Unlabeled Ones Research Articles

Contrastive subspace distribution learning for novel category discovery in high-dimensional visual data

Discovering novel visual categories, particularly in recognition tasks, is a prominent area of research in artificial intelligence. Visual data, such as images and videos, is inherently high-dimensional and they often exhibit low-dimensional subspace structures, making the task of uncovering novel low-dimensional subspace structures a substantial challenge. To address this challenge, we develop a novel subspace discovery method, consisting of four steps. First, we introduce a base-expressive network (BENet) designed to learn coefficients that express the relationships between samples and bases. Additionally, these coefficients are normalized as a probability distribution using the softmax function. Second, to enhance this process, we propose a Contrastive Subspace Distribution Learning (CSDL) technique, utilizing Rényi Divergence. CSDL combines self-supervised and supervised contrastive distances between the normalized coefficients, considering all samples and labeled samples, respectively. Third, the base-expressive coefficients are subsequently computed by the learned SENet, which effectively transfers subspace information from labeled samples to unlabeled ones. Finally, we employ spectral clustering on the predicted coefficients to automatically assign categories to unlabeled samples. Extensive experimental results demonstrate that our approach surpasses related methods in the realms of category discovery, unsupervised clustering, and semi-supervised (subspace) clustering.

Weakly Supervised Transformer for Radar Jamming Recognition

Radar jamming recognition is a key step in electronic countermeasures, and accurate and sufficient labeled samples are essential for supervised learning-based recognition methods. However, in real practice, collected radar jamming samples often have weak labels (i.e., noisy-labeled or unlabeled ones), which degrade recognition performance. Additionally, recognition performance is hindered by limitations in capturing the global features of radar jamming. The Transformer (TR) has advantages in modeling long-range relationships. Therefore, a weakly supervised Transformer is proposed to address the issues of performance degradation under weak supervision. Specifically, complementary label (CL) TR, called RadarCL-TR, is proposed to improve radar jamming recognition accuracy with noisy samples. CL learning and a cleansing module are successively utilized to detect and remove potentially noisy samples. Thus, the adverse influence of noisy samples is mitigated. Additionally, semi-supervised learning (SSL) TR, called RadarSSL-PL-TR, is proposed to boost recognition performance under unlabeled samples via pseudo labels (PLs). Network generalization is improved by training with pseudo-labeling unlabeled samples. Moreover, the RadarSSL-PL-S-TR is proposed to further promote recognition performance, where a selection module identifies reliable pseudo-labeling samples. The experimental results show that the proposed RadarCL-TR and RadarSSL-PL-S-TR outperform comparison methods in recognition accuracy by at least 7.07% and 6.17% with noisy and unlabeled samples, respectively.

Scribble Hides Class: Promoting Scribble-Based Weakly-Supervised Semantic Segmentation with Its Class Label

Scribble-based weakly-supervised semantic segmentation using sparse scribble supervision is gaining traction as it reduces annotation costs when compared to fully annotated alternatives. Existing methods primarily generate pseudo-labels by diffusing labeled pixels to unlabeled ones with local cues for supervision. However, this diffusion process fails to exploit global semantics and class-specific cues, which are important for semantic segmentation. In this study, we propose a class-driven scribble promotion network, which utilizes both scribble annotations and pseudo-labels informed by image-level classes and global semantics for supervision. Directly adopting pseudo-labels might misguide the segmentation model, thus we design a localization rectification module to correct foreground representations in the feature space. To further combine the advantages of both supervisions, we also introduce a distance entropy loss for uncertainty reduction, which adapts per-pixel confidence weights according to the reliable region determined by the scribble and pseudo-label's boundary. Experiments on the ScribbleSup dataset with different qualities of scribble annotations outperform all the previous methods, demonstrating the superiority and robustness of our method. The code is available at https://github.com/Zxl19990529/Class-driven-Scribble-Promotion-Network.

Efficient and Inexpensive Synthesis of 15N-Labeled 2-Azido-1,3-dimethylimidazolinium Salts Using Na15NO2 Instead of Na15NNN.

