Labeling Cost Research Articles

Segmentation of Pulmonary Cavity in Lung CT Scan for Tuberculosis Disease

The complexity of pulmonary tuberculosis (TB) lung cavity lesion features significantly increase the cost of semantic segmentation and labelling. However, the high cost of semantic segmentation has limited the development of TB automatic recognition to some extent. To address this issue, we developed an algorithm that automatically generates a semantic segmentation mask of TB from the TB target detection boundary box. Pulmonologists only need to identify and label the location of TB, and the algorithm can automatically generate the semantic segmentation mask of TB lesions in the labelled area. The algorithm, first, calculates the optimal threshold for separating the lesion from the background region. Then, based on this threshold, the lesion tissue within the bounding box is extracted and forms a mask that can be used for semantic segmentation tasks. Finally, we use the generated TB semantic segmentation mask to train Unet and Vnet models to verify the effectiveness of the algorithm. The experimental results demonstrate that Unet and Vnet achieve mean Dice coefficients of 0.612 and 0.637, respectively, in identifying TB lesion tissue.

A novel semi-supervised fault diagnosis method for chillers based on neighbor-optimized graph convolutional network

The fault diagnosis of chillers is of great significance in reducing building energy consumption and extending the operational lifespan of refrigeration equipment. Several popular machine learning-based fault diagnosis methods rely heavily on many labeled samples. However, such samples are difficult to obtain in practice due to sparse fault data and high labeling costs. This limits the application of ML-based fault diagnosis methods based on supervised learning. To reduce the dependence on labeled samples, this paper proposes a novel chiller fault diagnosis method based on neighbor-optimized graph convolutional network. The method improves the utilization efficiency of unlabeled samples by mining the spatio-temporal relationship between a large number of unlabeled samples and a limited number of labeled samples. And it dynamically adjusts the graph's structure by optimizing the number of adjacent samples in the correlation graph to obtain better diagnostic results. Its effectiveness is validated on the authoritative dataset ASHRAE RP-1043 and a more challenging dataset of real-world chillers in a building. Experimental results show that the proposed method can achieve better diagnostic performance than the state-of-the-art methods.

A Deep CNN Framework for UAV Intrusion Detection in Intelligent Systems

Unmanned Ariel Vehicle (UAV) s are dealing with several safety and protection issues including internal hardware/software and potential attacks. In addition, detecting UAV anomalies will be a crucial responsibility to defend against hostile enemies and prevent accidents. In this research, we present a UAV and an Automatic Dependent (AD) system using surveillance and Machine Learning (ML) algorithms to analyze data from their detectors in real-time. Proposed Improved Region based Convolutional Neural Network (IRCNN) model used to generate and acquire the characteristics of untreated sensor information and characteristics to facilitate AD. The proposed model creating an Inertial Measurement Unit (IMU) & UAV sensors dataset using cyber security simulation system and Active Learning (AL) identifies aggressions based on the least probable interrogation method. This proposed model enables the identification to efficiently improve the occurrences of unexplained aggressions discovered of IRCNN at reduced labeling cost. A thorough trial showed that IRCNN-AL is effective at detecting unknown threats with frequency improvements of between 9% and 30% on comparison approaches. The AL methodology presented with as few as 1% of a labeled unexpected aggressions.

Advantages of harmonised front-of-pack nutrition labelling for stakeholders

Abstract The adoption of a harmonised front-of-pack nutrition labelling (FoPNL) scheme has significant advantages from at least four perspectives. From the perspective of consumers, effective FoPNL can help inform them on the nutritional composition of the food they purchase and consume. From the perspective of economic operators, a harmonised FoPNL scheme results in a level playing field, increased legal certainty and reduced labelling costs. From the perspective of Member States, a mandatory, EU-wide scheme will contribute to policies intended to reduce the prevalence of obesity and diet-related diseases, and will facilitate compliance with the commitments they have made at international level to promote healthier food environments. From the perspective of the EU itself, a harmonised FoPNL scheme will promote the proper functioning of the internal market in line with the EU's mandate to ensure a high level of consumer and health protection in all its policies.

Label-free grain segmentation for optical microscopy images via unsupervised image-to-image translation

Grain boundaries play an important role in governing the mechanical and physical properties of polycrystalline materials. Therefore, quantitative analysis of grain structure is an important prerequisite for establishing structure-property relationships. However, this quantitative analysis is hampered by the low contrast and non-uniform illumination of optical microscopy images. Although previous studies have used neural networks to detect grain boundaries in optical micrographs, their practical applications have been restricted due to labeling costs for supervised learning and their sensitivity to defects present in a grain structure image. In this paper, we approach grain boundary detection as a real-to-virtual (R2V) translation problem, mapping a single real microstructure to virtual microstructures. With this perspective, our framework benefits from two learning schemes: unsupervised learning and physics-inspired learning. Extensive experiments demonstrate the superiority and generality of our framework. We expect that our approach will facilitate the use of data-driven techniques in the field of quantitative microstructure analysis.

