LDAAD: An effective label de-noising approach for anomaly detection

Automatic modulation classification (AMC) has a wide range of applications in both civilian and military fields, such as industrial Internet of Things (IIoT) security, communication spectrum management, and military electronic countermeasures. However, label mislabeling often occurs in practical scenarios, significantly impacting the performance and robustness of deep neural networks (DNNs). In this article, we propose a meta-learning guided label noise distillation method to enhance the robustness of AMC models against label noise or errors. Specifically, we propose a teacher-student heterogeneous network (TSHN) to discriminate and distill label noise. Following the notion that labels represent information, a teacher network, utilizing trusted few-shot labeled samples, reevaluates and corrects labels for a considerable number of untrusted labeled samples through meta-learning. By dividing and conquering untrusted labeled samples according to their confidence levels, the student network learns more effectively. Additionally, we propose a multiview signal (MVS) method to further enhance the performance of hard-to-classify categories with few-shot trusted labeled samples. Extensive experiments on the RadioML2016 and HisarMod2019.1 data sets demonstrate that our methods significantly improve accuracy and robustness in signal AMC across diverse label noise scenarios, including symmetric, asymmetric, and mixed label noise. For example, compared to the baseline convolutional neural network with the cross-entropy loss, our proposed TSHN achieves a remarkable 1.26% to 36.84% accuracy improvement under symmetric label noise and 0.12% to 38.59% accuracy improvement under mixed label noise. Moreover, TSHN exhibits greater robustness to varying label noise rates compared to existing methods.

Objective. Learning to classify cardiac abnormalities requires large and high-quality labeled datasets, which is a challenge in medical applications. Small datasets from various sources are often aggregated to meet this requirement, resulting in a final dataset prone to label noise due to inter- and intra-observer variability and different expertise. It is well known that label noise can affect the performance and generalizability of the trained models. In this work, we explore the impact of label noise and self-learning label correction on the classification of cardiac abnormalities on large heterogeneous datasets of electrocardiogram (ECG) signals. Approach. A state-of-the-art self-learning multi-class label correction method for image classification is adapted to learn a multi-label classifier for electrocardiogram signals. We evaluated our performance using 5-fold cross-validation on the publicly available PhysioNet/Computing in Cardiology (CinC) 2021 Challenge data, with full and reduced sets of leads. Due to the unknown label noise in the testing set, we tested our approach on the MNIST dataset. We investigated the performance under different levels of structured label noise for both datasets. Main results. Under high levels of noise, the cross-validation results of self-learning label correction show an improvement of approximately 3% in the challenge score for the PhysioNet/CinC 2021 Challenge dataset and an improvement in accuracy of 5% and reduction of the expected calibration error of 0.03 for the MNIST dataset. We demonstrate that self-learning label correction can be used to effectively deal with the presence of unknown label noise, also when using a reduced number of ECG leads.

Collecting large-scale data with clean labels for supervised training of neural networks is practically challenging. Although noisy labels are usually cheap to acquire, existing methods suffer a lot from label noise. This paper targets at the challenge of robust training at high label noise regimes. The key insight to achieve this goal is to wisely leverage a small trusted set to estimate exemplar weights and pseudo labels for noisy data in order to reuse them for supervised training. We present a holistic framework to train deep neural networks in a way that is highly invulnerable to label noise. Our method sets the new state of the art on various types of label noise and achieves excellent performance on large-scale datasets with real-world label noise. For instance, on CIFAR100 with a $40\%$ uniform noise ratio and only 10 trusted labeled data per class, our method achieves $80.2{\pm}0.3\%$ classification accuracy, where the error rate is only $1.4\%$ higher than a neural network trained without label noise. Moreover, increasing the noise ratio to $80\%$, our method still maintains a high accuracy of $75.5{\pm}0.2\%$, compared to the previous best accuracy $48.2\%$. Source code available: https://github.com/google-research/google-research/tree/master/ieg