15N-Labeled azides are important probes for infrared and magnetic resonance spectroscopy and imaging. They can be synthesized by reaction of primary amines with a 15N-labeled diazo-transfer reagent. We present the synthesis of 15N-labeled 2-azido-1,3-dimethylimidazolinium salts 1 as a 15N-labeled diazo-transfer reagent. Nitrosation of 1,3-dimethylimidazolinium-2-yl hydrazine (2) with Na15NO2 under acidic conditions gave 1 as a 1:1 mixture of α- and γ-15N-labeled azides, α- and γ-1, rather than γ-1 alone. The isotopomeric mixture thus obtained was then subjected to the diazo-transfer reaction with primary amines 3 to afford azides 4 as a 1:1 mixture of β-15N-labeled azides β-4 and unlabeled ones 4'. The efficient and inexpensive synthesis of 1 as a 1:1 mixture of α- and γ-1 using Na15NO2 instead of Na15NNN facilitates their wide use as a 15N-labeled diazo-transfer reagent for preparing 15N-labeled azides as molecular probes.

SPLAL: Similarity-based pseudo-labeling with alignment loss for semi-supervised medical image classification

Medical image classification presents significant challenges due to limited labeled samples and class imbalance resulting from varying disease prevalence. In many real-world scenarios, a scarcity of labeled images pertaining to specific ailments coexists with an abundance of unlabeled ones. The process of annotating these unlabeled medical images by domain experts is both time-intensive and financially burdensome. Consequently, medical image classification remains an active and evolving research area. Semi-supervised learning (SSL) techniques hold promise in addressing these challenges by effectively leveraging both labeled and unlabeled data. However, the application of SSL to medical image classification necessitates overcoming two primary obstacles: (1) the reliable estimation of pseudo-labels for unlabeled images and (2) the mitigation of biases stemming from class imbalance. In this work, we introduce a novel semi-supervised learning approach, denoted as SPLAL, which adeptly addresses these challenges. SPLAL leverages the concept of class prototypes and employs a weighted combination of classifiers to predict dependable pseudo-labels for a subset of unlabeled medical images. Additionally, we introduce an alignment loss term designed to alleviate model biases, particularly those arising from the overrepresentation of majority classes. This alignment loss mechanism ensures the model’s consistency in delivering the same classification output for diverse augmented images of the same disease. To evaluate the effectiveness of our proposed approach, we conducted comprehensive experiments on two publicly accessible benchmark datasets for medical image classification. Our empirical findings unequivocally demonstrate that SPLAL outperforms several state-of-the-art semi-supervised learning methods across various evaluation metrics.

Mapping user behaviors to identify professional accounts in Ethereum using semi-supervised learning

Ethereum is one of the largest blockchain platforms currently that has become a digital business environment. This platform allows for decentralized transactions between anonymous users. Thus, the development of methods to identify users’ behaviors and keep them anonymous can potentially leverage business on this platform. In this work, we aim to combine different categories of machine learning approaches, namely, unsupervised and semi-supervised, to map the behaviors of users’ owned accounts and identify users with professional activities in Ethereum. In addition, we provide here data to the community and analyze different machine learning techniques to characterize the users of Ethereum. These are challenging tasks due to the small fraction of publicly labeled data referring to users’ accounts that provide services on this platform, such as exchange, payment, and entertainment, among most casual behavior users. Initially, we use unsupervised learning techniques to cluster the unlabeled users’ accounts and to identify a set of them with casual behavior. As an outcome, a dataset containing labeled (casual or professional) and unlabeled instances is obtained. Semi-supervised learning methods are then applied (i) to generate models that classify accounts’ behaviors into casual or professional ones and (ii) to discover accounts with professional behaviors among the unlabeled ones. Computational experiments were conducted, and the results obtained by the proposed procedure are compared to those achieved by supervised learning techniques from the literature. The proposal outperformed those from the literature and reached values higher than 95% for the accuracy, precision, recall, Fβ-scores, MCC, and AUC-ROC.

Interactive Lexical and Semantic Graphs for Semisupervised Relation Extraction.