Green level and environmental labeling strategies in a dual‐channel green supply chain considering consumer trust

AbstractEnvironmental labeling has emerged as a crucial method for consumers to gain information related to the eco‐friendly attributes of a product. This paper examines environmental labeling strategies in a green supply chain. The manufacturer has the option to use either self‐labeling or external certified labeling, and consumers have different levels of trust in the two types of environmental labels. We find that under both strategies, supply chain decision makers should fully consider the impacts of consumer trust in environmental labeling, green research and development (R&D) cost, and the demand share of the direct channel on the supply chain. Besides, the retailer realizes higher profit under external certified labeling. A higher green R&D cost or smaller market proportion of the direct channel would dampen the motivation of the manufacturer to choose external certified labeling. Lastly, we explore the case of a risk‐averse retailer and find that the external certified environmental labeling is optimal when the risk aversion is high. Moreover, under both strategies, the risk‐averse attitude of the retailer leads to a more significant channel utility.

Semi-supervised ensemble fault diagnosis method based on adversarial decoupled auto-encoder with extremely limited labels

Intelligent fault diagnosis can enhance the reliability of mechanical equipment. However, only a few labels are available in a large amount of fault data due to high labeling costs in practical engineering. The fault recognition capability of existing semi-supervised diagnosis methods is severely insufficient with limited labels, especially with extremely limited labels that only a single labeled sample available per fault type. To address this issue, a novel semi-supervised ensemble fault diagnosis framework termed ADAE-LFDM is proposed based on adversarial decoupled auto-encoder (ADAE) and low-dimensional feature distance metric (LFDM). Firstly, the locally selective combination in parallel outlier ensembles (LSCP) method is introduced to efficiently separate normal and fault samples. Subsequently, an ADAE with branching structure and latent space feature regularization strategy is proposed to decouple and capture the fault feature. Finally, a LFDM strategy that contains feature dimensionality reduction, and centroid-based metric is performed to achieve high-accuracy fault diagnosis. Experimental results based on two rotating machinery datasets have demonstrated that the proposed method achieves a diagnostic accuracy of over 97 % when there is only a single labeled sample available per fault type, and an average diagnostic accuracy of 85 % under cross-operating condition, showing the superiority compared to other methods.

Label-free model evaluation and weighted uncertainty sample selection for domain adaptive instance segmentation

This paper addresses the challenges of model evaluation and optimization that arise from domain differences between the target and source domains during model deployment. Current methods for model accuracy evaluation require a fully annotated test set. However, obtaining additional human labels for every unique application scenario can be costly and time-intensive. To tackle this problem, this paper proposes an instance segmentation model evaluation method based on domain differences, which can give the prediction accuracy of the model on unlabeled test sets. Moreover, to enhance deployment accuracy cost-effectively, this paper proposes an “effective operation”-based labeling cost computation method and a weighted uncertainty sample selection method. The former accurately computes labeling costs for instance segmentation, while the latter selects the most valuable samples from the target domain for labeling and training. Model evaluation experiments demonstrate that this method’s root mean square error (RMSE) on Cityscapes is approximately 4% less than that of other existing model evaluation methods. Model optimization experiments demonstrate that the proposed method achieves greater model accuracy than comparative methods under four distinct data partitioning protocols. The code is available at https://github.com/licongguan/Lamer.

Online drift compensation framework based on active learning for gas classification and concentration prediction

Sensor drift is an urgent issue in the machine olfaction community. To date, most studies have focused on gas classification tasks based on an offline method, while neglecting concentration prediction and labeling cost. To permit multitasking including sensor drift, gas classification, concentration prediction, and labeling cost, this paper presents a novel online drift compensation framework based on active learning. Specifically, a Query Strategy for Gas Classification (QSGC) and a Query Strategy for Concentration Prediction (QSCP) are designed respectively, and an Online Domain-adaptive Extreme Learning Machine (ODELM) is proposed. First, the QSGC/QSCP is employed to select the most valuable samples for labeling in the gas classification task/concentration prediction task. Second, the ODELM utilizes only one labeled sample to update the prediction model, and thus adapts to evolving sensor drift. The proposed framework is compared with several state-of-the-art methods. Experimental results demonstrate that the proposed method achieves the best generalization ability with the minimum labeling cost.