Automated detection of vulnerabilities in source code is an essential cybersecurity challenge, underpinning trust in digital systems and services. Graph Neural Networks (GNNs) have emerged as a promising approach as they can learn the structural and logical code relationships in a data-driven manner. However, the performance of GNNs is severely limited by training data imbalances and label noise. GNNs can often learn “spurious” correlations due to superficial code similarities in the training data, leading to detectors that do not generalize well to unseen real-world data. In this work, we propose a new unified framework for robust and interpretable vulnerability detection—that we call VISION—to mitigate spurious correlations by systematically augmenting a counterfactual training dataset. Counterfactuals are samples with minimal semantic modifications that have opposite prediction labels. Our complete framework includes: (i) generating effective counterfactuals by prompting a Large Language Model (LLM); (ii) targeted GNN model training on synthetically paired code examples with opposite labels; and (iii) graph-based interpretability to identify the truly crucial code statements relevant for vulnerability predictions while ignoring the spurious ones. We find that our framework reduces spurious learning and enables more robust and generalizable vulnerability detection, as demonstrated by improvements in overall accuracy (from 51.8% to 97.8%), pairwise contrast accuracy (from 4.5% to 95.8%), and worst-group accuracy increasing (from 0.7% to 85.5%) on the widely popular Common Weakness Enumeration (CWE)-20 vulnerability. We also demonstrate improvements using our proposed metrics, namely, intra-class attribution variance, inter-class attribution distance, and node score dependency. We provide a new benchmark for vulnerability detection, CWE-20-CFA, comprising 27,556 samples from functions affected by the high-impact and frequently occurring CWE-20 vulnerability, including both real and counterfactual examples. Furthermore, our approach enhances societal objectives of transparent and trustworthy AI-based cybersecurity systems through interactive visualization for human-in-the-loop analysis.

Large-scale management of surface water resources in urban areas can be difficult, especially if the region is subject to monsoonal waterlogging. Deep learning-based methods for computer vision tasks, such as image segmentation, can effectively be applied to remote sensing data for generating water body maps of large cities, aiding managerial entities, urban planners and policymakers. The robustness of these models to erroneous pixel level training class labels has not been studied in water body segmentation context. Label noise is commonly experienced in classification tasks and may hinder the performance. We collected and densely labeled Sentinel-2 images over Dhaka, one of the most densely populated and flood prone megacities in the world. We synthetically injected four types of label noise viz., i) Gaussian noise, ii) translation, ii) rotation, and iv) mirroring. Our primary objective is to observe and quantitatively analyze the effects of label noise on remote sensing data-driven deep learning models for water body segmentation. Our results show that salt and pepper noise (injected artificially using Gaussian noise) of only 50% can cause a massive 48.55% drop in Intersection-over-Union score. The consequences of learning from training data with different magnitudes and settings of label noise have been explored.

In many applications of classifier learning, training data suffers from label noise. Deep networks are learned using huge training data where the problem of noisy labels is particularly relevant. The current techniques proposed for learning deep networks under label noise focus on modifying the network architecture and on algorithms for estimating true labels from noisy labels. An alternate approach would be to look for loss functions that are inherently noise-tolerant. For binary classification there exist theoretical results on loss functions that are robust to label noise. In this paper, we provide some sufficient conditions on a loss function so that risk minimization under that loss function would be inherently tolerant to label noise for multiclass classification problems. These results generalize the existing results on noise-tolerant loss functions for binary classification. We study some of the widely used loss functions in deep networks and show that the loss function based on mean absolute value of error is inherently robust to label noise. Thus standard back propagation is enough to learn the true classifier even under label noise. Through experiments, we illustrate the robustness of risk minimization with such loss functions for learning neural networks.

Semi-supervised person re-identification (Re-ID) is an important approach for alleviating annotation costs when learning to match person images across camera views. Most existing works assume that training data contains abundant identities crossing camera views. However, this assumption is not true in many real-world applications, especially when images are captured in nonadjacent scenes for Re-ID in wider areas, where the identities rarely cross camera views. In this work, we operate semi-supervised Re-ID under a relaxed assumption of identities rarely crossing camera views, which is still largely ignored in existing methods. Since the identities rarely cross camera views, the underlying sample relations across camera views become much more uncertain, and deteriorate the noise accumulation problem in many advanced Re-ID methods that apply pseudo labeling for associating visually similar samples. To quantify such uncertainty, we parameterize the probabilistic relations between samples in a relation discovery objective for pseudo label training. Then, we introduce reward quantified by identification performance on a few labeled data to guide learning dynamic relations between samples for reducing uncertainty. Our strategy is called the Rewarded Relation Discovery (R 2 D), of which the rewarded learning paradigm is under-explored in existing pseudo labeling methods. To further reduce the uncertainty in sample relations, we perform multiple relation discovery objectives learning to discover probabilistic relations based on different prior knowledge of intra-camera affinity and cross-camera style variation, and fuse the complementary knowledge of different probabilistic relations by similarity distillation. To better evaluate semi-supervised Re-ID on identities rarely crossing camera views, we collect a new real-world dataset called REID-CBD, and perform simulation on benchmark datasets. Experiment results show that our method outperforms a wide range of semi-supervised and unsupervised learning methods.