The performance of relation extraction (RE) is hindered by the lack of sufficient labeled data. Semisupervised methods can offer to help hands with this problem by augmenting high-quality unlabeled samples into the training data. However, existing semisupervised RE methods either need a set of manually defined rules or rely on the classifier trained on the small labeled data, i.e., the former requires the heavy intervention of human knowledge, and the latter is bound to the number and the quality of the labeled data. In this article, we present a novel semisupervised RE method that involves small human efforts and is robust to the size of the initial set of labeled data. Specifically, we adopt only two simple rules to build the lexical and semantic graphs which connect the labeled samples with the unlabeled ones. In this way, the graphs are much easier to construct yet keep the ability to transfer knowledge from labeled samples to unlabeled ones. We then develop a graph interaction module to fully exploit the reference information in lexical and semantic graphs, which is used to jointly recognize the high-quality unlabeled samples with the classifier. We conduct extensive experimental results on two public datasets. The results demonstrate that our proposed method significantly outperforms the state-of-the-art baselines.

An Adaptive Layer to Leverage Both Domain and Task Specific Information from Scarce Data

Many companies make use of customer service chats to help the customer and try to solve their problem. However, customer service data is confidential and as such, cannot easily be shared in the research community. This also implies that these data are rarely labeled, making it difficult to take advantage of it with machine learning methods. In this paper we present the first work on a customer’s problem status prediction and identification of problematic conversations. Given very small subsets of labeled textual conversations and unlabeled ones, we propose a semi-supervised framework dedicated to customer service data leveraging speaker role information to adapt the model to the domain and the task using a two-step process. Our framework, Task-Adaptive Fine-tuning, goes from predicting customer satisfaction to identifying the status of the customer’s problem, with the latter being the main objective of the multi-task setting. It outperforms recent inductive semi-supervised approaches on this novel task while only considering a relatively low number of parameters to train on during the final target task. We believe it can not only serve models dedicated to customer service but also to any other application making use of confidential conversational data where labeled sets are rare. Source code is available at https://github.com/gguibon/taft

VILL: Toward Efficient and Automatic Visual Landmark Labeling

Of all indoor localization techniques, vision-based localization emerges as a promising one, mainly due to the ubiquity of rich visual features. Visual landmarks, which present distinguishing textures, play a fundamental role in visual indoor localization. However, few researches focus on visual landmark labeling. Preliminary arts usually designate a surveyor to select and record visual landmarks, which is tedious and time-consuming. Furthermore, due to structural changes (e.g., renovation), the visual landmark database may be outdated, leading to degraded localization accuracy. To overcome these limitations, we propose VILL , a user-friendly, efficient, and accurate approach for visual landmark labeling. VILL asks a user to sweep the camera to take a video clip of his/her surroundings. In the construction stage, VILL identifies unlabeled visual landmarks from videos adaptively according to the graph-based visual correlation representation. Based on the spatial correlations with selected anchor landmarks, VILL estimates locations of unlabeled ones on the floorplan accurately. In the update stage, VILL formulates an alteration identification model based on the judgments from different users to identify altered landmarks accurately. Extensive experimental results in two different trial sites show that VILL reduces the site survey substantially (by at least 65.9%) and achieves comparable accuracy.

A Dual Neighborhood Hypergraph Neural Network for Change Detection in VHR Remote Sensing Images

The very high spatial resolution (VHR) remote sensing images have been an extremely valuable source for monitoring changes occurring on the Earth’s surface. However, precisely detecting relevant changes in VHR images still remains a challenge, due to the complexity of the relationships among ground objects. To address this limitation, a dual neighborhood hypergraph neural network is proposed in this article, which combines multiscale superpixel segmentation and hypergraph convolution to model and exploit the complex relationships. First, the bi-temporal image pairs are segmented under two scales and fed to a pre-trained U-net to obtain node features by treating each object under the fine scale as a node. The dual neighborhood is then defined using the father-child and adjacent relationships of the segmented objects to construct the hypergraph, which permits models to represent higher-order structured information far more complex than the conventional pairwise relationships. The hypergraph convolutions are conducted on the constructed hypergraph to propagate the label information from a small amount of labeled nodes to the other unlabeled ones by the node-edge-node transformation. Moreover, to alleviate the problem of imbalanced sampling, the focal loss function is adopted to train the hypergraph neural network. The experimental results on optical, SAR and heterogeneous optical/SAR data sets demonstrate that the proposed method offersbetter effectiveness and robustness compared to many state-of-the-art methods.

Adaptive Betweenness Clustering for Semi-Supervised Domain Adaptation.

