we exploit a commitment scheme to ensure that no information about sensitive data is leaked to a third party. We conduct extensive experiments to evaluate the performance of our scheme. For federated learning with 100 clients, our scheme only takes 13.1s to generate evidence and 8.8s to verify it. In addition, we compared our scheme with other schemes, and we observed that our scheme satisfies both the security properties of verifiability and privacy. [ABSTRACT FROM AUTHOR]

Recently, contrastive learning has been proven to be powerful in cross-domain feature learning and has been widely used in image translation tasks. However, these methods often overlook the differences between positive and negative samples regarding model optimization ability and treat them equally. This weakens the feature representation ability of the generative models. In this paper, we propose a novel image translation model based on asymmetric slack contrastive learning. We design a new contrastive loss asymmetrically by introducing a slack adjustment factor. Theoretical analysis shows that it can adaptively optimize and adjust according to different positive and negative samples and significantly improve optimization efficiency. In addition, to better preserve local structural relationships during image translation, we constructed a regional differential structural consistency correction block using differential vectors. Comparative experiments were conducted using seven existing methods on five datasets. The results indicate that our method can maintain structural consistency between cross-domain images at a deeper level. Furthermore, it is more effective in establishing real image-domain mapping relations, resulting in higher-quality images being generated. • Image translation tasks based on contrastive learning often ignore the difference between different positive/negative samples. • Our asymmetric slack contrastive learning adaptively improves the optimization efficiency of contrastive loss. • The local consistency can improve the global consistency of the generated image. • Preserving differential structural consistency can help eliminate local distortions in image translation. [ABSTRACT FROM AUTHOR]

The fusion technology, combining infrared and visible modes, has the potential to enhance the semantic content of backgrounds, thereby improving scene interpretability. However, most existing image fusion algorithms primarily concentrate on the fusion process, often neglecting the importance of preprocessing the source images to enhance their visual fidelity. Additionally, these algorithms frequently overlook the distinct characteristics of infrared and visible modes, leading to suboptimal weight allocations that do not correspond with human perception. To tackle these issues, this paper proposes a fusion algorithm that emphasizes visual fidelity and the rectification of salient information. More specifically, we improve fusion algorithms by designing an adaptive enhancement method based on Taylor approximation and visual compensation, which proves particularly effective in complex environments. Our proposed multi-scale decomposition approach extracts salient information from the transmission map, thereby enriching fusion results with finer details to accentuate target features. Drawing inspiration from the distinctive attributes of infrared and visible image modes, we devise a fusion weight calculation method grounded in similarity measurements to effectively convey significant information from the source images. To validate the effectiveness of our proposed method, we conducted validation experiments using publicly available datasets. Our experimental findings exhibit a prominent advantage over fifteen state-of-the-art fusion algorithms in both subjective and objective assessments. Our code is publicly available at: https://github.com/VCMHE/FVC_SIR. [ABSTRACT FROM AUTHOR]

In this paper, we propose a novel self-supervised Gaussian Restricted Boltzmann Machine with contrastive learning (CL-GRBM), which fuses contrastive representation learning and contrastive divergence to optimize and enhance the representation of GRBM. Built upon the concept of contrastive representation learning, CL-GRBM aims to enhance the representation capacity of the GRBM. Firstly, a pair of positive samples are constructed by one times Gibbs sampling with the original data. Then, the contrastive loss is used to pull the positive samples closer and push other samples further apart, making similar representations closer and different representations farther apart. In summary, during the training process of CL-GRBM using the CD algorithm, the objective is to align the sampling probability distribution of the visible layer in CL-GRBM as closely as possible with the empirical distribution of the original data. In the feature space of the sampling probabilities in the hidden layer, the distance between positive samples is minimized to capture the intrinsic structure of the data. According to the experimental verification, the proposed CL-GRBM shows better performance than contrastive models. As a shallow self-supervised model, it has even better performance than some excellent deep self-supervised models. [ABSTRACT FROM AUTHOR]

• Two optimization strategies for preventing over-fitting, ensuring reasonable rendering for arbitrary novel view synthesis and robust depth prediction. • An effective IBVH-based initialization approach for directly optimizing explicit 3D structures of high-quality novel view synthesis. This initialization strategy aims to maximize the utilization of object grid resolution under resolution-limited situations and improve rendering quality in grid-based methods. • Successfully eliminate evident ghosting artifacts by directly optimizing 3D structure, yielding a more realistic global rendering. Directly optimizing radiance fields with explicit 3D primitives, such as voxels and 3D Gaussians, demonstrates impressive effectiveness in both training and rendering. However, optimizing explicit 3D structures from limited training views using differentiable volumetric rendering often results in ghosting artifacts (floaters), i.e., it suffers from early-stage errors in unseen space and remains as evident ghosting artifacts as gradient becomes very sparse spatially with optimization procedure processes. Thus, to address this problem, this paper proposes Clear-Plenoxels, which involves three simple yet non-trivial components: 1) a Visual Hull-based initialization strategy for maximizing the utilization of object resolution and effectively rejecting false updates during early-stage optimization by learning rate assignment per voxel grid; 2) an effective penalty function on local grid discrepancy for ensuring consistent rendering at different view directions, which can remove unnecessary voxels; 3) a mask guided transmittance supervision for each training ray, significantly guaranteeing depth prediction precision when the object's surface color is close to the background. Experiments on a public dataset demonstrate that our method successfully overcomes ghosting artifacts by directly optimizing explicit primitive strategy. The proposed method achieves approximately 1 dB improvement in voxel grid-based methods and matches, if not surpasses, state-of-the-art quality regarding novel view rendering. The code and visualization videos are publicly available at https://github.com/nortonBryan/Clear-plenoxels. [ABSTRACT FROM AUTHOR]

The Set-Union Knapsack Problem (SUKP) is a complex combinatorial optimisation problem with applications in resource allocation, portfolio selection, and logistics. This paper presents a parallel local search algorithm for solving SUKP on the Compute Unified Device Architecture (CUDA) platform in Graphics Processing Units (GPUs). The proposed method employs a compact algorithm that divides the search space into smaller regions. For diversity, each thread in a GPU block starts the search process from a different location in a region using a different initial solution. Each thread then searches the local optimum by utilising communication between individuals through a crossover operator exploiting the best solution within the GPU block. Through extensive experiments on a set of SUKP benchmark instances ranging in size from small to large, we demonstrate the effectiveness of the proposed algorithm in finding high-quality solutions within comparable time frames. Furthermore, a comparative performance analysis with the current state-of-the-art SUKP algorithms reveals the competitive advantage of the proposed method. The GPU-based parallel local search algorithm using uniform crossover is a valuable addition to the repertoire of algorithms addressing SUKP, highlighting its potential for practical applications in real-world decision-making scenarios. [ABSTRACT FROM AUTHOR]

Fuzzy Cognitive Maps (FCMs) hold promise as a mathematical tool for modeling and simulating complex systems due to their transparency, flexibility to operate on prior knowledge structures and recurrent reasoning characteristics. However, they suffer from significant shortcomings that have prevented them from being more widely used. Some of these issues include discrepancies in component interpretation, saturation of neural concepts, arbitrary nonlinearities, and dynamic behaviors that are difficult to align with the problem domain. By integrating theoretical advances with practical needs, this paper proposes a revised modeling and simulation methodology termed " neural cognitive mapping " that addresses these issues holistically. Firstly, we redefine concepts' activation values in terms of changes rather than absolute values, ensuring a unified interpretation of the model's components. Secondly, we propose a parameterized activation function, called " exponential normalized activator ", which allows experts to control the neurons' nonlinearities while avoiding saturation states. Furthermore, we provide a twofold reasoning rule that simultaneously computes the concepts' changes and the amounts of resources attached to problem variables. Thirdly, we introduce a framework for interpreting simulation results across various dynamic behaviors, including scenarios with unique fixed-point attractors. The simulations using both real-world case studies and synthetically generated data illustrate the superiority of our proposal compared with the traditional approach in terms of clarity, usefulness, consistency, and controllability. Moreover, the empirical studies opened new research directions to be explored in future research. • " Neural cognitive mapping " methodology for modeling and simulating complex systems. • Addresses interpretation, nonlinearity, and dynamic behavior issues of FCM. • Introduces " exponential normalized function " to control the model's nonlinearity. • Systematic framework for interpreting simulation results across dynamic behaviors. [ABSTRACT FROM AUTHOR]

