Partial multilabel feature selection (PMLFS) is a prevalent subject that aims to enhance the performance of multilabel learning (MLL) in the context of noisy labels. In PMLFS, a crucial aspect is handling the false positive labels hidden in the candidate label set, as the imprecise annotations could mislead the feature selection process. However, many existing approaches for partial label disambiguation rely on topology information and tend to be error-prone. Besides, feature selection frameworks are often built upon a linear regression model, leading to a reliance on the classifier and a deficiency in exploring local structures. Focusing on the issues above, this article proposes a novel two-stage PMLFS method, resorting to the ideology of granular computing. In the first stage, a label disambiguation method is developed using label-specific information. Specifically, a specific granular ball computing model is designed to characterize the distribution of datapoints labeled differently, and therefore, using the affinity relationships among samples and balls, the label-specific information concealed in the data distribution can be captured for label disambiguation. In the second stage, a filter-based feature selection method that explores the local structure of samples is presented. This method relies on a devised fuzzy decision neighborhood rough set (FDNRS) to capture more detailed membership information by maximizing the neighborhood consistency of samples' related labels. Simultaneously, the feature selection method minimizes the uncertainty derived from unrelated labels. Extensive experiments on 12 datasets in terms of four evaluation metrics demonstrated the effectiveness of the proposed approach.

Partial label learning (PLL) is a complicated weakly supervised multi-classification task compounded by class imbalance. Currently, existing methods only rely on inter-class pseudo-labeling from inter-class features, often overlooking the significant impact of the intra-class imbalanced features combined with the inter-class. To address these limitations, we introduce Granular Ball Representation for Imbalanced PLL (GBRIP), a novel framework for imbalanced PLL. GBRIP utilizes coarse-grained granular ball representation and multi-center loss to construct a granular ball-based feature space through unsupervised learning, effectively capturing the feature distribution within each class. GBRIP mitigates the impact of confusing features by systematically refining label disambiguation and estimating imbalance distributions. The novel multi-center loss function enhances learning by emphasizing the relationships between samples and their respective centers within the granular balls. Extensive experiments on standard benchmarks demonstrate that GBRIP outperforms existing state-of-the-art methods, offering a robust solution to the challenges of imbalanced PLL.

Partial multilabel learning (PML) addresses the issue of noisy supervision, which contains an overcomplete set of candidate labels for each instance with only a valid subset of training data. Using label enhancement techniques, researchers have computed the probability of a label being ground truth. However, enhancing labels in the noisy label space makes it impossible for the existing partial multilabel label enhancement methods to achieve satisfactory results. Besides, few methods simultaneously involve the ambiguity problem, the feature space's redundancy, and the model's efficiency in PML. To address these issues, this article presents a novel joint partial multilabel framework using broad learning systems (namely BLS-PML) with three innovative mechanisms: 1) a trustworthy label space is reconstructed through a novel label enhancement method to avoid the bias caused by noisy labels; 2) a low-dimensional feature space is obtained by a confidence-based dimensionality reduction method to reduce the effect of redundancy in the feature space; and 3) a noise-tolerant BLS is proposed by adding a dimensionality reduction layer and a trustworthy label layer to deal with PML problem. We evaluated it on six real-world and seven synthetic datasets, using eight state-of-the-art partial multilabel algorithms as baselines and six evaluation metrics. Out of 144 experimental scenarios, our method significantly outperforms the baselines by about 80%, demonstrating its robustness and effectiveness in handling partial multilabel tasks.

Mirror detection is of great significance for avoiding false recognition of reflected objects in computer vision tasks. Existing mirror detection frameworks usually follow a supervised setting, which relies heavily on high quality labels and suffers from poor generalization. To resolve this, we instead propose the first weakly-supervised mirror detection framework and also provide the first scribble-based mirror dataset. Specifically, we relabel 10,158 images, most of which have a labeled pixel ratio of less than 0.01 and take only about 8 seconds to label. Considering that the mirror regions usually show great scale variation, and also irregular and occluded, thus leading to issues of incomplete or over detection, we propose a local-global feature enhancement (LGFE) module to fully capture the context and details. Moreover, it is difficult to obtain basic mirror structure using scribble annotation, and the distinction between foreground (mirror) and background (non-mirror) features is not emphasized caused by mirror reflections. Therefore, we propose a foreground-aware mask attention (FAMA), integrating mirror edges and semantic features to complete mirror regions and suppressing the influence of backgrounds. Finally, to improve the robustness of the network, we propose a prototype contrast loss (PCL) to learn more general foreground features across images. Extensive experiments show that our network outperforms relevant state-of-the-art weakly supervised methods, and even some fully supervised methods. The dataset and codes are available at https://github.com/winter-flow/WSMD.

