Label polysemy, where an instance can be associated with multiple labels, is common in real-world tasks. LDL (label distribution learning) is an effective learning paradigm for handling label polysemy, where each instance is associated with a label distribution. Although numerous LDL algorithms have been proposed and achieved satisfactory performance on most existing datasets, they are typically trained directly on the collected label distributions which often lack quality guarantees in real-world tasks due to annotator subjectivity and algorithm assumptions. Consequently, direct learning from such uncertain label distributions can lead to unpredictable generalization performance. To address this problem, we propose an adaptive-grained label distribution learning framework whose main idea is to extract relatively reliable supervision information from unreliable label distributions, and thus the label distribution learning task can be decomposed into three subtasks: coarsening label distributions, learning coarse-grained labels and refining coarse-grained labels. In this framework, we design an adaptive label coarsening algorithm to extract an optimal coarsen-grained labels and a label refining function to enhance the coarse-grained label into the final label distributions. Finally, we conduct extensive experiments on real-world datasets to demonstrate the advantages of our proposal.

Automatic Speech Recognition (ASR) systems are vulnerable to adversarial examples (AEs), where small, carefully designed perturbations are added to original audio to mislead the systems into generating target commands. Existing adversarial attacks typically initialize perturbations either as zero or as Text-to-Speech clips of the target command. The former accumulates the features of the command in the perturbed audio, while the latter constantly reduces the features of the command, resulting in the generation of AEs. Although most target commands in the AEs are imperceptible to humans, the audio often exhibits noticeable distortions or disruptions, making it apparent that the sound has been tampered with. This work aims to retain only the essential features of adversarial audio, minimizing distortions from unnecessary elements to improve quality and make the attack less detectable. Our findings highlight the importance of formants as critical features for black-box adversarial attacks, motivating the development of a novel Formant Filter Bank (FFB) tailored to the target command. By inputting musical audio into the FFB, we utilize the filtered output as the perturbation seed, which retains the formant features of the target command and blends in certain features of the original music. Then we search for a minimum enhancement factor for the perturbation seed to generate high-quality AEs. Our perturbation can be regarded as local amplitude modulation of the music, so we define the AE as EvilHarmony. Experimental results demonstrate that our method successfully attacks commercial black-box ASR models, including Microsoft, Google, Amazon, Tencentyun, Aliyun, and OpenAI Whisper-V3. Compared to existing approaches, our AEs achieve significantly greater stealth, with 53% to 77% of participants perceiving them as indistinguishable from normal audio across the six ASR API services. Additionally, our approach successfully attacks Google Assistant and voice assistants on Surface Pro 9 in the real world. Demos are uploaded at https://sites.google.com/view/evilharmony.

Three-way decision (3WD) is a trisecting-and-acting model that is in line with the decision-making logic of human beings. Bayesian risk decision (BRD) theory is a statistical method that optimizes decisions by quantifying and minimizing uncertainty-induced losses. As a typical BRD method, cost-sensitive 3WD aims to reduce decision risks and enhance decision credibility by incorporating an uncertainty decision item, known as boundary decision. However, in most existing 3WD-based multiclassification (MC) models, there is redundancy in both the decision item set and cost matrix, which may lead to decision conflict, ambiguity, and redundancy. Moreover, determining the cost matrix under the intuitionistic fuzzy (IF) environment remains a challenging task. To address these problems, we propose a reliable three-way multiclassification (R3WMC) model under the IF environment. Specifically, an IF cost transformation mechanism is proposed, and a novel boundary decision associated with the corresponding compact cost matrix is established to achieve a conflict-free decision set. Motivated by confidence-based classifiers, we propose an effective decision rule which is time-efficient and able to solve decision conflicts. We also demonstrate that our model is an enhanced BRD model, contributing to a solid theoretical basis for the proposed R3WMC method. Finally, a comparative analysis of multiple models across various datasets validates the reliability and superior performance of our model.