In this paper, we propose a pioneering approach for blind Hyperspectral Image (HSI) unmixing named Multi-Features Graph Deep Fusion Learning Networks for HSI Unmixing (MF-GDL). Our method leverages the power of multi-feature graph deep fusion learning networks. MF-GDL integrates spectral HSI graph features with complementary spatial graph features, overcoming the fixed window size limitation observed in prior deep learning approaches in capturing spatial relationships effectively. A key contribution lies in the introduction of multi-view graph learning, a novel aspect in HSI unmixing. By incorporating graph similarity-based spatial information, we address an important gap in the existing literature, marking the first attempt, to our knowledge, to tackle this challenge using a graph deep learning method. Furthermore, we introduce auxiliary information based on Extended Morphological Profiles (EMPs) to enrich spatial context beyond high-level spatial relationships between neighboring pixels. This multi-view fusion learning not only improves accuracy in material abundance estimation but also demonstrates the significance of considering multiple perspectives in HSI unmixing. We validate the efficacy of MF-GDL on four real datasets: Jasper Ridge, Samson, Urban, and Apex showcasing its robust performance and highlighting the substantial contribution of the multi-view aspect in advancing HSI unmixing techniques. • A multi-view graph neural networks is proposed for hyperspectral image unmixing. • Spectral- and Morphological Profiles-based graphs are built with relevant features. • Unmixing accuracy is improved by the learned shared latent representation. • Superior unmixing performance demonstrated on real HSI datasets. [ABSTRACT FROM AUTHOR]

Benefitting from the development of pyramidal feature learning, current state-of-the-art multi-scale detection paradigm has become proficient in detecting objects of varying scales. However, feature pyramid network (FPN), in spite of constructing multi-scale features with strong semantics, still suffers from limited performance caused by insufficient detail exploitation, information loss, limited receptive fields and hard proposal assignment, which can be mainly categorized into semantic level and instance level. To address these limitations, this paper analyzes the structural components that inhibit multi-scale feature representation and then presents a multi-stage progressive FPN (ProFPN) along with a novel RoI feature representation method called soft proposal assignment. In the semantic level, the bottom-up interaction module is first proposed to address to insufficient exploitation of high resolution features. In the bottom-up interaction module, global context attention blocks are utilized to interact adjacent-level features with detail information in a bottom-up progressive manner. After that, the top-down transfer module is designed to mitigate semantic information loss of high-level features. In the top-down transfer module, multi-branch asymmetric dilated blocks are adopted in a top-down progressive manner, which expands receptive fields to capture more object poses. In the instance level, to overcome the hard assignment of object proposals, a nonparametric strategy named soft proposal assignment is proposed to leverage the scale of each object proposal to generate dynamic weights for RoI features from adjacent levels. Comprehensive experiments conducted on MS COCO dataset demonstrate the superiority of ProFPN. By adding negligible extra FLOPs, the proposed ProFPN outperforms most pyramid-based methods. Moreover, due to the design of inherited feature utilization in ProFPN, transformer-based detectors have witnessed a substantial increase in detecting small objects while simultaneously achieving significant reductions in FLOPs. The source code of the proposed method is available at https://github.com/GingerCohle/ProFPN. [Display omitted] • SPA leverages the scale of object proposals to generate RoI features. • Without adding computational costs, SPA boosts several detectors by 0.3 AP ∼ 2.1 AP. • ProFPN combines regional contextual details and semantic information in a progressive manner. • ProFPN boosts transformer-based detectors while reducing FLOPs. [ABSTRACT FROM AUTHOR]

Transfer learning-based methods hold promise for enhancing classification task performance. However, a transfer learning mechanism for hard-to-classify time series classification tasks caused by limited samples in training set is still to be accomplished. Drawing inspiration from domain adaptation and style transfer methods in computer vision, we aim to address sample scarcity and enhance classifier generalization by leveraging abundant unrelated time series datasets. In this paper, a transfer learning mechanism called feature-level style transfer, where the feature alignment is conducted in a domain adaptation-inspired manner, has been proposed to target initially hard-to-classify time series tasks with limited training samples. The proposed mechanism eliminates constraints on source dataset selection, enabling the utilization of weakly-related or unrelated datasets to enhance classifier performance. Furthermore, a voting mechanism has been formulated to achieve more accurate predictions by taking transferable information learnt from different source domains into comprehensive consideration. • The TSC transfer learning using weakly-related datasets has been achieved. • Focusing on TSC datasets with limited training samples. • Accommodating the difference in label sets, lengths and channel numbers. • A voting mechanism has been proposed for multi-source transfer learning. [ABSTRACT FROM AUTHOR]

Convolutional Neural Networks (CNNs), particularly UNet, have become prevalent in medical image segmentation tasks. However, CNNs inherently struggle to capture global dependencies owing to their intrinsic localities. Although Transformers have shown superior performance in modeling global dependencies, they encounter the challenges of high model complexity and dependencies on large-scale pre-trained models. Furthermore, the current attention mechanisms of Transformers only consider single-scale feature interactions, making it difficult to analyze feature correlations at different scales in the same attention layer. In this paper, we propose DMSA-UNet, which strengthens the global analysis capability and maximally preserves the local inductive bias capability while maintaining low model complexity. Specifically, we reformulate vanilla self-attention as efficient Dual Multi-Scale Attention (DMSA) that captures multi-scale-enhanced global information along both spatial and channel dimensions with linear complexity and pixel granularity. We also introduce a context-gated linear unit in DMSA for each feature to obtain adaptive attention based on neighboring contexts. To preserve the convolutional properties, DMSAs are inserted directly between the UNet's convolutional blocks rather than replacing them. Because DMSA has multi-scale adaptive aggregation capability, the deepest convolutional block of UNet is removed to mitigate the noise interference caused by fixed convolutional kernels with large receptive fields. We further leverage efficient convolution to reduce computational redundancy. DMSA-UNet is highly competitive in terms of model complexity, with 33% fewer parameters and 15% fewer FLOPs (at 22 4 2 resolution) than UNet. Extensive experimental results on four different medical datasets demonstrate that DMSA-UNet outperforms other state-of-the-art approaches without any pre-trained models. • Our method combines CNNs and Transformers for medical image segmentation. • An efficient Dual Multi-Scale Attention (DMSA) module is proposed. • Our method is highly competitive in terms of model complexity. • Our method outperforms other SOTA approaches without any pre-trained models. [ABSTRACT FROM AUTHOR]

This paper proposes a new neural-based algorithm applied to online identifying, observing, and adaptively controlling of uncertain nonlinear systems via which the system transient and residual values independently regulated. The Lyapunov stability concept is used for enhancing the control performance. Especially, distinguished from recent related studies, not only identified model but the learning method and control rules are also innovatively implemented through which it independently separates the transient-time regulation from residual error adjustment. Sequentially, an advanced adaptive neural controller for an uncertain nonlinear plant with state feedback is implemented using the open-loop estimation result. The key contribution of the online estimator will be verified in this proposed control algorithm. Eventually, as to verify the prescribed performance and the flexibility of the suggested control approach, the estimation and control of an uncertain hyper-chaotic weld plant will be tested. [ABSTRACT FROM AUTHOR]