Voice-over-Internet-Protocol (VoIP) calls are prone to various speech impairments due to environmental and network conditions resulting in bad user experience. A reliable audio impairment classifier helps to identify the cause for bad audio quality. The user feedback after the call can act as the ground truth labels for training a supervised classifier on a large audio dataset. However, the labels are noisy as most of the users lack the expertise to precisely articulate the impairment in the perceived speech. In this paper, we analyze the effects of massive noise in labels in training dense networks and Convolutional Neural Networks (CNN) using engineered features, spectrograms and raw audio samples as inputs. We demonstrate that CNN can generalize better on the training data with a large number of noisy labels and gives remarkably higher test performance. The classifiers were trained both on randomly generated label noise and the label noise introduced by human errors. We also show that training with noisy labels requires a significant increase in the training dataset size, which is in proportion to the amount of noise in the labels.

Classification model performance can be degraded by label noise in the training set. The sentiment classification domain also struggles with this issue, whereby customer reviews can be mislabeled. Some customers give a rating score for a product or service that is inconsistent with the review content. If business owners are only interested in the overall rating picture that includes mislabeling, this can lead to erroneous business decisions. Therefore, this issue became the main challenge of this study. If we assume that customer reviews with noisy labels in the training data are validated and corrected before the learning process, then the training set can generate a predictive model that returns a better result for the sentiment analysis or classification process. Therefore, we proposed a mechanism, called polarity label analyzer, to improve the quality of a training set with noisy labels before the learning process. The proposed polarity label analyzer was used to assign the polarity class of each sentence in a customer review, and then polarity class of that customer review was concluded by voting. In our experiment, datasets were downloaded from TripAdvisor and two linguistic experts helped to assign the correct labels of customer reviews as the ground truth. Sentiment classifiers were developed using the k-NN, Logistic Regression, XGBoost, Linear SVM and CNN algorithms. After comparing the results of the sentiment classifiers without training set improvement and the results with training set improvement, our proposed method improved the average scores of F1 and accuracy by 20.59%.

Supervised classification of remotely sensed images is a classical method to update topographic geospatial databases. The task requires training data in the form of image data with known class labels, whose generation is time-consuming. To avoid this problem one can use the labels from the outdated database for training. As some of these labels may be wrong due to changes in land cover, one has to use training techniques that can cope with wrong class labels in the training data. In this paper we adapt a label noise tolerant training technique to the problem of database updating. No labelled data other than the existing database are necessary. The resulting label image and transition matrix between the labels can help to update the database and to detect changes between the two time epochs. Our experiments are based on different test areas, using real images with simulated existing databases. Our results show that this method can indeed detect changes that would remain undetected if label noise were not considered in training.

Cancer is a highly heterogeneous condition that can occur almost anywhere in the human body. [<sup>18</sup>F]fluorodeoxyglucose Positron Emission Tomography (<sup>18</sup>F-FDG PET) is a imaging modality commonly used to detect cancer due to its high sensitivity and clear visualisation of the pattern of metabolic activity. Nonetheless, as cancer is highly heterogeneous, it is challenging to train general-purpose discriminative cancer detection models, with data availability and disease complexity often cited as a limiting factor. Unsupervised learning methods, more specifically anomaly detection models, have been suggested as a putative solution. These models learn a healthy representation of tissue and detect cancer by predicting deviations from the healthy norm, which requires models capable of accurately learning long-range interactions between organs, their imaging patterns, and other abstract features with high levels of expressivity. Such characteristics are suitably satisfied by transformers, which have been shown to generate state-of-the-art results in unsupervised anomaly detection by training on normal data. This work expands upon such approaches by introducing multi-modal conditioning of the transformer via cross-attention i.e. supplying anatomical reference information from paired CT images to aid the PET anomaly detection task. Furthermore, we show the importance and impact of codebook sizing within a Vector Quantized Variational Autoencoder, on the ability of the transformer network to fulfill the task of anomaly detection. Using 294 whole-body PET/CT samples containing various cancer types, we show that our anomaly detection method is robust and capable of achieving accurate cancer localization results even in cases where normal training data is unavailable. In addition, we show the efficacy of this approach on out-of-sample data showcasing the generalizability of this approach even with limited training data. Lastly, we propose to combine model uncertainty with a new kernel density estimation approach, and show that it provides clinically and statistically significant improvements in accuracy and robustness, when compared to the classic residual-based anomaly maps. Overall, a superior performance is demonstrated against leading state-of-the-art alternatives, drawing attention to the potential of these approaches.