Harmonic diffractive optical elements (HDOEs) are characterized by their lightweight and compact size, making them promising candidates for applying in future large-aperture space optical imaging systems. However, its lower focusing efficiency and unavoidable manufacturing errors can result in degraded and blurred imaging. To effectively improve the imaging quality of HDOE optical systems, this study proposes an image super-resolution (SR) method based on plug-and-play (PnP) technology, referred to as HDSR. Specifically, the study first establishes the objective function for image SR and then introduces a Poissonian-Gaussian noise model to describe the noise in HDOE optical imaging systems. Based on this, a denoiser based on a convolutional neural network (CNN) is trained and used as the prior term in the optimization function. In addition, the study proposes a learning-based parameter auto-estimation and updating mechanism to reduce the complexity of manually tuning iterative parameters in the PnP technology. In the experimental section, the study explores and verifies the role and importance of the adopted noise model and parameter estimation mechanism. The results show that the proposed HDSR method significantly enhances the imaging quality of the HDOE optical system. In outdoor scenes, the natural image quality evaluator (NIQE) metric average value after SR using this method is 9.28, which is a 49.62% improvement compared to the bicubic method.

Abstract Text-to-image (TTI) models provide huge innovation ability for many industries, while the content security triggered by them has also attracted wide attention. Considerable research has focused on content security threats of large language models (LLMs), yet comprehensive studies on the content security of TTI models are notably scarce. This paper introduces a systematic tool, named EvilPromptFuzzer, designed to fuzz evil prompts in TTI models. For 15 kinds of fine-grained risks, EvilPromptFuzzer employs the strong knowledge-mining ability of LLMs to construct seed banks, in which the seeds cover various types of characters, interrelations, actions, objects, expressions, body parts, locations, surroundings, etc. Subsequently, these seeds are fed into the LLMs to build scene-diverse prompts, which can weaken the semantic sensitivity related to the fine-grained risks. Hence, the prompts can bypass the content audit mechanism of the TTI model, and ultimately help to generate images with inappropriate content. For the risks of violence, horrible, disgusting, animal cruelty, religious bias, political symbol, and extremism, the efficiency of EvilPromptFuzzer for generating inappropriate images based on DALL.E 3 are greater than 30%, namely, more than 30 generated images are malicious among 100 prompts. Specifically, the efficiency of horrible, disgusting, political symbols, and extremism up to 58%, 64%, 71%, and 50%, respectively. Additionally, we analyzed the vulnerability of existing popular content audit platforms, including Amazon, Google, Azure, and Baidu. Even the most effective Google SafeSearch cloud platform identifies only 33.85% of malicious images across three distinct categories.

Recently, the three-way decision models based on prospect theory have attracted much attention, because they can well consider the risk attitude of decision-makers when decisions involve gains and losses. However, in terms of reference point selection for prospect theory, these models adopt a unified strat-egy for all alternatives without considering the characteristics of the alternatives themselves. To solve this problem, a reference point selection method based on fuzzy information granules is proposed in this paper. Furthermore, a prospect theory-based generalized intuitionistic fuzzy three-way decision model is established. Specifically, firstly, a reference point selection method based on fuzzy information granules is constructed for the evaluation values of alternatives. Secondly, based on the value function of prospect theory, an intuitionistic fuzzy scoring function is proposed. Meanwhile, based on the difference between the two intuitionistic fuzzy scoring values, a PROMETHEE-based trisecting method is designed. In addition, utilizing the priority relationship existing in three divisions, a generalized ranking strategy using recursive trisecting is established for obtaining the ranking result of alternatives. Finally, based on the priority relationship of the classification attribute, an effectiveness index is proposed to evaluate the effectiveness of the decision-making method in decision-making cases. On this basis, the proposed model is used to process different decision-making cases, verifying its effectiveness, superiority, and feasibility.

On the basis of the support vector machine (SVM), Large Margin Distribution Machine (LDM) improves the generalization performance by incorporating the marginal distribution theory. Nevertheless, the current LDM models (LDMs) still exhibit limitations when it comes to handling noisy data, such as: i) LDMs fails to effectively discern the samples being noise and consequently falls short in robust defenses. ii) The hinge loss of LDMs is predicated upon the minimal inter-category separation, rendering the corresponding classifier highly susceptible to perturbations induced by noise. To address these limitations, we leverage the fuzzy set theory and pinball loss function, and propose a novel Fuzzy Large Margin Distribution Machine with Unified Pinball Loss (FUPLDM), which is performed as: i) An innovative fuzzy membership function is developed, utilizing two distinct types of feature centers and their associations with the samples. The function assigns a probability to each sample, indicating the likelihood of it being classified as noise. As a result, the model gains the remarkable ability to accurately identify and distinguish noise from other data. ii) Replace the hinge loss function with the unified pinball loss. The pinball loss function is based on interquartile distance, which is less affected by noise and can well improve the noise immunity of the classifier at the boundary. Therefore, FUPLDM has superior noise recognition capabilities and substantial noise resistance against its detrimental effects. Furthermore, We also analyzed the properties of FUPLDM, including noise insensitivity, intra-class distance, inter-class scatter, and misclassification error. At last, we conduct a series of comparative experiments on artificial synthetic datasets, UCI benchmark datasets, and noise-added UCI datasets, which demonstrate the effectiveness and superiority of FUPLDM.