The classic problems in itemset mining involve finding frequent itemsets and high-utility itemsets. However, frequent itemset mining has the disadvantage of not paying attention to the profit of products, while high-utility itemset mining does not address the issue of the cost price of the products. Therefore, neither can locate products with high-efficiency value on investment. To overcome these problems, the high-efficiency itemset mining (HEIM) problem was proposed. Despite its practicality, this issue has received little attention. The algorithms proposed to exploit high-efficiency itemsets (HEI) still use ineffective strategies on dense databases and unstrict upper bounds, requiring a lot of time and memory. To address the current issues with HEIM, the paper proposes tight upper bounds for the early pruning of candidates. Several techniques are also proposed, such as combining similar transactions and saving promising transaction locations, to reduce the cost of database scanning. Finally, the techniques are combined to propose a novel way to implement the MHEI (an efficient strategy for Mining High-Efficiency Itemsets in quantitative databases) to optimize the HEIM process. The experimental process also shows that the proposed algorithm has performance better than the state-of-the-art algorithm in HEIM. [ABSTRACT FROM AUTHOR]

Internet of Things (IoT) technology increases connectivity between devices; however, IoT device networks are less secure, making them highly susceptible to attack. Most of the network security defense techniques are passive defense with inevitable delays. For instance, in the case of intrusion detection systems, during the time interval between detecting an attack and implementing defenses, the attacker may inflict damage on the target. Therefore, the adoption of active defense techniques becomes particularly important. Network Security Situation Prediction (NSSP) is an active defense technology capable of predicting future states of network security. Currently, most methods in the NSSP field focus on single-step prediction, while research related to multi-step prediction is scarce. Hence, this paper presents an NSSP method based on the PatchTST framework(PatchesNet), which exhibits good multi-step predictive performance. The method uses Variational Mode Decomposition (VMD) to deal with sequence instability. Multi-scale Patch delivery (MSP), Multi-scale Second Decomposition (MSD), and Period Embedding (PeE) are used to obtain more sequence features. Correlation Fusion (CF) can reduce the noise effect on prediction. Experiments are conducted on three datasets. The results demonstrate that PatchesNet exhibits favorable performance, relatively reducing the MSE and MAE by 39.243% and 26.033%, respectively, compared to the previous state-of-the-art method. [ABSTRACT FROM AUTHOR]

In real-world scenarios, multiple subtasks of legal judgment (such as law article, charge, and term of penalty) have strict task logical order and label topological relation, and their results influence and validate each other. However, most existing methods model them as simple classification problems, which ignores the logical and semantic constraints between subtasks. Besides, they mainly focus on the fact description for judgment results, and ignore the standard legal documents (i.e., the established law articles). To this end, we propose a H ierarchical D ependency-based L egal J udgment P rediction framework (HD-LJP), which integrates task judicial logic, label topological relation, and hierarchical semantics knowledge in legal text effectively. Specifically, HD-LJP employs consistency and distinction distillation to model label topological relation among multiple subtasks, and improve the differentiation of each subtask itself respectively. In addition, for simulating the judicial logic of human judges, we define logical dependencies between subtasks, and then utilize the results of intermediate subtasks to make auxiliary prediction of other subtasks. And, hierarchical semantics knowledge is fully integrated and applied in these two processes, which will profoundly affect the creditability and interpretability of the judgment results. The experimental results show that HD-LJP can improve the prediction performance of three LJP subtasks, especially in the term of penalty. Compared with the existing methods, the macro-F1 on CAIL-small is increased by 13.4%, and 6.9% on CAIL-big. In addition, through further case studies, this paper demonstrates that HD-LJP performs better for tail classes and confusing labels than the current SOTA R-former. • A hierarchical LJP framework for co-embedding four auxiliary prediction information. • Applying legal provision knowledge with hierarchical dependencies into LJP. • Employing relation distillation to model the label dependence among multiple subtasks. • Adopting case attention mechanism to simulate the task judicial logic of human judges. [ABSTRACT FROM AUTHOR]

The main contribution of this paper is introduction of descriptors based on the moments of the Bernstein–Durrmeyer polynomials. As it is demonstrated, their distinctive feature is their ability to preserve knowledge about the shapes of the signals such as monotonicity and convexity, without imposing any a priori constraints. These descriptors act in time-domain as shape-preserving knowledge extractors. For their empirical version, the asymptotic unbiasedness and MSE consistency are also proved without additional pre-filtering. Another significant contribution is the method of designing an autoencoder for selecting the number of descriptors suitable for a whole class of underlying signals. For this purpose, a generalized version of Akaike's information criterion (AIC) is derived, and its approximate version is proposed and tested. It serves as the main nonlinear part of the autoencoder. Its linear part computes the descriptors, while the decoder part computes the AIC for all particular signals from a considered family. To demonstrate the usefulness of the proposed methodology, we consider the problem of classifying signals based on knowledge about their shapes gathered by the descriptors. First, a general scheme of selecting an appropriate classifier from a predefined list is proposed. Then, the proposed approach is applied to classifying vibrations of a large excavator in an open pit mine. • Signal descriptors, preserving monotonicity and convexity, are described. • The estimates are shown to be consistent in the mean squared error sense. • An algorithm for designing autoencoder is proposed. [ABSTRACT FROM AUTHOR]

To solve the problem of target discrimination being ignored in developing the feature of the current frame, this paper proposes the effective feature representation via memory (EFRM) method to form the effective and discriminative feature representation of the current frame from global and local perspectives by fully benefiting from the rich information contained in the memorized frames. First, the global discriminative feature, representing the differences between the foreground and background, is generated through the space–time memory read (STMR) with the nonlocal matching scheme. Second, the local discriminative feature, which has feature differences among targets that appear in the current frame, is generated by the designed per-object memory enhancement (PoME) by relying only on the diverse representations of the targets shown in the memorized frames. Finally, the segmentation of the current frame is generated based on the effective feature representation formed by the concatenation of global and local discriminative features. Evaluations on DAVIS 16, 17 and YouTube-VOS 18, 19 demonstrate the competitive performance of the proposed method. • This study proposes to form an effective feature representation of the current frame through global and local discriminative perspectives. • PoME with reverse matching is designed in the local discriminative perspective to enhance the feature discrimination of the targets that appear in the current frame. • QaWA is integrated into PoME to score the memorized frames according to their masks' qualities; then, PoME utilizes these scores to form a more desirable local discriminative feature. [ABSTRACT FROM AUTHOR]

Bitcoin price volatility fascinates both researchers and investors, studying features that influence its movement. This paper expends on previous research and examines time series data of various exogenous and endogenous factors: Bitcoin, Ethereum, S&P 500, and VIX closing prices; exchange rates of the Euro and GPB to USD; and the number of Bitcoin-related tweets per day. A period of three years (from September 2019 to September 2022) is covered by the research dataset. A two-layer framework is introduced tasked with accurately forecasting Bitcoin price. In the first layer, to account for complexities in the analyzed data, variational mode decomposition (VMD) extracts trends from the time series. In the second layer, Long short-term memory and hybrid Bidirectional long short-term memory networks were used to forecast prices several steps ahead. This work also introduced an enhanced variant of the sine cosine algorithm to tune the control parameters of VMD and both neural networks for attaining the best possible performance. The main focus is on combining VMD with modified metaheuristics to improve cryptocurrency closing value forecast. Two sets of experiments were conducted, with and without VMD. The results have been contrasted with models tuned by seven other cutting-edge optimizers. Extensive experimental outcomes indicate that Bitcoin price can be forecasted with great accuracy using selected features and time series decomposition. Additionally, the best model was analyzed, and Shapley values indicated that features such as EUR/USD exchange rates, Ethereum closing prices, and GBP/USD exchange rates, have a significant impact on forecasts. • VMD, LSTM and BiLSMT are used in this study to predict Bitcoin price. • Employed enhanced metaheuristics to improve VMD, LSTM and BiLSTM model precision. • SHAP analysis was conducted to explore variable effects. • An efficient price forecasting tool is offered for investors and traders. • Provides valuable contributions to digital currency traders and investors. [ABSTRACT FROM AUTHOR]