In real-world scenarios, collected and annotated data often exhibit the characteristics of multiple classes and long-tailed distribution. Additionally, label noise is inevitable in large-scale annotations and hinders the applications of learning-based models. Although many deep learning based methods have been proposed for handling long-tailed multi-label recognition or label noise respectively, learning with noisy labels in long-tailed multi-label visual data has not been well-studied because of the complexity of long-tailed distribution entangled with multi-label correlation. To tackle such a critical yet thorny problem, this paper focuses on reducing noise based on some inherent properties of multi-label classification and long-tailed learning under noisy cases. In detail, we propose a Stitch-Up augmentation to synthesize a cleaner sample, which directly reduces multi-label noise by stitching up multiple noisy training samples. Equipped with Stitch-Up, a Heterogeneous Co-Learning framework is further designed to leverage the inconsistency between long-tailed and balanced distributions, yielding cleaner labels for more robust representation learning with noisy long-tailed data. To validate our method, we build two challenging benchmarks, named VOC-MLT-Noise and COCO-MLT-Noise, respectively. Extensive experiments are conducted to demonstrate the effectiveness of our proposed method. Compared to a variety of baselines, our method achieves superior results.

Incorrect labels as well as the discrepancy between training and test domain data distributions can significantly affect the effectiveness of supervised data-driven models in machine fault diagnosis applications. Such a challenge can be characterized as the noisy label-domain generalization (NL-DG) problem. In this article, the extended invariant risk minimization (EIRM) is developed, which incorporates flat minima seeking to address the NL-DG challenge. The ability of handling NL-DG is realized by shifting the gradient penalty base from the dummy classifier to the entire model. EIRM is shown to be closely related to locating a flat minimum, which is crucial for label noise (LN) robustness and model generalization. Explorations on function smoothness and algorithm convergence are offered to understand EIRM from the theoretical aspect. An efficient implementation of EIRM is also developed to construct the fault diagnosis model. The EIRM-based fault diagnosis method is compared with strong benchmarks on multiple NL-DG tasks using actuator and gearbox fault datasets. Results indicate that the EIRM-based method on average is more effective than the benchmarks. The code is available at https://github.com/mozhenling/doge-eirm.

Existing multi-label learning algorithms mostly assume that all class labels in an obtained data set are completely accurate. This assumption is rarely valid. As the scale of the label space increases, owing to limited domain knowledge and differences among methods of data acquisition, some examples will be unlabeled or mislabeled. The introduction of such noise results in the poor performance of trained multi-label classifiers. This paper proposes a new method for multi-label learning with label noise, called low-rank-constraint-based multi-label learning with label noise (ML-LRC), by incorporating low-rank matrix recovery and features selection into the learning process. First, we assume that two strongly correlated labels share more samples than two weakly correlated labels. Based on this assumption, the ML-LRC applies low-rank constraints to the label matrix to mine the local correlations of class labels and remove noise from the observed label matrix. Second, we believe that each class label only correlates to some features. So we control the sparsity of the coefficient matrix to filter out label-specific features, which can reduce the complexity of the model and greatly improve the performance of the classifier. The accelerated proximal gradient method and alternating direction method of multipliers are adopted to solve the problem. Experiments conducted on four benchmark multi-label datasets demonstrate the competitive performance of our proposed method compared with several state-of-the-art methods.

When applying learning to rank algorithms in real search applications, noise in human labeled training data becomes an inevitable problem which will affect the performance of the algorithms. Previous work mainly focused on studying how noise affects ranking algorithms and how to design robust ranking algorithms. In our work, we investigate what inherent characteristics make training data robust to label noise. The motivation of our work comes from an interesting observation that a same ranking algorithm may show very different sensitivities to label noise over different data sets. We thus investigate the underlying reason for this observation based on two typical kinds of learning to rank algorithms (i.e.~pairwise and listwise methods) and three different public data sets (i.e.~OHSUMED, TD2003 and MSLR-WEB10K). We find that when label noise increases in training data, it is the \emph{document pair noise ratio} (i.e.~\emph{pNoise}) rather than \emph{document noise ratio} (i.e.~\emph{dNoise}) that can well explain the performance degradation of a ranking algorithm.