Compared to unsupervised domain adaptation, semi-supervised domain adaptation (SSDA) aims to significantly improve the classification performance and generalization capability of the model by leveraging the presence of a small amount of labeled data from the target domain. Several SSDA approaches have been developed to enable semantic-aligned feature confusion between labeled (or pseudo-labeled) samples across domains; nevertheless, owing to the scarcity of semantic label information of the target domain, they were arduous to fully realize their potential. In this study, we propose a novel SSDA approach named Graph-based Adaptive Betweenness Clustering (G-ABC) for achieving categorical domain alignment, which enables cross-domain semantic alignment by mandating semantic transfer from labeled data of both the source and target domains to unlabeled target samples. In particular, a heterogeneous graph is initially constructed to reflect the pairwise relationships between labeled samples from both domains and unlabeled ones of the target domain. Then, to degrade the noisy connectivity in the graph, connectivity refinement is conducted by introducing two strategies, namely Confidence Uncertainty based Node Removal and Prediction Dissimilarity based Edge Pruning. Once the graph has been refined, Adaptive Betweenness Clustering is introduced to facilitate semantic transfer by using across-domain betweenness clustering and within-domain betweenness clustering, thereby propagating semantic label information from labeled samples across domains to unlabeled target data. Extensive experiments on three standard benchmark datasets, namely DomainNet, Office-Home, and Office-31, indicated that our method outperforms previous state-of-the-art SSDA approaches, demonstrating the superiority of the proposed G-ABC algorithm.

A Semi-Supervised Inspection Approach of Textured Surface Defects under Limited Labeled Samples

Defect inspection is a key step in guaranteeing the surface quality of industrial products. Based on deep learning (DL) techniques, related methods are highly effective in defect classification tasks via a supervision process. However, collecting and labeling many defect samples are usually harsh and time-consuming processes, limiting the application of these supervised classifiers on various textured surfaces. This study proposes a semi-supervised framework, based on a generative adversarial network (GAN) and a convolutional neural network (CNN), to classify defects of a textured surface, while a novel label assignment scheme is proposed to integrate unlabeled samples into semi-supervised learning to enhance the overall performance of the system. In this framework, a customized GAN uses limited labeled samples to generate unlabeled ones, while the proposed label assignment scheme makes the generated data follow different label distributions in such a way that they can participate in training with labeled data. Finally, a CNN is proposed for semi-supervised training and the category identification of each defect sample. Experimental results show the effectiveness and robustness of the proposed framework even if original samples are limited. We verify our approach on four different surface defect datasets, achieving consistently competitive performances.

Ensemble-Based Semi-Supervised Learning for Milling Chatter Detection

Chatter is one of the most deleterious phenomena during the machining process, and leads to a low quality of workpiece surface, a noisy workplace, and decreases in tool and machine life. In order to overcome these limitations and improve the machining performance, various effective methods have been developed for chatter detection. The main shortcoming of such methods is that they require all the data to be labeled. However, the labeled data that accurately reflect the chatter states are hard to collect in practical application. This paper proposes a semi-supervised method to classify chatter states with a small quantity of labeled data and large quantity of unlabeled ones. In order to improve the classification accuracy and generalization ability, ensemble learning is combined with the semi-supervised method, and an EB-SSL model is proposed in this paper. Take the non-stationarity and multiple scaling behaviors of chatter data into consideration, multifractal detrended fluctuation analysis (MF-DFA) is utilized to extract distinguished features from raw chatter detection signals. Experimental results show that this method can identify the chatter states more accurately. The performance analysis indicates that the proposed method is applicable in different milling conditions.

Mathematical Foundations of Graph-Based Bayesian Semi-Supervised Learning

In recent decades, science and engineering have been revolutionized by a momentous growth in the amount of available data. However, despite the unprecedented ease with which data are now collected and stored, labeling data by supplementing each feature with an informative tag remains to be challenging. Illustrative tasks where the labeling process requires expert knowledge or is tedious and time-consuming include labeling X-rays with a diagnosis, protein sequences with a protein type, texts by their topic, tweets by their sentiment, or videos by their genre. In these and numerous other examples, only a few features may be manually labeled due to cost and time constraints. How can we best propagate label information from a small number of expensive labeled features to a vast number of unlabeled ones? This is the question addressed by semi-supervised learning (SSL). This article overviews recent foundational developments on graph-based Bayesian SSL, a probabilistic framework for label propagation using similarities between features. SSL is an active research area and a thorough review of the extant literature is beyond the scope of this article. Our focus will be on topics drawn from our own research that illustrate the wide range of mathematical tools and ideas that underlie the rigorous study of the statistical accuracy and computational efficiency of graph-based Bayesian SSL.