Few-shot learning (FSL) methods have attracted great attention in recent years. However, under the cross-domain few-shot learning (CDFSL) setting, where domain gaps exist between the source and target domains, many FSL methods face performance degradation. One possible reason is the well-trained model can easily learn discriminative features on the source domain but may fail on target domains. To alleviate this problem, it is desirable to learn a model that can extract useful general features from target tasks regardless of domain gaps during the source domain training. In this paper, we propose a dynamic representation enhancement (DRE) framework that maximizes the model's representation ability to learn diverse and meaningful features on unknown domains. First, we proposed a stage-wise cycling and warm-up feature activation training strategy to make the model continuously explore useful general features during training, instead of just satisfying with fitting well on the limited training data. Second, we designed a task-adaptive loss function to enhance the learning of diverse features by making the model focus on inter-class discriminative features during training. Experimental analysis shows that the proposed DRE framework effectively extracts informative features across various target domains. On the popular cross-domain BSCD-FSL benchmark, the performance of the DRE framework is competitive with the state-of-the-art methods with an average classification accuracy of 51.01%, 66.71% and 74.13% in the 5-way 1-shot, 5-way 5-shot and 5-way 20-shot settings, respectively. Our code is publicly available at https://github.com/ideal-123/Dynamic-Representation-Enhancement-framework. [ABSTRACT FROM AUTHOR]

With the rapid development of multimedia technologies, the efficient retrieval of large-scale multimedia information is regarded as a critical area to research. Hashing methods have achieved superiority as an effective solution for multimedia retrieval while offering the advantages of reduced storage demands and faster retrieval speeds. However, traditional hashing methods face challenges in addressing cross-modal heterogeneity and associating modal features with label information. In this paper, we introduce a novel supervised cross-modal hashing method named Supervised Semantic-Embedded Hashing (SSEH) for Multimedia Retrieval. Firstly, we designed a Semantic-Enhanced Representation module based on label mapping, which achieved deep integration of label information with multimedia features to ensure the completeness of semantic information in the hash codes. Secondly, a Class Structure Preservation module was constructed to comprehensively extract precise semantic information and association relationships from the labels, thereby ensuring the accuracy of semantic information and association relationships in the hash codes. We conducted extensive experiments on widely recognized datasets to demonstrate our SSEH method's efficacy and robustness. The code for our experiments is available at https://github.com/YunfeiChenMY/SSEH. • Proposed a novel supervised semantic-embedding hashing method. • Designed a Semantic-Enhanced Representation module. • Proposed a Class Structure Preservation to extract the semantic relationships. • Evaluated the effectiveness of the proposed method via three famous datasets. [ABSTRACT FROM AUTHOR]

Existing few-shot segmentation (FSS) only considers learning support-query correlation and segmenting unseen categories under the precise pixel masks. However, the cost of a large number of pixel masks during training is expensive. This paper considers a more challenging scenario, weakly-supervised few-shot segmentation (WS-FSS), which only provides category (i. e. image-level) labels. It requires the model to learn robust support-query information when the generated mask is inaccurate. In this work, we design a Correlation Enhancement Network (CORENet) with foundation model, which utilizes multi-information guidance to learn robust correlation. Specifically, correlation-guided transformer (CGT) utilizes self-supervised ViT tokens to learn robust correlation from both local and global perspectives. From the perspective of semantic categories, the class-guided module (CGM) guides the model to locate valuable correlations through the pre-trained CLIP. Finally, the embedding-guided module (EGM) implicitly guides the model to supplement the inevitable information loss during the correlation learning by the original appearance embedding and finally generates the query mask. Extensive experiments on PASCAL-5 i and COCO-20 i have shown that CORENet exhibits excellent performance compared to existing methods. • We propose a Correlation Enhancement Network with foundation model assistance to accurately segment query samples in weakly-supervised few-shot segmentation (WS-FSS). • We propose a Correlation-Guided Transformer, a Class-Guided Module and an Embedding-Guided Module to mine target information in correlation features. • Our CORENet achieved state-of-the-art results in two WS-FSS scenarios (i.e. PASCAL-5 i and COCO-20 i). [ABSTRACT FROM AUTHOR]

Long-term time series forecasting is widely used in various real-world applications, such as weather, traffic, energy, healthcare, etc. Recently, time series decomposition techniques have been adopted in many mainstream forecasting models, such as the prevalent Transformer-based models, to help capture sophisticated temporal patterns and achieve success in several benchmark tasks. However, conventional decomposition algorithms are often based on simple operations or limited to specific fields, and therefore are not effective and applicable enough, especially for complex time series. In this paper, we propose Mode Decomposition and 2D Convolutional Network (MDCNet), a structure-simple yet effective forecasting architecture based on a more effective decomposition method and a multi-frequency time series feature extraction network with multi-scale 2D convolution. Specifically, we first introduce a Variational Mode Decomposition Block to discover intricate time patterns, which decompose time series into trend components and stationary modal components at different main frequencies. Then, we design a Trend Prediction Block and an Intrinsic Mode Functions Prediction Block to capture global correlation and hidden dependencies within different main frequencies, respectively. Furthermore, a Frequency Enhancement Module is designed to further reduce the impact of noise in long-term time series. Experiments on eight benchmark datasets show that MDCNet significantly reduces the error of the previous state-of-the-art method by 15.1% and 11.5% for multivariate and univariate time series, respectively. • A simple, efficient and robust time series forecasting network, MDCNet, is proposed. • The mode decomposition block is designed to decompose complex time series smoothly. • MDCNet uses 2D convolution to extract inter-pattern and intra-pattern dependencies. • MDCNet outperforms SOTA methods and achieves 15.1% relative performance improvement. [ABSTRACT FROM AUTHOR]

The rapid development and increasing evolution of malware necessitate novel defensive techniques with high accuracy and minimal false positives to safeguard information systems against potential threats. Unfortunately, current malware detection methods, primarily relying on deep learning to identify malicious fingerprints from extensive training sets, prove ineffective against few-shot and evolving malware variants. To address these challenges, this paper introduces MalIRL, which designs a model-free inverse reinforcement learning (IRL) mechanism to automatically capture the evolving attack intent of malware. Specifically, MalIRL explores six representative categories of malware actions and employs sliding windows to organically divide the massive malware execution event stream into multiple attack stages, achieving a reduced action space and state space. To model dynamic malicious environments, MalIRL proposes an instant dynamic heterogeneous graph representation learning technique. This technique learns state representations of different malware attack stages, enhancing detection accuracy and efficiency by incrementally capturing the newly added contextual semantics of diverse malware entities and relations. In experiments with three real-world malware datasets, MalIRL surpasses existing state-of-the-art methods, particularly in few-shot malware detection scenarios, MalIRL exhibits performance benefits of up to 18.9%. [ABSTRACT FROM AUTHOR]

Video-based person re-identification (Re -ID) aims to retrieve the target person from video sequences captured by a distributed camera system. It remains a challenging task due to the reasons such as occlusion and misalignment in the video. To address above problem, many methods are proposed to exploit multi-scale spatio-temporal features in videos. However, established methods typically assign the equal weights to temporal or spatial features at different scales, which significantly diminishes the distinct roles of each feature. In this paper, we propose a novel Multi-scale Feature Aggregation Network (MFANet) for video-based person Re -ID. Specifically, we propose two flexible modules, Multi-scale Temporal Feature Aggregation (MTFA) and Multi-scale Spatial Feature Aggregation (MSFA). These two modules first extract different scales of temporal (dynamic, static) and spatial (coarse and fine) features, and then adaptively assign weights to each feature according to the video sequence. Both of these lightweight modules can be incorporated with 3D Convolutional Neural Network to build our MFANet. Extensive experiments on four public benchmarks demonstrate that MTFA and MSFA improve the performance of baseline architectures, and our MFANet achieves the best performance compared to other state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Federated learning (FL) has been used as a promising approach to breaking the dilemma between the data privacy and the learning from large collections of distributed data. Without sharing data, the server trains a global model with high accuracy only by accessing the gradient of each client. However, recent research has revealed that FL is vulnerable to gradient inversion attacks, where the server can reconstruct clients' training data by inverting their uploaded gradients. Current defense methods share the common limitation that they need to either perturb the training data or the gradients. However, this perturbation will inevitably reduce the accuracy of the global model. To overcome these limitations, this paper proposes a defense method by allowing clients to train local models without using training data but rather using the statistical information of the data. Inspired by the statistical machine unlearning method, which uses statistical information to train models, combined with knowledge distillation to lightweight local models, our method further isolates the relationship between gradients and training data by transferring the model to a simpler model. Knowledge distillation typically involves extracting knowledge from a teacher model and transferring it to a student model without the need for the original training data. With this method, on the one hand, the server can reconstruct only the statistical information of the student model data, which is semantically meaningless to humans. On the other hand, using the statistical information of the data to train the global model can guarantee the model's accuracy. Experimental results show that our method outperforms existing defense methods in various aspects, including reconstruction accuracy, model accuracy and training efficiency. [ABSTRACT FROM AUTHOR]

