Semi-supervised Learning Approach Research Articles

Video violence recognition and localization using a semi-supervised hard attention model

The significant growth of surveillance camera networks necessitates scalable AI solutions to efficiently analyze the large amount of video data produced by these networks. As a typical analysis performed on surveillance footage, video violence detection has recently received considerable attention. The majority of research has focused on improving existing methods using supervised methods, with little, if any, attention to the semi-supervised learning approaches. In this study, a reinforcement learning model is introduced that can outperform existing models through a semi-supervised approach. The main novelty of the proposed method lies in the introduction of a semi-supervised hard attention mechanism. Using hard attention, the essential regions of videos are identified and separated from the non-informative parts of the data. A model’s accuracy is improved by removing redundant data and focusing on useful visual information in a higher resolution. Implementing hard attention mechanisms using semi-supervised reinforcement learning algorithms eliminates the need for attention annotations in video violence datasets, thus making them readily applicable. The proposed model utilizes a pre-trained I3D backbone to accelerate and stabilize the training process. The proposed model achieved state-of-the-art accuracy of 90.4% and 98.7% on RWF and Hockey datasets, respectively.

A microRNA Prognostic Signature in Patients with Diffuse Intrinsic Pontine Gliomas through Non-Invasive Liquid Biopsy.

Simple SummaryDiffuse intrinsic pontine glioma (DIPG) is a neuro-radiologically defined tumor of the brainstem, primarily affecting children, with most diagnoses occurring between 5 and 7 years of age. Surgical removal in DIPGs is not feasible. Subsequent tumor progression is almost universal and no biomarker for predicting the course of the disease has entered into clinical practice so far. Under these premises, it is essential to develop reliable biomarkers that are able to improve outcomes and stratify patients using non-invasive methods to determine tumor profiles. We designed a study assessing circulating miRNA expression by a high-throughput platform and divided patients into training and validation phases in order to disclose a potential signature with clinical impact. Our results for the first time have proved the usefulness of blood-circulating nucleic acids as powerful, easy-to-assay molecular markers of disease status in DIPG.Diffuse midline gliomas (DMGs) originate in the thalamus, brainstem, cerebellum and spine. This entity includes tumors that infiltrate the pons, called diffuse intrinsic pontine gliomas (DIPGs), with a rapid onset and devastating neurological symptoms. Since surgical removal in DIPGs is not feasible, the purpose of this study was to profile circulating miRNA expression in DIPG patients in an effort to identify a non-invasive prognostic signature with clinical impact. Using a high-throughput platform, miRNA expression was profiled in serum samples collected at the time of MRI diagnosis and prior to radiation and/or systemic therapy from 47 patients enrolled in clinical studies, combining nimotuzumab and vinorelbine with concomitant radiation. With progression-free survival as the primary endpoint, a semi-supervised learning approach was used to identify a signature that was also tested taking overall survival as the clinical endpoint. A signature comprising 13 circulating miRNAs was identified in the training set (n = 23) as being able to stratify patients by risk of disease progression (log-rank p = 0.00014; HR = 7.99, 95% CI 2.38–26.87). When challenged in a separate validation set (n = 24), it confirmed its ability to predict progression (log-rank p = 0.00026; HR = 5.51, 95% CI 2.03–14.9). The value of our signature was also confirmed when overall survival was considered (log-rank p = 0.0021, HR = 4.12, 95% CI 1.57–10.8). We have identified and validated a prognostic marker based on the expression of 13 circulating miRNAs that can shed light on a patient’s risk of progression. This is the first demonstration of the usefulness of nucleic acids circulating in the blood as powerful, easy-to-assay molecular markers of disease status in DIPG. This study provides Class II evidence that a signature based on 13 circulating miRNAs is associated with the risk of disease progression.