Semi-supervised semantic segmentation with cross teacher training

Convolutional neural networks can achieve remarkable performance in semantic segmentation tasks. However, such neural network approaches heavily rely on costly pixel-level annotation. Semi-supervised learning is a promising resolution to tackle this issue, but its performance still far falls behind the fully supervised counterpart. This work proposes a cross-teacher training framework with three modules that significantly improves traditional semi-supervised learning approaches. The core is a cross-teacher module, which could simultaneously reduce the coupling among peer networks and the error accumulation between teacher and student networks. In addition, we propose two complementary contrastive learning modules. The high-level module can transfer high-quality knowledge from labeled data to unlabeled ones and promote separation between classes in feature space. The low-level module can encourage low-quality features learning from the high-quality features among peer networks. In experiments, the cross-teacher module significantly improves the performance of traditional student–teacher approaches, and our framework outperforms state-of-the-art methods on benchmark datasets.

Inter-subject registration-based one-shot segmentation with alternating union network for cardiac MRI images.

Medical image segmentation based on deep-learning networks makes great progress in assisting disease diagnosis. However, currently, the training of most networks still requires a large amount of data with labels. In reality, labeling a considerable number of medical images is challenging and time-consuming. In order to tackle this challenge, a new one-shot segmentation framework for cardiac MRI images based on an inter-subject registration model called Alternating Union Network (AUN) is proposed in this study. The label of the source image is warped with deformation fields discovered from AUN to segment target images directly. Initially, the volumes are pre-processed by aligning affinely and adjusting the global intensity to simplify subsequent deformation registration. AUN consists of two kinds of subnetworks trained alternately to optimize segmentation gradually. The first kind of subnetwork takes a pair of volumes as inputs and registers them using global intensity similarity. The second kind of subnetwork, which takes the predicted labels generated from the previous subnetwork and the labels refined using the information of intrinsic anatomical structures of interest as inputs, is intensity-independent and focuses attention on registering structures of interest. Specifically, the input of AUN is a pair of a labeled image with the texture in regions of interest removed and a target image. Additionally, a new similarity measurement more appropriate for registering such image pair is defined as Local Squared Error (LSE). The proposed registration-based one-shot segmentation pays attention to the problem of the lack of labeled medical images. In AUN, only one labeled volume is required and a large number of unlabeled ones can be leveraged to improve segmentation performance, which has great advantages in clinical application. In addition, the intensity-independent subnetwork and LSE proposed in this study empower the framework to segment medical images with complicated intensity distribution.

Semi-Supervised Cloud Detection in Satellite Images by Considering the Domain Shift Problem

In terms of semi-supervised cloud detection work, efforts are being made to learn a promising cloud detection model via a limited number of pixel-wise labeled images and a large number of unlabeled ones. However, remote sensing images obtained from the same satellite sensor often show a data distribution drift problem due to the different cloud shapes and land-cover types on the Earth’s surface. Therefore, there are domain distribution gaps between labeled and unlabeled satellite images. To solve this problem, we take the domain shift problem into account for the semi-supervised learning (SSL) network. Feature-level and output-level domain adaptations are applied to reduce the domain distribution gaps between labeled and unlabeled images, thus improving predicted results accuracy of the SSL network. Experimental results on Landsat-8 OLI and GF-1 WFV multispectral images demonstrate that the proposed semi-supervised cloud detection network (SSCDnet) is able to achieve promising cloud detection performance when using a limited number of labeled samples and outperforms several state-of-the-art SSL methods.