Bayesian optimization is a popular framework for the optimization of black box functions. Multifidelity methods allow to accelerate Bayesian optimization by exploiting low-fidelity representations of expensive objective functions. Popular multifidelity Bayesian strategies rely on sampling policies that account for the immediate reward obtained evaluating the objective function for a specific input, precluding greater informative gains that might be obtained looking ahead more steps. This paper proposes a Non-Myopic Multifidelity Bayesian Optimization framework (NM2-BO) to grasp the long-term reward from future steps of the optimization. Our computational strategy comes with a two-step lookahead multifidelity acquisition function that maximizes the cumulative reward obtained measuring the improvement of the solution over two steps ahead. Our NM2-BO algorithm is demonstrated for a large set of benchmark problems to stress test across a broad spectrum of mathematical properties, and a physics-based application relevant for the engineering community. In all the cases, we observe that the proposed algorithm outperforms standard MFBO frameworks. [ABSTRACT FROM AUTHOR]

Weakly Supervised Video Anomaly Detection (WSVAD) plays a crucial role in visual surveillance by effectively distinguishing anomalies from normality with only video-level annotations; nevertheless, due to inherent limitations, including imbalanced data, intra-bag similarity , and snippet entanglement , existing methodologies prone to a bias towards normality. This paper proposes a novel framework, Visual Prompt Exemplars-based WSVAD termed VPE-WSVAD, to enhance discriminative representations of anomalies against normality by incorporating scenario-awareness visual prompt exemplars. The proposed VPE-WSVAD framework comprises three key components. Firstly, the prompt retrieval network generates potential abnormal proposals consistent with visual prompt exemplars at the frame level. Secondly, the proposal filtering mechanism aims to select the proposal with the highest confidence score to represent potential anomalies in the snippet. Finally, prompt likelihood learning is designed to capture the correlations between proposals (w.r.t prompt exemplars) and snippets, generating discriminative representations for each snippet. By incorporating visual prompt exemplars, our method provides more detailed reporting of abnormal events, including when, where, and what. To validate the efficacy of our method, we conducted comprehensive evaluations on three publicly available datasets. The experimental results conclusively demonstrate the superiority of our approach, achieving an impressive frame-level area under the curve (AUC) of 96.88% on the ShanghaiTech dataset and 99.86% on the UCSD Ped2 dataset. Our code will be released in the future. [ABSTRACT FROM AUTHOR]

Node representation learning plays a crucial role in addressing entity-related tasks in a relational network. However, current methods for node representation learning based on graph neural networks often rely on a layer-by-layer perception strategy. This approach compresses node features during propagation, hindering the direct perception of essential information and leading to over-smoothing issues. Moreover, many methods lack a comprehensive consideration of network topology and node semantics, limiting the model's performance. In this paper, we break through the traditional layer-by-layer perception strategy and propose TSCNet, a node representation learning method founded on direct perception strategy and topological and semantic collaborative mining. Initially, TSCNet establishes direct connections between the target node and sampled neighbor nodes at different orders, facilitating the target node directly perceiving neighborhood information. Subsequently, TSCNet introduces static learnable parameters to weigh different-order topologies and utilizes attention mechanisms to weigh the semantic relevance between nodes. Finally, based on the direct connections graph, TSCNet aggregates neighborhood information according to topology and semantic weights, achieving task-adaptive topology and semantic co-mining. We conduct extensive experiments on ten real-world datasets. The model achieves top-ranking results in eight datasets in which the node classification accuracy reaches 41.27% on the Actor dataset and 94.78% on the Texas dataset, significantly outperforming mainstream models. The results demonstrate the effectiveness of the proposed TSCNet. [ABSTRACT FROM AUTHOR]

The feature selection problem in multi-view data has garnered widespread attention and research in recent years, leading to the development of numerous feature selection algorithms tailored for multi-view data. However, existing methods often focus solely on known data, overlooking the potential distribution information of unknown data. Additionally, these methods inevitably introduce a large number of parameters to fully utilize the information from different views in multi-view data, thereby reducing the efficiency of model training. To address these issues comprehensively, we propose a novel framework called Multi-view Stable Feature Selection with Adaptive Optimization of View Weights (MvSFS-AOW). Specifically, the framework first employs the Multi-view Stable Feature Selection (MvSFS) algorithm to evaluate and select features from different views. Subsequently, it dynamically adjusts view weights using the Adaptive Optimization of View Weights (AOW) algorithm to achieve optimal generalization performance. By incorporating unknown data into the training process, we enhance the reliability of the framework in practical applications. Furthermore, our framework achieves competitive performance without requiring extensive parameter tuning. Experimental results demonstrate that the proposed framework achieves promising classification and clustering performance on multiple datasets, surpassing other state-of-the-art algorithms. Code for this paper available on: https://github.com/boredcui/MvSFS-AOW. [ABSTRACT FROM AUTHOR]

Pre-trained language models have demonstrated remarkable performance in few-shot learning through the emergence of "prompt-based learning" methods, where the performance of these tasks highly rely on the quality of prompts. Existing prompt learning methods typically customize a single prompt to each few-shot learning task and all the examples in the task share the universal prompt. However, a fine-grained prompt design can enhance the performance of few-shot learning task by leveraging more diverse information hidden in the set of examples. In light of this motivation, this paper introduce an example-specific prompt learning method to embody fine-grained self-adapting prompts for few-shot learning with pre-trained models. Specifically, we introduce the concept of the "weak consistency assumption", to trade-off the task-specific consistent and example-specific diversity. Based on this assumption, a novel method called Self-adapting Continuous Prompt Learning (SP-learning) to learn example-specific prompts is proposed. It employs a cross-attention prompt generator that considers the characteristics of input samples and utilizes a diversity calibration technique to adjust the prompt generator accordingly. By personalizing prompts for each example, SP-learning aims to improve few-shot learning performance. We perform a systematic evaluation on 10 public benchmark tasks and our method outperforms 8 of those tasks. Our research sheds light on the importance of personalized prompts and opens up new possibilities for improving few-shot learning tasks. • We present the weak consistency assumption that the prompts in a task are consistent in the whole task but diverse in individuals, which allows us to learn example-specific templates. • We propose to train a prompt generator with the sample as input to generate an example-specific prompt for each example, by generating a meaningful prompt token set and selecting appropriate prompt tokens to form a proper prompt for the input sample. • We propose the diversity calibration technique to force the prompts to be diverse on the individual examples. [ABSTRACT FROM AUTHOR]