The weight of traditional space optical imaging systems often increases nearly cubically with the increase in aperture size. To overcome this limitation, this study proposes the use of a lightweight harmonic diffractive optical imaging system combined with deep denoiser prior image restoration, to achieve high-resolution, lightweight remote sensing.This research first designed a novel 150-order harmonic diffractive optical element (H-DOE) featuring a 40-mm aperture size and a 320-mm focal length, which is equipped with seven annular zones covering a broad spectral band (500-800 nm). Its slim structure significantly reduces weight and volume, thereby lowering its launch costs. Furthermore, to address the blurring issues encountered in H-DOE imaging tasks and attain enhanced image resolution, this study incorporates an advanced image restoration technique. This technique employs a deep denoiser as a prior module, which is embedded into a model-based image restoration optimization framework. The newly trained deep denoiser utilizes a U-Net architecture integrating a transformer and residual structures and is adept at handling complex noise during the optical imaging process. Experimental results demonstrate that the performance of the proposed image restoration strategy based on a deep denoiser surpasses that of the existing technologies, elevating the resolvable frequency of the modulation transfer function (MTF) of an H-DOE imaging system from 40.58 lp/mm to 98.55 lp/mm, an enhancement of 142.9%. This significant image quality improvement showcases its vast potential for use in future high-resolution, lightweight remote sensing applications.

Competitor identification is an essential component of corporate strategy. With the rapid development of artificial intelligence, various data-mining methodologies and frameworks have emerged to identify competitors. In general, the competitiveness among companies is determined by both market commonality and resource similarity. However, because resource information is more difficult to obtain than market information, existing studies primarily identify competitors via market commonality. To address this limitation, we introduce multisource company descriptions as well as heterogeneous business relationships, and we propose a novel method for simultaneously mining the market commonality and resource similarity. First, we use multisource company descriptions to represent companies and transform the heterogeneous business relationships into a heterogeneous business network. Then, we propose a novel multisource data fusion-based heterogeneous graph attention network (MHGAT) to learn the pairwise competitive relationships between companies. Specifically, a graph neural network-based model is proposed to learn the embeddings of companies by preserving their competition, and a multilevel attention framework is designed to integrate the embeddings from neighboring company level, heterogeneous relationship level, and multisource description level. Finally, experiments on a real-world dataset verify the effectiveness of our proposed MHGAT and demonstrate the usefulness of company descriptions and business relationships in competitor identification.

Intuitionistic fuzzy (IF) set theory combined with twin support vector machines (TSVM) has shown highly advantageous performance in robust and fast classification. However, the existing methods have some inherent flaws: 1) The inaccurate confidence caused by the single-center-based IF strategy degrades the classification performance of the model. 2) The nonmembership degree is time-consuming and the threshold is difficult to determine, as it depends on sample's local neighborhood set. 3) Recent theoretical studies have demonstrated that optimizing the margin distribution achieve better generalization performance than maximizing the minimum margin in SVM-based models (SVMs). In view of the aforementioned shortcomings, we propose a novel dual-center-based IF (DC-IF) function, and introduce the margin distribution into IFTSVM, forming a DC-IF twin bounded large margin distribution machines (DC-IFTBLDM). Initially, the DC-IF strategy references the contribution of one sample to both categories, which is time efficient and more accurately reflects the credibility of the samples. Moreover, the margin mean and variance are used to optimize the margin distribution, and a regularization term is utilized to minimize structural risk. Therefore, DC-IFTBLDM has a good performance in noise immunity and classification capability. Comparative experiments based on synthetic datasets, UCI benchmark datasets, and datasets with different noise ratios verify the effectiveness of the proposed model. The parameter sensitivity analysis further proves the high accuracy and stability of DC-IFTBLDM.