Generic semi-supervised adversarial subject translation for sensor-based activity recognition

Performance of Human Activity Recognition (HAR) models, particularly deep neural networks, is highly contingent upon the availability of the massive amount of annotated training data. Though, data collection and manual labeling in the HAR domain are prohibitively expensive due to human resource dependence in both steps. Hence, domain adaptation techniques are proposed to adapt the knowledge from the existing source of data. More recently, adversarial transfer learning methods have shown promising results for visual classification, yet limited for HAR problems, which are still prone to the unfavorable effects of the imbalanced distribution of samples. This paper presents a novel generic semi-supervised approach that takes advantage of the adversarial framework to tackle these shortcomings by leveraging knowledge from annotated samples exclusively from the source subject and unlabeled ones of the target subject. An extensive subject translation experiments is conducted on three large, middle, and small-size datasets with different levels of imbalance to assess the robustness of the proposed model to the scale as well as the imbalance in the data. The results demonstrate the effectiveness of our proposed algorithms over state-of-the-art methods, which led to up to 13%, 4%, and 13% improvement of our high-level activities recognition metrics for Opportunity, LISSI, and PAMAP2 datasets, respectively.

Semi-supervised EEG emotion recognition model based on enhanced graph fusion and GCN

Objective. To take full advantage of both labeled data and unlabeled ones, the Graph Convolutional Network (GCN) was introduced in electroencephalography (EEG) based emotion recognition to achieve feature propagation. However, a single feature cannot represent the emotional state entirely and precisely due to the instability of the EEG signal and the complexity of the emotional state. In addition, the noise existing in the graph may affect the performance greatly. To solve these problems, it was necessary to introduce feature/similarity fusion and noise reduction strategies. Approach. A semi-supervised EEG emotion recognition model combining graph fusion, network enhancement, and feature fusion was proposed. Firstly, different features were extracted from EEG and then compacted by Principal Component Analysis (PCA), respectively. Secondly, a Sample-by-sample Similarity Matrix (SSM) was constructed based on each feature, and similarity network fusion (SNF) was adopted to fuse the graphs corresponding to different SSMs to take advantage of their complementarity. Then, Network Enhancement (NE) was performed on the fused graph to reduce the noise in it. Finally, GCN was performed on the concatenated features and the enhanced fused graph to achieve feature propagation. Main results. Experimental results demonstrated that: (a) When 5.30 of SEED and 7.20 of SEED-IV samples were chosen as the labeled samples, respectively, the minimum classification accuracy improvement achieved by the proposed scheme over state-of-the-art schemes were 1.52 on SEED and 13.14 on SEED-IV, respectively. (b) When 8.00 of SEED and 9.60 of SEED-IV samples were chosen as the labeled samples, respectively, the minimum training time reduction achieved by the proposed scheme over state-of-the-art schemes were 46.75 s and 22.55 s, respectively. (c) Graph fusion, network enhancement, and feature fusion all contributed to the performance enhancement. (d) The key hyperparameters that affect the performance were relatively few and easy to set to obtain outstanding performance. Significance. This paper demonstrated that the combination of graph fusion, network enhancement, and feature fusion help to enhance GCN-based EEG emotion recognition.

Multiple-instance domain adaptation for cost-effective sensor-based human activity recognition

Machine learning-based human activity recognition (HAR) is important as the means of human–computer interaction to empower the existing systems in many areas, such as healthcare, entertainment, logistics, and manufacturing. To build such a recognition tool, it is clear that sufficient labeled samples are required. Oftentimes, it is more difficult to obtain labeled samples rather than to obtain unlabeled ones due to the prohibitive conditions (e.g., financial cost, time, hazardous environment, and human labor), so we end up with incomplete or weakly labeled data. The multiple-instance learning technique (MIL) alleviates such issue by allowing us to leverage weakly labeled data by performing the classification of a bag of instances rather than a single instance. However, since multiple-instance learning is intrinsically the generalization of supervised learning, it may face the same problem as the usual supervised learning approaches: performance degradation on the data with different distribution. In fact, such distribution difference is common in sensor-based HAR which makes it difficult for a classifier model to perform predictions. For example, the difficulties happen when the current data distribution is shifted due to sensor deterioration, or when the model that is generated from a certain domain is applied to a different domain (e.g., different person with different device placement, posture, and gait). In this work, we propose a multiple-instance domain adaptation approach that handles weakly annotated data for model training, while providing adaptation mechanism to deal with data distribution difference. We incorporate “high-level adaptation” and “bag-level adaptation” to find a robust sensor data representation which minimizes distribution difference. The proposed approach is tested on standard sensor-based HAR datasets, conditioned on weak annotation and cross-domain settings. Our experimental result shows promising recognition performance improvements compared to the classical MIL and domain adaptation approaches.