Establishing high-quality correspondences is a fundamental step for many computer vision tasks. Leveraging the local consistency of correct correspondences (i.e., inliers) has been a prevalent approach to distinguish them from incorrect correspondences (i.e., outliers). However, the random distribution of a significant proportion of outliers complicates the local neighborhood construction of inliers, making it inevitably contain many outliers, which compromises the reliability of the consistency information. In this paper, we propose a Progressive Multi-Stage Adaptive Feature Learning Network (PMA-Net) to solve this problem, which progressively and adaptively learns over multiple stages to remove random outliers embedded in the initial consistency. Specifically, we propose an Adaptive Dynamic Graph Construction (ADGC) module to construct a global topological graph with a high inlier ratio by calculating the affinity between the corresponding neighbors. Additionally, we also design a Feature Mapping Processing (FMP) block and Multi-Stage Prediction (MSP) block to improve the accuracy of subsequent local neighborhood information aggregation, as well as information loss due to channel dimension compression in the prediction phase. Experimental results in camera pose estimation, remote sensing image registration, and homography estimation demonstrate that the proposed PMA-Net performs better than other state-of-the-art methods on various public datasets. Code: https://github.com/XiaojieLi11/PMA-Net [ABSTRACT FROM AUTHOR]

Noise feature modules play a prevalent role in image manipulation localization. However, existing approach involves utilizing a specific noise feature module tailored to address distinct categories of tampering procedures, limiting its adaptability across diverse scenarios. This limitation becomes particularly pronounced in image manipulation localization where a singular noise module struggles to comprehensively represent all potential tampering types, given the inherent uncertainty surrounding the nature of tampering. A novel Contribution-Aware Noise Feature (CANF) representation model, specifically designed to tackle the intricacies of image tamper noise, is introduced in this paper. The proposed model integrates existing noise feature modules, autonomously enhancing their capabilities. This enhancement not only contributes significantly to image manipulation localization during the inference phase but also enables a quantitative evaluation of the distinct contributions of various noise modules to the inference results. To optimize the utilization of image information, the Red–Green–Blue (RGB) and CANF data are fused, leading to the development of a standardized image manipulation localization architecture. This architecture exhibits heightened efficiency in harnessing diverse types of information, thereby enhancing the overall performance of image manipulation localization. Experimental results demonstrate that our approach, alongside established methods IF-OSN, MSFF, PCL, and NCL, surpasses the state-of-the-art NCL by 8.5% and 5.4% concerning average F1 and AUC metrics across four standard image manipulation localization datasets: NIST16, CASIAv1, Columbia, and Coverage. In addition, we comprehensively explore the impact of various noise characterization modules on different types of tampering and the fact that Late Fusion is more conducive to image tampering localization. • A contribution-aware noise feature representation (CANF) model for analyzing image tampering noise. • Incorporates CANF and RGB data to pinpoint exact tampered areas in images. • Method adaptable to various types of image tampering, versatile for forensics and authentication. • CANF effectively integrates various noise modules, enhancing their collective impact. • Comprehensive testing on standard datasets and challenging online scenario, outperforms state-of-the-art. [ABSTRACT FROM AUTHOR]

Given a graph G = (V , E) , the minimum dominating set (MinDS) problem is to identify a smallest set of vertices D such that every vertex in V ∖ D is adjacent to at least one vertex in D. The MinDS problem is a classic NP -hard problem and has been extensively studied because of its many disparate applications in network analysis. To solve this problem in practice, many heuristic approaches have been proposed to obtain a high-quality solution within a given time limit. However, existing heuristic algorithms are limited by various tie-breaking cases when selecting vertices, which slows down the effectiveness of the algorithms. In this paper, we design an efficient local search algorithm for MinDS, named DmDS – a dual-mode local search framework that probabilistically chooses between two distinct vertex-swapping schemes. We further address limitations of other algorithms by introducing vertex selection criterion based on the frequency of vertices added to solutions to address tie-breaking cases, and by improving the quality of the initial solution via a greedy strategy with perturbation. We evaluate DmDS against the state-of-the-art algorithms on real-world datasets, consisting of 382 instances (or families) with up to tens of millions of vertices. Experimental results show that DmDS obtains the best performance in accuracy for almost all instances and finds significantly better solutions than state-of-the-art MinDS algorithms on a broad range of large real-world graphs; specifically, DmDS computes the smallest solution on 352 (out of 382) instances, and on 119 instances DmDS finds smaller solutions than all other algorithms in our comparison. [ABSTRACT FROM AUTHOR]

Multi-modal emotion recognition in conversation is challenging because of the difficulty to jointly leverage the information from heterogeneous text, acoustic, and visual modalities. Recent context-aware methods usually design a graph structure to model dependencies of utterances and speakers, or integrate Multi-modal information. However, they typically lack a sufficient extraction of unimodal context, and rarely explore the emotion consensus prototypes among different samples with the same label. For solving these problems, in this paper, we propose a Graph Context extraction and Consensus-aware Learning (GCCL) framework to excavate context-sensitive fusion features and simulate the emotion evocation process during the emotion consensus learning. Specifically, GCCL contains a well-designed graph-based module to capture speaker, temporal and modality dependencies and integrate information from different modalities. Then, we design an emotion consensus learning unit to mine the most typical feature of each category in each modality. A speaker-guided contrastive learning loss is further proposed to guarantee the diversity between different individuals and the semantic consistency between distinct modalities. Moreover, we construct a consensus-aware unit with an attention-based memory mechanism to preserve semantic correlations among different samples on the category-level. Extensive experimental results on two conversational datasets demonstrate that the proposed GCCL outperforms the state-of-art methods. Code is available at https://github.com/gityider/GCCL. [ABSTRACT FROM AUTHOR]

• A Transformer based underwater image enhancement for vision-guided AUVs is proposed. • Construct a Dimension Decomposition Transformer to compute global dependencies. • Design a multi-stage Transformer strategy to extract multi-scale global information. • Build a semi-supervised learning framework to reduce the reliance on paired datasets. • Excellent performance is achieved in underwater visual enhancement engineering tasks. Vision-guided Autonomous Underwater Vehicles (AUVs) have gradually become significant tools for human exploration of the ocean. However, distorted images severely limit the visual ability, making it difficult to meet the needs of complex underwater environment perception. Fortunately, recent advancements in deep learning have led to rapid developments in underwater image enhancement. The emergence of the Transformer architecture has further enhanced the capabilities of deep learning. However, the direct application of Transformer to underwater image enhancement presents challenges in computing pixel-level global information and extracting local features. In this paper, we present a novel approach that merges dimension decomposition Transformer with semi-supervised learning for underwater image enhancement. To begin, dimension decomposition attention is proposed, which enables Transformer to compute global dependencies directly at the original scale and correct color distortions effectively. Concurrently, we employ convolutional neural networks to compensate for Transformer's limitations in extracting local features, thereby enriching details and textures. Subsequently, a multi-stage Transformer strategy is introduced to divide the network into high- and low-resolution stages for multi-scale global information extraction. It helps correct color distortions while enhancing the network's focus on regions with severe degradation. Moreover, we design a semi-supervised learning framework to reduce the reliance on paired datasets and construct a corresponding multi-scale fusion discriminator to enhance the sensitivity to input data. Experimental results demonstrate that our method outperforms state-of-the-art approaches, showcasing excellent learning and generalization capabilities on subjective perception and overall evaluation metrics. Furthermore, outstanding results highlight the significant improvements it brings to downstream visual engineering applications. The code of the proposed DDformer is available at https://github.com/ZhiGao-hfuu/DDformer. [ABSTRACT FROM AUTHOR]

Image watermarking technology poses a significant challenge, requiring a delicate balance between robustness, imperceptibility, and capacity. To solve the balancing problem, a statistical learning based blind image watermarking technique is proposed in this paper. The idea of the proposed method to solve the balancing problem is to enhance imperceptibility and robustness while accommodating sufficient watermarks. Firstly, the local low-order pseudo-Zernike moments (PZM) magnitude in the discrete non-separable Shearlet transform (DNST) domain is constructed as the embedding domain. To ensure stability in embedding positions, we propose an edge embedding strategy. Secondly, by analyzing the statistical characteristics and multi-correlations of DNST-PZM magnitudes, a multi-correlation vector model based on the skew student's-t mixture (SSM) and hidden Markov tree (HMT) is designed. Finally, leveraging the vector SSM-HMT model and the maximum likelihood (ML) criterion, we derive the closed-form decoder expression. Comparative analysis with state-of-the-art methods demonstrates the superior imperceptibility and robustness of our proposed approach in accommodating the same capacity watermark. [ABSTRACT FROM AUTHOR]