Accurate estimation of the Lithium-ion batteries (LiBs) state of health (SOH) is essential to ensure the safe and reliable operation of battery-powered devices. Most of the current data-driven SOH estimation models are designed under fixed ambient temperatures, overlooking the high sensitivity of LiBs to changing ambient temperatures. To bridge this gap, a novel method is proposed with transfer learning (TL) to model and estimate the SOH at varying ambient temperatures. First, canonical variate analysis is employed to capture the temporal dynamics in the time-series data of LiBs and extract temporal features. Second, at the reference temperature, an interpretable SOH estimation model with a long short-term memory network is utilized to learn the regression relationship between the temporal features and the labeled capacities. Third, at the new temperature, a small amount of data is projected into the key feature domain at the reference temperature to obtain the common temporal features. Afterward, TL is employed to take advantage of the existing process knowledge through keeping all functional layers of the previous well-trained SOH estimation model. Finally, the efficacy of the proposed method is verified on the NASA dataset with two discrete temperatures, 24°C and 44°C. Evaluated by the index of root mean squared error, the proposed method is capable of improving prediction accuracy by 94.18% when the model is transferred from 44°C to 24°C.

Partial label learning (PLL) tackles scenarios where the unique ground-truth label of each sample is concealed within a candidate label set. Dimensionality reduction, considering labeling confidence estimation, has become a promising strategy to enhance the generalization performance of PLL models. However, current studies achieve dimensionality reduction, often relying on kNN-based labeling confidence estimation or disregarding potential labeling information. To address this issue, this paper proposes a novel Confidence-induced granular Partial label feature selection method using Dependency and Similarity (CPDS), which consists of two phases: Labeling Confidence Estimation (LCE) and Feature Selection (FS). For LCE, through granular ball computing, the feature space's similarity and the label space's correlation between the training data and the granular ball can be fused simultaneously, thereby effectively reconstructing more credible labeling confidence from candidate labels with more diverse semantic representation information. In the FS stage, by leveraging the LC with more diverse information, the proposed PLL neighborhood decision system further effectively combines feature dependency and label similarity to identify a feature subset with more discriminative capabilities, thereby achieving better performance for classification tasks. Among them, feature dependency effectively utilizes the dependency between neighborhoods and equivalence relations, while label similarity fully exploits the similarity between each sample and its neighbors. Extensive experiments show that CPDS significantly outperforms the compared approaches in most cases on nine controlled UCI datasets and five real-world datasets, demonstrating the superiority of the proposed method.

Label distribution learning (LDL), leveraging the label significance (LS), is more appropriate for solving label ambiguity problems than multilabel learning (MLL). However, directly obtaining the LS of LDL is extremely expensive and challenging. Thus, label enhancement (LE) algorithms are effectively proposed to acquire inherent LS from MLL for training the LDL models. Nevertheless, most existing LE models will suffer from low accuracy and low efficiency with following issues: ignoring mapping relationship between feature and label space, resulting in inaccurate enhanced data; designing independently apart from LDL models, resulting in an inability of unified LE-LDL learning; and requiring to optimize numerous parameters iteratively, resulting in worse training efficiency. Consequently, a novel unified LE-LDL learning framework, namely stacked graph-regularized polynomial-based fuzzy broad learning system (SGP-FBLS), is proposed by following three innovations: polynomial-based fuzzy system is introduced to enhance feature mapping ability while improving the learning performance effectively; graph regularized-based optimization objective function (GP-FBLS) is presented by considering interinstance correlation and label correlation to mine potential LS, thereby improving the accuracy of subsequent LDL tasks; and a weight stacked strategy is innovatively proposed to directly transmit LS and weighted parameters from GP-FBLS to SGP-FBLS without retraining, achieving the most satisfactory performance while significantly improving the training efficiency. Finally, comparative studies on 19 practical datasets demonstrate the effectiveness and superiority of proposed methods.