An Online Unsupervised Streaming Features Selection Through Dynamic Feature Clustering

Streaming feature selection (SFS) is emerging as a key research direction which addresses the non-stationary property of feature streams when the sample size is fixed. Most existing SFS techniques are supervised methods, and ignore the label scarcity. Real-world datasets are typically unlabeled and the labeling costs are expensive. Although some unsupervised SFS approaches are proposed, these approaches are either limited to the homogeneous feature types or require substantial computational complexity. To address these problems, we propose an online unsupervised feature selection framework using dynamic feature clustering in this paper. We derived a recursive density lower bound to estimate the density distribution of feature streams and developed a density-based dynamic clustering method to perform the online feature stream clustering for exploring feature redundancy. An unsupervised online feature relevance maximization and redundancy minimization strategy is introduced to extract a subset of important features with low redundancy from the feature stream. Experimental results on thirteen well-known benchmark datasets and comparison studies with seven state-of-the-art supervised SFS methods demonstrate that the proposed unsupervised method provides statistically comparable performance with the supervised SFS techniques while the label information is unknown.

Federated Semi-Supervised Learning for Medical Image Segmentation via Pseudo-Label Denoising.

Distributed big data and digital healthcare technologies have great potential to promote medical services, but challenges arise when it comes to learning predictive model from diverse and complex e-health datasets. Federated Learning (FL), as a collaborative machine learning technique, aims to address the challenges by learning a joint predictive model across multi-site clients, especially for distributed medical institutions or hospitals. However, most existing FL methods assume that clients possess fully labeled data for training, which is often not the case in e-health datasets due to high labeling costs or expertise requirement. Therefore, this work proposes a novel and feasible approach to learn a Federated Semi-Supervised Learning (FSSL) model from distributed medical image domains, where a federated pseudo-labeling strategy for unlabeled clients is developed based on the embedded knowledge learned from labeled clients. This greatly mitigates the annotation deficiency at unlabeled clients and leads to a cost-effective and efficient medical image analysis tool. We demonstrated the effectiveness of our method by achieving significant improvements compared to the state-of-the-art in both fundus image and prostate MRI segmentation tasks, resulting in the highest Dice scores of 89.23% and 91.95% respectively even with only a few labeled clients participating in model training. This reveals the superiority of our method for practical deployment, ultimately facilitating the wider use of FL in healthcare and leading to better patient outcomes.

Autonomous Cross Domain Adaptation Under Extreme Label Scarcity.

A cross domain multistream classification is a challenging problem calling for fast domain adaptations to handle different but related streams in never-ending and rapidly changing environments. Notwithstanding that existing multistream classifiers assume no labeled samples in the target stream, they still incur expensive labeling costs since they require fully labeled samples of the source stream. This article aims to attack the problem of extreme label shortage in the cross domain multistream classification problems where only very few labeled samples of the source stream are provided before process runs. Our solution, namely, Learning Streaming Process from Partial Ground Truth (LEOPARD), is built upon a flexible deep clustering network where its hidden nodes, layers, and clusters are added and removed dynamically with respect to varying data distributions. A deep clustering strategy is underpinned by a simultaneous feature learning and clustering technique leading to clustering-friendly latent spaces. A domain adaptation strategy relies on the adversarial domain adaptation technique where a feature extractor is trained to fool a domain classifier by classifying source and target streams. Our numerical study demonstrates the efficacy of LEOPARD where it delivers improved performances compared to prominent algorithms in 15 of 24 cases. Source codes of LEOPARD are shared in https://github.com/wengweng001/LEOPARD.git to enable further study.

An Unsupervised Gradient-Based Approach for Real-Time Log Analysis From Distributed Systems

We consider the problem of real-time log anomaly detection for distributed system with deep neural networks by unsupervised learning. There are two challenges in this problem, including detection accuracy and analysis efficacy. To tackle these two challenges, we propose GLAD, a simple yet effective approach mining for anomalies in distributed systems. To ensure detection accuracy, we exploit the gradient features in a well-calibrated deep neural network and analyze anomalous pattern within log files. To improve the analysis efficacy, we further integrate one-class support vector machine (SVM) into anomalous analysis, which significantly reduces the cost of anomaly decision boundary delineation. This effective integration successfully solves both accuracy and efficacy in real-time log anomaly detection. Also, since anomalous analysis is based upon unsupervised learning, it significantly reduces the extra data labeling cost. We conduct a series of experiments to justify that GLAD has the best comprehensive performance balanced between accuracy and efficiency, which implies the advantage in tackling practical problems. The results also reveal that GLAD enables effective anomaly mining and consistently outperforms state-of-the-art methods on both recall and F1 scores.