The approximation learning of fuzzy concepts associated with fuzzy rough sets and three-way decisions is a useful technology for the representation, learning, and transformation of uncertain knowledge. However, how to integrate the advantages of fuzzy rough approximations and three-way approximations has rarely been carefully investigated so far. In this article, we focus on exploring the connection and interplay of these two methods, and further propose the hierarchical fuzzy rough approximations in the dynamic fuzzy open-world environment. We utilize the temporal-spatial perspectives of three-way decisions to construct multigranularity structures and implement multigranularity learning. The temporality of data and the spatiality of model parameters are both considered in such frameworks. Subsequently, we discuss the interpretation and representation of fuzzy three-way regions in fuzzy rough sets with some definitions and properties. The max, double, and aggregated evaluation-based models of three-way approximations are proposed to gradually transform a fuzzy concept to a crisp concept by the time-evolving attributes. Finally, the comparative experimental results between static and dynamic approaches demonstrate the effectiveness of our proposed hierarchical approximation learning models.

Decision-theoretic rough sets (DTRSs) capture the decision semantics and can deduce three-way decisions with respect to the minimum expected risk. Considering the new evaluation format of hesitant fuzzy sets, we extend the range of applications of DTRSs to hesitant fuzzy information systems. In this case, the integrated approach by considering the interaction between information systems and loss functions becomes one of challenges. Different from the results reported in most of the existing papers, we combine the hesitant fuzzy information system and loss functions together via error analysis. In the hesitant fuzzy information system, a new binary relation is first defined by utilizing the normalization of hesitant fuzzy elements and the distance function. Then, the calculations of the similarity class and the conditional probability are discussed. With the aid of the error analysis method, we effectively aggregate the loss functions presented in the similarity class and determine the expected losses in the format of the intervals. Based on the possibility degree, we further explore the decision rules by comparing the expected losses. With the above analysis, we design a decision-making procedure of three-way decisions in a hesitant fuzzy information system. Finally, we elaborate the application of three-way decisions in hesitant fuzzy information systems by an example of the security evaluation of peer-to-peer lending platforms and validate our methods.

Three-way decisions play an important role in rough sets and decision theory. As a representative model, decision-theoretic rough sets (DTRSs) provide a sound interpretation of thresholds used in three-way decisions. This problem is associated with the determination of the loss function of DTRSs. In this paper, we investigate a novel way of determining the loss functions of DTRSs with relative values. More specifically, with the aid of analytic hierarchy process (AHP) method, the determination of loss functions is realized in the context of DTRSs. First, a hierarchical structure of DTRSs is constructed. Second, along the hierarchical structure, pairwise comparison matrices are analyzed in a top-down fashion. In light of the generic condition imposed on the loss functions of DTRSs, some constraints on relative values between loss functions are introduced. The relative ratios at each level are computed. Considering the consistency ratio (CR) of the reciprocal matrices, two mathematical programming approaches are developed by exploiting the flexibility of information granularity. Then, we design a decision procedure for the determination of loss functions and deduce three-way decisions. The loss functions are calculated by aggregating the relative ratios being available at each level. With regard to the loss functions, we finally compare the existing studies with the AHP method. We demonstrate that the relative value with AHP improves the restriction of the existing studies and exhibits a certain level of tolerance to inconsistency.

Decision-theoretic rough sets (DTRSs) play a crucial role in risk decision-making problems. With respect to the minimum expected risk, DTRSs deduce the rules of three-way decisions. Considering the new expression of evaluation information with hesitant fuzzy sets (HFSs), we introduce HFSs into DTRSs and explore their decision mechanisms. More specifically, we take into account the losses of DTRSs with hesitant fuzzy elements and propose a new model of hesitant fuzzy decision-theoretic rough sets (HFDTRSs). Some properties of the expected losses and their corresponding scores are carefully investigated under the hesitant fuzzy information. Three-way decisions and the associated cost of each object are further derived. With the above analysis, a novel risk decision-making method with the aid of HFDTRSs is developed. Besides the three-way decisions with DTRSs, the method investigates the ranking and resource allocation by utilizing the associated costs of alternatives and multiobjective 0-1 integer programming. Our study also offers a solution in the aspect of determining losses of DTRS and extends the range of applications.