Feature selection can reduce the dimensionality of data effectively. Most of the existing feature selection approaches using rough sets focus on the static single type data. However, in many real-world applications, data sets are the hybrid data including symbolic, numerical and missing features. Meanwhile, an object set in the hybrid data often changes dynamically with time. For the hybrid data, since acquiring all the decision labels of them is expensive and time-consuming, only small portion of the decision labels for the hybrid data is obtained. Therefore, in this paper, incremental feature selection algorithms based on information granularity are developed for dynamic partially labeled hybrid data with the variation of an object set. At first, the information granularity is given to measure the feature significance for partially labeled hybrid data. Then, incremental mechanisms of information granularity are proposed with the variation of an object set. On this basis, incremental feature selection algorithms with the variation of a single object and group of objects are proposed, respectively. Finally, extensive experimental results on different UCI data sets demonstrate that compared with the non-incremental feature selection algorithms, incremental feature selection algorithms can select a subset of features in shorter time without losing the classification accuracy, especially when the group of objects changes dynamically, the group incremental feature selection algorithm is more efficient.

With the development of deep learning, synthetic aperture radar (SAR) image ship detection based on the convolutional neural network has made significant progress. However, there are two problems. 1) The false alarm detection rate is high due to complex background and coherent speckle noise interference. 2) For smaller ship targets, missed detection is prone to occur. In this letter, a novel ship detection model (MFTF-Net) based on multi-feature transformation and fusion is proposed to address the issues. First, to avoid the randomness of initial point selection and the influence of outlier points, the anchor frame clustering approach based on the K-medians++ algorithm is presented to cluster the object candidate frames. Second, the low-level feature information is passed to the high level by constructing a local enhancement network; then, an improved Transformer structure is introduced to replace the last convolutional block of the backbone network to obtain rich contextual information. Finally, a four-scale residual feature fusion network is designed, which fully fuses the object’s detailed and semantic information. In addition, improved convolutional block attention module (CBAM) and squeeze and excitation (SE) attention mechanisms are applied in the lower two layers and upper two layers of the network output to reduce the interference of confusing information, respectively. The experimental results demonstrate that the proposed method is superior to the state-of-the-art thirteen baseline models on SAR ship detection dataset (SSDD), high-resolution SAR images dataset (HRSID), and SAR-ship-dataset public datasets in terms of the mAP, recall, accuracy, and F1 metrics.

Attribute reduction plays an important role in knowledge discovery and data mining. Confronted with data characterized by the interval and missing values in many data analysis tasks, it is interesting to research the attribute reduction for interval-valued data with missing values. Uncertainty measures can supply efficient viewpoints, which help us to disclose the substantive characteristics of such data. Therefore, this paper addresses the attribute reduction problem based on uncertainty measure for interval-valued data with missing values. At first, an uncertainty measure is provided for measuring candidate attributes, and then an efficient attribute reduction algorithm is developed for the interval-valued data with missing values. To improve the efficiency of attribute reduction, the objects that fall within the positive region are deleted from the whole object set in the process of selecting attributes. Finally, experimental results demonstrate that the proposed algorithm can find a subset of attributes in much shorter time than existing attribute reduction algorithms without losing the classification performance.

Rough set theory has been applied successfully in knowledge discovery, computational intelligence and decision analysis. It can only deal with features of symbolic type for complete data. However, incomplete data with fuzzy decision under a preference-ordered relation is common in real-world applications. In this paper, we propose a feature selection framework for such data by combining the dominance-based rough sets. At first, the judgment theorems are established by the Boolean reasoning techniques. Then, the discernibility matrix and the discernibility function approach are proposed to find all subsets of features. In addition, an efficient feature selection algorithm to find a feature subset is proposed. Finally, the experimental results show that, in most cases for different data sets, the proposed algorithm is effective and efficient for feature selection from the incomplete data with fuzzy decision.

In many machine learning and data mining applications, it may happen that the data acquired for classification analysis are set-valued, i.e., The feature values of an object set are set-valued, which can be used to characterize uncertain information in decision making tasks. Set-valued data are the generalized models of single-valued data. Some mutual information-based feature selection algorithms have been extensively studied, but less effort has been made to investigate the feature selection issue with the mutual information analysis in set-valued data. Just owing to these, mutual information is firstly introduced in the set-valued data in this paper. Unlike the traditional computations, the mutual information is estimated on the unmarked objects. Correspondingly, a feature selection algorithm based on mutual information is developed, which is implemented in a dwindling universe to quicken the feature selection process. Compared with the state-of-the-art methods, the experimental results on different data sets demonstrate the efficiency and effectiveness of the proposed algorithm in set-valued data.