Boosting SAR Aircraft Detection Performance with Multi-Stage Domain Adaptation Training

Deep learning has achieved significant success in various synthetic aperture radar (SAR) imagery interpretation tasks. However, automatic aircraft detection is still challenging due to the high labeling cost and limited data quantity. To address this issue, we propose a multi-stage domain adaptation training framework to efficiently transfer the knowledge from optical imagery and boost SAR aircraft detection performance. To overcome the significant domain discrepancy between optical and SAR images, the training process can be divided into three stages: image translation, domain adaptive pretraining, and domain adaptive finetuning. First, CycleGAN is used to translate optical images into SAR-style images and reduce global-level image divergence. Next, we propose multilayer feature alignment to further reduce the local-level feature distribution distance. By applying domain adversarial learning in both the pretrain and finetune stages, the detector can learn to extract domain-invariant features that are beneficial to the learning of generic aircraft characteristics. To evaluate the proposed method, extensive experiments were conducted on a self-built SAR aircraft detection dataset. The results indicate that by using the proposed training framework, the average precision of Faster RCNN gained an increase of 2.4, and that of YOLOv3 was improved by 2.6, which outperformed other domain adaptation methods. By reducing the domain discrepancy between optical and SAR in three progressive stages, the proposed method can effectively mitigate the domain shift, thereby enhancing the efficiency of knowledge transfer. It greatly improves the detection performance of aircraft and offers an effective approach to address the limited training data problem of SAR aircraft detection.

DNAzyme-driven bipedal DNA walker and catalytic hairpin assembly multistage signal amplified electrochemical biosensor based on porous AuNPs@Zr-MOF for detection of Pb2+

As a highly toxic and refractory heavy metal contaminant, Pb2+ seriously endangers human health. The problems of low sensitivity and high cost of signal labeling widely exist in common electrochemical biosensors. Herein, a Pb2+ electrochemical biosensor was constructed using a DNAzyme-driven bipedal DNA Walker and catalytic hairpin assembly as the multistage signal amplification strategy. Compared with Zr-MOF, AuNPs@Zr-MOF has a larger porosity and specific surface area, which can effectively load MB to amplify the current signal. Pb2+ can trigger a dual signal amplification reaction to gradually accumulate the signal of methylene blue/gold nanoparticle @ zirconium-based metal organic frameworks (MB/AuNPs@Zr-MOF) on the electrode. The ingeniously designed sensing strategy realized the analysis of Pb2+ with a wide linear range from 0.05 to 1000 nmol/L and a lower limit of detection (LOD) of 4.65 pmol/L. In addition, the sensor has strong anti-interference ability and can accurately detect Pb2+ in various food samples.

Controlled graph neural networks with denoising diffusion for anomaly detection

Leveraging labels in a supervised learning framework as prior knowledge to enhance network anomaly detection has become a trend. Unfortunately, just a few labels are typically available due to the expensive labeling cost. The limited labeled data might not adequately represent the underlying distribution of the data, making the trained model fail to accurately detect anomalies in test data that fall outside the labeled data’s distribution. Recently, data augmentation has been widely used to address this issue. However, general graph data augmentation methods might lead to suboptimal performance when directly applying to network anomaly detection as they might alter graph semantics. In this paper, we provide a denoising diffusion probabilistic model (DDPM)-based Controlled Graph Neural Networks (ConGNN) that can address the problem of insufficient labeled data. In particular, we propose a graph-specific diffusion model-based generator which can inject the characteristics of a reference node into another source node. This generator is adopted to steer the procedure of neighborhood aggregation in GNN to build a controlled GNN for generating augmented data. Based on these augmented data, we present a data-enclosing hypersphere model with our designed consistency regularization term to perform anomaly detection. Extensive experiments demonstrate the superior performance of our model compared to the state-of-the-art baselines.