Community search over bipartite graphs is a crucial issue with significant applications in various domains, including group recommendations, fraud detection and representation learning. Current research primarily focuses on the structural cohesiveness constraint between two sets of vertices, often neglecting the community's size constraint. Unfortunately, this oversight may lead to communities of sizes ranging from excessively small to substantially large, thereby affecting their applicability. Therefore, in this paper, we focus on the maximal size constraint community search over bipartite graphs, which not only takes into account the (α , β) degree constraints for each set of vertices but also ensures that the number of vertices in each set is limited to ξ and ζ , respectively. To solve this problem, we first introduce two competitive algorithms, namely Ordering-based Expansion Algorithm (OEA) and Followers-based Peeling Algorithm (FPA), wherein they greedily expand or peel vertices, respectively, to identify the maximal one in a breadth-first manner, utilizing a novel vertex ordering and followers concept. To further accelerate the computation, we propose an advanced Expand-and-Filter Algorithm (EFA++) that first acquires a slightly larger community and subsequently executes a filtering phase employing a dynamic programming technique. Furthermore, we define the "Affected Area" for edge insertion or deletion that caters for dynamic scenarios. Finally, the theoretical analysis and extensive experimental evaluation on several real-life datasets demonstrate the effectiveness and efficiency of the proposed algorithms. [ABSTRACT FROM AUTHOR]

Although great efforts have been made in single source adaptation, most methods do not consider a more practical scenario where the source domain comprises multiple data distributions. In this paper, we propose a new multiple teacher–student framework (MTSF) for multi-source domain adaptive object detection. Our MTSF adopts multiple teacher–student sub-networks to distill domain specific knowledge from different source domains, yielding multiple precise pseudo labels. Considering that the samples from different domains are biased, we propose a new knowledge fusion module namely hybrid knowledge distillation (HKD) to fuse the knowledge by considering the contribution of each sample from different domains to the final decision on the target domain. Our HKD consists of a multi-source prototype construction (MSPC) module and a dense-to-sparse cross-domain pseudo label refinement (DCPR) module. According to the prototypes measured by MSPC, the DCPR distills the knowledge from dense to sparse in a hierarchical way, which makes the generated pseudo labels more reliable. When evaluated on benchmark adaptation scenarios (i.e. cross-camera and cross-time), our proposed method outperforms previous methods, establishing a new state-of-the-art. [ABSTRACT FROM AUTHOR]

Feature selection plays a key role in preprocessing, effectively addressing the curse of dimensionality in multi-label learning. While current approaches commonly utilize feature or label similarity to construct the weight matrix, typically through manifold learning regularization, there has been a dearth of progress in developing similarity-constrained regularization terms for multi-label feature selection. To address this gap, this paper conducts a comprehensive investigation and proposes a novel similarity constraint leveraging non-negative matrix factorization techniques. Subsequently, a new algorithm is introduced termed Multi-label Feature Selection via Similarity Constraints with Non-negative Matrix Factorization (SCNMF). Initially, the Gaussian similarity matrix among features is computed and factorized into a weight matrix using non-negative matrix factorization. Subsequently, the Cosine distance matrix among labels is computed to constrain the weight matrix. Finally, an objective function is formulated based on the aforementioned constraint mechanisms and iteratively optimized. Extensive experiments conducted across more than 10 multi-label datasets demonstrate the superior performance of our proposed approach. The source code is accessible at the URL: https://github.com/BIGOatMNNU/SCNMF. [ABSTRACT FROM AUTHOR]

Detecting digital audio tampering is essential for ensuring judicial fairness and societal security. Traditional methods, primarily based on Electric Network Frequency (ENF), have been limited by their reliance on singular, static features, which overlook critical temporal dynamics inherent in ENF data. This results in suboptimal detection accuracy. Addressing these limitations, this paper introduces a novel transformer model, ENFformer, designed for the detection of digital audio tampering by leveraging both short and long-term temporal features of ENF data. Initially, the ENFformer model extracts traditional zero-order and first-order phase features using discrete Fourier transforms (D F T 0 and D F T 1 ), along with frequency features obtained through the Hilbert transform. These features are processed using a frame-based algorithm to develop their temporal counterparts. To enhance feature extraction, the model employs a two-layer one-dimensional Convolutional Long Short-Term Memory (ConvLSTM) network to assimilate short-term temporal features, followed by a Bidirectional Long Short-Term Memory (BiLSTM) network for long-term feature integration. A branch attention mechanism then synergizes these long-term features, which are further refined by a transformer module for accurate tampered audio identification. Our empirical evaluations on the Carioca and ENF-EDIT1 databases demonstrate that ENFformer achieves detection accuracies of 97.33% and 93.50% respectively, surpassing existing state-of-the-art methods. These results confirm the effectiveness of our approach, which significantly advances the field of digital audio tampering detection by incorporating a comprehensive analysis of temporal information in ENF features. The source code of this study is publicly available at https://github.com/CCNUZFW/ENFformer. • Shallow ENF features extraction aids deep exploration, enhancing tampering detection. • ConvLSTM and BiLSTM networks extract deep ENF features, improved by branch attention. • The proposed ENFformer, incorporating transformers, outperforms current SOTA models. [ABSTRACT FROM AUTHOR]

Recently, pixel-value-ordering (PVO) based reversible data hiding (RDH) methods have become hotspots in spatial domain RDH research. In PVO-based methods, the pixel value correlation within a local region is exploited for data embedding. To embed secret data, the cover images are partitioned into non-overlapped rectangle blocks to be sorted locally, and a lower embedding distortion is thus obtained. However, their embedding capacities are often limited by the fixed local blocks from which pixels are sorted. To maintain the advantages and overcome disadvantages of the PVO-based methods, a new global pixel-value-ordering (GPVO) framework is proposed in this paper. By applying our GPVO framework, a PVO-based method can utilize pixels from any position in the cover image rather than from local pixel blocks. Then, through the dynamic sequence partition realized by the GPVO framework for the first time, the embedding capacity of PVO-based methods can be greatly improved while reducing embedding distortion further reduced. Finally, we propose a two-stage pairwise embedding scheme to be applied to the sequences, aiming to achieve an advanced embedding performance. Experimental results illustrate that the proposed GPVO works better than other related state-of-the-art RDH methods. [ABSTRACT FROM AUTHOR]

Link prediction within social networks represents a notable research area. Leveraging semantic text and multidimensional user interaction features, this paper introduces a hierarchical and progressive user interaction matrix-based link prediction method in social networks. First, considering the dynamic nature of a user's semantic text features, the word2vec algorithm was employed to extract text features. Given the long short-term memory algorithm's proficiency in learning long-term sequence dependencies, we proposed a method to represent time-series text called "TT2vec" in order to capture the evolution pattern of a user's semantic text features. Second, to address the multidimensionality of user interactions, we constructed a hierarchical and progressive interaction network adjacency matrix, thereby deriving advanced user interaction features using an enhanced graph convolutional network. Then, using TT2vec to further capture the interaction network's evolution, we proposed HPIN2vec, which is a method for representing hierarchical and progressive user interaction networks. Finally, for the users' semantic text features and multidimensional interaction features, we proposed a link prediction method based on graph attention networks (GATs) and users' dynamic features, considering the advantage of GAT aggregating neighbor node features. Experimental findings indicated that the proposed method not only identifies immediate dynamics in user features but also enhances link prediction accuracy through complex hierarchical and progressive interaction patterns. [ABSTRACT FROM AUTHOR]