For most attribute reduction in Pawlak rough set model (PRS), monotonicity is a basic property for the quantitative measure of an attribute set. Based on the monotonicity, a series of attribute reductions in Pawlak rough set model such as positive-region-preserved reductions and condition entropy-preserved reductions are defined and the corresponding heuristic algorithms are proposed in previous rough sets research. However, some quantitative measures of attribute set may be non-monotonic in probabilistic rough set model such as decision-theoretic rough set (DTRS), and the non-monotonic definition of the attribute reduction should be reinvestigated and the heuristic algorithm should be reconsidered. In this paper, the monotonicity of the positive region in PRS and DTRS are comparatively discussed. Theoretic analysis shows that the positive region in DTRS model may be expanded with the decrease of the attributes, which is essentially different from that in PRS model. Hereby, a new non-monotonic attribute reduction is presented for the DTRS model in this paper, and a heuristic algorithm for searching the newly defined attribute reduction is proposed, in which the positive region is allowed to be expanded instead of remaining unchanged in the process of attribute reduction. Experimental analysis is included to validate the theoretic analysis and quantify the effectiveness of the proposed attribute reduction algorithm.

Dominance-based rough sets approach (DRSA) is an effective tool to deal with information with preference-ordered attribute domains and decision classes. Any information system may evolve when new objects enter into or old objects get out. Approximations of DRSA need update for decision analysis or other relative tasks. Incremental updating is a feasible and effective technique to update approximations. The purpose of this paper is to present an incremental approach for updating approximations of DRSA. The approach is applicable to dynamic information systems when the set of objects varies over time. In this paper, we discuss the principles of incrementally updating P-dominating sets and P-dominated sets and propose an incremental approach for updating approximations of DRSA. A numerical example is given to illustrate the incremental approach. The experimental evaluations on data sets from UCI show that the incremental approach outperforms the original nonincremental one.

Incremental learning is an efficient technique for knowledge discovery in a dynamic database, which enables acquiring additional knowledge from new data without forgetting prior knowledge. Rough set theory has been successfully used in information systems for classification analysis. Set-valued information systems are generalized models of single-valued information systems, which can be classified into two categories: disjunctive and conjunctive. Approximations are fundamental concepts of rough set theory, which need to be updated incrementally while the object set varies over time in the set-valued information systems. In this paper, we analyze the updating mechanisms for computing approximations with the variation of the object set. Two incremental algorithms for updating the approximations in disjunctive/conjunctive set-valued information systems are proposed, respectively. Furthermore, extensive experiments are carried out on several data sets to verify the performance of the proposed algorithms. The results indicate the incremental approaches significantly outperform non-incremental approaches with a dramatic reduction in the computational speed.

By considering the levels of tolerance for errors and the cost of actions in real decision procedure, a new two-stage approach is proposed to solve the multiple-category classification problems with Decision-Theoretic Rough Sets (DTRS). The first sta

Approximations of a concept in rough set theory induce rules and need to update for dynamic data mining and related tasks. Most existing incremental methods based on the classical rough set model can only be used to deal with the categorical data. This paper presents a new dynamic method for incrementally updating approximations of a concept under neighborhood rough sets to deal with numerical data. A comparison of the proposed incremental method with a nonincremental method of dynamic maintenance of rough set approximations is conducted by an extensive experimental evaluation on different data sets from UCI. Experimental results show that the proposed method effectively updates approximations of a concept in practice. © 2012 Wiley Periodicals, Inc.

Knowledge in an information system evolves with its dynamical environment. A new concept of interesting knowledge based on both accuracy and coverage is defined in this paper for dynamic information systems. An incremental model and approach as well as its algorithm for inducing interesting knowledge are proposed when the object set varies over time. A case study validates the feasibility of the proposed method.

This paper presents a development of component synthesis vibration suppression (CSVS) method for control of flexible structures. The proposed method eliminates unwanted flexible modes of vibration while achieving the desired rigid body motion. The robustness to uncertainties of dynamic modeling parameters is analyzed. Unlike traditional input shaping, in which a numerical optimization is used, design for CSVS commands is based on analytic methodology and is relatively easy to implement. A case study is performed on a time-fuel optimal control strategy using constant amplitude reaction jet thrusters. Both simulation and experimental results validate the effectiveness of the proposed approach.