Polyaniline and polystyrene sulfonic acid modified on conductive biochar nanoparticle with strong signal enhancement for trace atrazine detection

A label and reagent free and cost effective immunosensing system was newly developed for detection of trace leveled atrazine in water. This was constructed by synthesis of polyaniline (PA) and polystyrene sulfonic acid (PSS) modified biochar (BC) nanoparticle composite (PA-PSS-BC) and coating them on screen printed electrode as conductive and electrochemically active layer, and covalently immobilizing anti-atrazine antibody on electrode for specific target binding. The low cost biochar from sugarcane was highly conductive and favorable for attaching PA and PSS due to large surface area and sp2 carbon structure. Crediting to dopant PSS, PA-PSS-BC achieved significant enhancement of redox current signal even in phosphate buffered saline of neutral pH compared to that of without PSS. And redox current of PA was extremely stable with negligible signal change even after twenty cycles repeated reactions owning to protection by BC. These essentially ensured sensing platform being sensitively performed in neutral solution with exemption of adding external redox reagent. As a result, fairly low concentrations of atrazine could be detected within 5 min, with limit of detection being 0.407 ng/L in a dynamic detection range varying from 1 to 100 ng/L. Moreover, the immunosensor exhibited reasonable signal recoveries for detecting trace leveled atrazine in real water samples, high selectivity, and good stability with dry and cold storage.

One Class One Click: Quasi scene-level weakly supervised point cloud semantic segmentation with active learning

Reliance on vast annotations to achieve leading performance severely restricts the practicality of large-scale point cloud semantic segmentation. For the purpose of reducing data annotation costs, effective labeling schemes are developed and contribute to attaining competitive results under weak supervision strategy. Revisiting current weak label forms, we introduce One Class One Click (OCOC), a low cost yet informative quasi scene-level label, which encapsulates both point-level and scene-level annotations. An active weakly supervised framework is proposed to leverage scarce labels by involving weak supervision from both global and local perspectives. Contextual constraints are imposed by an auxiliary scene classification task, respectively based on global feature embedding and point-wise prediction aggregation, which restricts the model prediction merely to OCOC labels within a sub-cloud. Furthermore, we design a context-aware pseudo labeling strategy, which effectively supplement point-level supervisory signals subject to OCOC labels. Finally, an active learning scheme with a uncertainty measure — temporal output discrepancy is integrated to examine informative samples and provides guidance on sub-clouds query, which is conducive to quickly attaining desirable OCOC annotations and reduces the labeling cost to an extremely low extent. Extensive experimental analysis using three LiDAR benchmarks respectively collected from airborne, mobile and ground platforms demonstrates that our proposed method achieves very promising results though subject to scarce labels. It considerably outperforms genuine scene-level weakly supervised methods by up to 25% in terms of average F1 score and achieves competitive results against full supervision schemes. On terrestrial LiDAR dataset — Semantics3D, using approximately 2‱ of labels, our method achieves an average F1 score of 85.2%, which increases by 11.58% compared to the baseline model. Codes are publicly available at https://github.com/PuzoW/One-Class-One-Click.

RNNtcs: A test case selection method for Recurrent Neural Networks

Recurrent Neural Network (RNN) is a typical feedback neural network, which is particularly effective in processing time-series data tasks such as image description, text generation and classification, etc. However, its safety and quality issues have aroused strong concern in academia and industry. One of the most commonly used methods to ensure RNN-driven software quality is testing and optimizing RNN. Its core idea is to introduce human efforts to label test cases so that many labeled test cases can be used to trigger faults in RNNs. But this brings a huge overhead. Therefore, it is of great importance to study the test case selection method for RNN. Current research on test case selection for RNN has achieved some results from the stateful view of RNN, but there are still some problems, such as a low bug detection rate, poor inclusiveness, and lack of diversity.In this paper, according to the structural characteristics of RNN, we propose a method of RNN test case selection, RNNtcs, which combines clustering and uncertainty, in order to select test cases that can touch RNN bugs as much as possible, thus reducing the cost of labeling. In order to evaluate RNNtcs, we conduct experiments on popular image datasets, text datasets and RNN models. The experimental results show that RNNtcs is superior to the state-of-the-art approach DeepState in bug detection effectiveness, bug detection inclusiveness and bug detection diversity. The test cases selected by RNNtcs can effectively improve the robustness of the RNN model.

Active learning with label quality control.

Training deep neural networks requires a large number of labeled samples, which are typically provided by crowdsourced workers or professionals at a high cost. To obtain qualified labels, samples need to be relabeled for inspection to control the quality of the labels, which further increases the cost. Active learning methods aim to select the most valuable samples for labeling to reduce labeling costs. We designed a practical active learning method that adaptively allocates labeling resources to the most valuable unlabeled samples and the most likely mislabeled labeled samples, thus significantly reducing the overall labeling cost. We prove that the probability of our proposed method labeling more than one sample from any redundant sample set in the same batch is less than 1/k, where k is the number of the k-fold experiment used in the method, thus significantly reducing the labeling resources wasted on redundant samples. Our proposed method achieves the best level of results on benchmark datasets, and it performs well in an industrial application of automatic optical inspection.