Unmanned Aerial Vehicles (UAVs), functioning as aerial robots, undertake multifarious missions within intricate and uncharted environments, necessitating autonomous exploration for efficient UAV navigation. This paper introduces a Deep Learning-based Autonomous UAV Exploration method (DLAE), which distinguishes itself from traditional autonomous ground robot exploration by employing cameras instead of radar sensors. This difference propels us toward proposing a hybrid action space that integrates both positional and yaw actions, aimed at circumventing the limitations imposed by the Field of View (FOV). Furthermore, to enhance learning efficiency in autonomous UAV exploration, we design an invalid action masking scheme. Significantly, our contribution includes the development of an autoregressive network model alongside a distinct training procedure specifically tailored for autonomous UAV exploration. Comprehensive experiments conducted across a variety of test maps reveal that our proposed method surpasses existing learning-based approaches, delivering superior exploration efficiency and reduced decision-making time in comparison to conventional strategies. Moreover, physical simulations have corroborated the effectiveness of our method in practical scenarios, showcasing its potential for real-world applications. [ABSTRACT FROM AUTHOR]

• A transfer learning-based digital multitwin adaptive fusion fault diagnosis framework is proposed, which is used for source domain sample generation and target domain feature extraction, enabling complementary fusion between models. • A multichannel, multiscale fault diagnosis network is designed, enabling multi-sensor fusion and fusion of multiple features within the network. • Two novel data twinning structures, 1DACGAN twinning and SWSLSTM twinning, are designed to extract different features from one-dimensional signals. • An effective transfer learning strategy is designed, enabling optimal parameter transfer and automatic sample selection. With the development of digital twin technology, mechanical failures can be more comprehensively described through the interaction between sensors and twin models, which thereby demonstrate significant potential in fault diagnosis. However, due to the limitations of modeling methods, digital twin model construction poses challenges and exhibits poor generalization, making it difficult to apply across domains. To address these issues, this paper proposes a digital multitwin fusion diagnostic algorithm based on transfer learning. The digital multitwin comprise the source domain mechanistic twin, the target domain distributional feature 1D Auxiliary Classifier GAN (1D-ACGAN) twin, and the target domain time-series feature shared weight stacking LSTM (SWSLSTM) twin. First, based on the fault features collected by multiple sensors in the mechanism twin, a multichannel, multiscale mid-fusion diagnostic network is constructed. Second, the network is hierarchically transferred to the target domain network using a frozen transfer technique, automatically finding the optimal network parameters. Furthermore, to enhance the data quality, a data twin composed of the SWSLSTM twin and 1D-ACGAN twin is proposed to capture the dependency relationships of target domain data features. Finally, the migrated network adaptively adjusts the generated samples of the data twin based on the designed indicators, achieving high-precision cross-domain application of the multitwin diagnostic model. The proposed algorithm has been validated utilising data from triple-piston pumps and openly accessible bearing datasets, achieving diagnostic accuracies of 91.6% and 95.28%, respectively. These results substantiate the efficacy of the algorithm, demonstrating its superior generalisation capability and robustness in comparison to extant classical methodologies. [ABSTRACT FROM AUTHOR]

Emotion recognition in conversation (ERC) has gained considerable attention owing to its extensive applications in the field of human-computer interaction. However, previous models have had certain limitations in exploring the potential emotional relationships within the conversation due to their inability to fully leverage speaker information. Additionally, information from various modalities such as text, audio, and video can synergistically enhance and supplement the analysis of emotional context within the conversation. Nonetheless, effectively fusing multimodal features to understand the detailed contextual information in the conversation is challenging. This paper proposes a Speaker-Aware Cognitive network with Cross-Modal Attention (SACCMA) for multimodal ERC to effectively leverage multimodal information and speaker information. Our proposed model primarily consists of the modality encoder and the cognitive module. The modality encoder is employed to fuse multimodal feature information from speech, text, and vision using a cross-modal attention mechanism. Subsequently, the fused features and speaker information are separately fed into the cognitive module to enhance the perception of emotions within the dialogue. Compared to seven common baseline methods, our model increased the Accuracy score by 2.71 % and 1.70 % on the IEMOCAP and MELD datasets, respectively. Additionally, the F1 score improved by 2.92 % and 0.70 % for each dataset. Various experiments also demonstrate the effectiveness of our method. [ABSTRACT FROM AUTHOR]

To address 3D object retrieval, substantial efforts have been made to generate highly discriminative descriptors for 3D objects represented by a single modality, such as voxels, point clouds, or multiview images. It is promising to leverage complementary information from multimodal representations of 3D objects to further improve retrieval performance. However, multimodal 3D object retrieval has rarely been developed or analyzed for large-scale datasets. In this paper, we propose a self-and-cross-attention-based aggregation of point clouds and multiview images (SCA-PVNet) for 3D object retrieval. With deep features extracted from point clouds and multi-view images, we design two types of feature aggregation modules, namely the in-modality aggregation module (IMAM) and the cross-modality aggregation module (CMAM), for effective feature fusion. IMAM leverages a self-attention mechanism to aggregate multiview features, whereas CMAM exploits a cross-attention mechanism to interact with point-cloud and multiview features. The final descriptor of a 3D object for object retrieval can be obtained by concatenating the aggregated feature outputs of both modules. Extensive experiments and analyses were conducted on four datasets ranging from small to large scales to demonstrate the superiority of the proposed SCA-PVNet over state-of-the-art methods. In addition to achieving state-of-the-art retrieval performance, our method is more robust in challenging scenarios when views or points are missing during inference. [ABSTRACT FROM AUTHOR]

Transformer-based detectors have recently achieved remarkable success in object detection, revolutionizing the field with their efficiency and accuracy. However, applying these models to high-resolution images presents significant challenges due to the increased computational demands and complexity of processing dense, high-resolution data. In this paper, we introduce a novel model specifically designed for object detection from high-resolution imagery. This model incorporates a multi-layer object-focus network along with a transformer encoder-decoder structure. Specifically, the model employs a dual-head strategy in the object-focus network to balance the detailed analysis of small objects with computational efficiency. This is achieved by leveraging data sparsity to reduce unnecessary computations in massive background regions. Additionally, to improve detection performance for small objects, we propose a method to effectively apply the transformer encoder-decoder structure combined with the object-focus network on the multi-layer feature maps of the feature pyramid. Our extensive evaluations of the VisDrone, MS-COCO, and UAVid datasets demonstrate that our model outperforms other DETR-based detectors in both detection accuracy and computational speed, highlighting its superior performance. These results indicate a significant advancement in the field of high-resolution object detection utilizing transformer architecture. [ABSTRACT FROM AUTHOR]

In recent years, classification problems have been widely studied as one of the critical research directions in artificial intelligence. Efficiently simulating the human brain's ability to solve deterministic classification problems has become a prominent research branch in the field of classification problems. Numerical spiking neural P systems (NSN P systems) with the characteristics of the highly flexible spiking neural P (SNP)systems architecture and addressing numerical variants have special advantages for solving classification problems in the real world. This paper proposes, a novel NSN P system, namely, a learning numerical spiking neural P system (LNSN P system), is proposed to solve classification problems. In LNSN P systems, each neuron contains not only discrete spikes and rules that must be described by regular expressions but also numerical variables and production functions that can be in any form, overcoming the discontinuity of traditional SN P systems. In addition, an ascending structure without an outer encoder has been designed in this study to encode information from low-dimensional indivisible space to high-dimensional divisible space. At the same time, an adaptive learning mechanism has been introduced to improve algorithm convergence and precision. Finally, the experimental results obtained using eight benchmark datasets from the machine learning repository demonstrate the feasibility and effectiveness of the proposed LNSN P system. Moreover, the proposed LNSN P system exhibits lower space complexity and better generalisation ability than the existing SN P systems with the learning ability to solve different classification problems. • An LNSN P system is proposed to solve classification problems. • An ascending structure without an outer encoder is designed to encode information. • An adaptive learning mechanism is introduced to improve algorithm performance. [ABSTRACT FROM AUTHOR]