Detect Surface Cracks Using Explainable Neural Networks

This presentation explores an end-to-end AI workflow by implementing and validating an explainable single-class classification algorithm. The neural network accurately detects cracks in images of concrete, while returning a heatmap of anomalies. A semi-supervised one-class learning approach is taken, meaning that the network is trained on data consisting only of examples of images without anomalies. Despite training only on samples of normal scenes, the model learns how to distinguish anomalies. One-class learning offers many advantages for anomaly detection problems such as: 1. Representations of anomalies can be scarce 2. Anomalies can represent expensive or catastrophic outcomes 3. There can be many kinds of anomalies, and these can change over the lifetime of the model. Describing what "good" looks like is often more feasible than providing data that represents all possible anomalies in real world settings A crucial part of detection is for a human observer to be able to understand why a trained network classifies images as anomalies. Explainable classification supplements the class prediction with information that justifies how the neural network reached its decision. The presentation further explores how to deploy the algorithm onto a portable device, so that the classification could happen, in real-time, on-prem.

Semi-supervised learning for the spatial extrapolation of soil information

Digital soil mapping (DSM) can be used to predict soils at unvisited sites, but problems arise when predictions are needed in areas without any soil observations. In such situations, DSM can still extend the results from reference areas with soil data to target areas that are alike in terms of soil-forming factors and obey the same rules. Such DSM methods have low accuracy due to the complexity of spatial variation in soil, and the difficulty of matching soil-forming factors exactly between reference and target areas. A new approach for extrapolating soil information from reference to target areas is proposed in the current research. We evaluated the ability of a semi-supervised learning (SSLR→T) approach compared to a supervised learning (SLR→T) approach for extrapolating soil classes in two areas (reference and target areas) in central Iran. The SSLR→T used soil observations from the reference area and covariates from both areas. Then, the learned knowledge produced by SSLR→T was transferred to the target area to estimate soil classes. The findings revealed that SSLR→T resulted in higher overall accuracy (0.65) and kappa index (0.44) in the target area compared to the SLR→T (overall accuracy = 0.40 and kappa index = 0.18). Furthermore, the SSLR→T produced the lower values of the confusion index (mean = 0.66) compared to the SLR→T (mean = 0.80). This indicated that the SSLR→T could not only increase the accuracy but also decrease the uncertainty of the soil class predictions, compared to the spatial extrapolation predictions derived from the SLR→T. Generally, these findings indicated that leveraging covariate information from the target area during the training of DSM models in the reference area could successfully improve the generalization power of the models, indicating the effectiveness of SSLR→T for spatial extrapolation.

A Semi-Supervised Learning Approach for Automatic Detection and Fashion Product Category Prediction with Small Training Dataset Using FC-YOLOv4

Over the past few decades, research on object detection has developed rapidly, one of which can be seen in the fashion industry. Fast and accurate detection of an E-commerce fashion product is crucial to choosing the appropriate category. Nowadays, both new and second-hand clothing is provided by E-commerce sites for purchase. Therefore, when categorizing fashion clothing, it is essential to categorize it precisely, regardless of the cluttered background. We present recently acquired tiny product images with various resolutions, sizes, and positions datasets from the Shopee E-commerce (Thailand) website. This paper also proposes the Fashion Category—You Only Look Once version 4 model called FC-YOLOv4 for detecting multiclass fashion products. We used the semi-supervised learning approach to reduce image labeling time, and the number of resulting images is then increased through image augmentation. This approach results in reasonable Average Precision (AP), Mean Average Precision (mAP), True or False Positive (TP/FP), Recall, Intersection over Union (IoU), and reliable object detection. According to experimental findings, our model increases the mAP by 0.07 percent and 40.2 percent increment compared to the original YOLOv4 and YOLOv3. Experimental findings from our FC-YOLOv4 model demonstrate that it can effectively provide accurate fashion category detection for properly captured and clutter images compared to the YOLOv4 and YOLOv3 models.

Simultaneous label inference and discriminant projection estimation through adaptive self-taught graphs

Processing structured data has become an interesting topic in recent years. The development of graph-based semi-supervised learning models has attracted much attention from machine learning researchers. In this paper, we present a novel approach for graph-based semi-supervised learning. We provide an effective method for simultaneous label recovery and linear transformation estimation. The targeted linear transformation is to obtain a discriminant subspace. The most important factor in this work to improve the semi-supervised learning is to exploit the data structure and soft labels of the available unlabeled samples. In the iterative optimization scheme used, the prior estimation of the labels increases the monitoring information in an indirect way through an introduced label-graph, avoiding the use of confidence-based hard decisions as used in self-supervised methods. It also enforces label smoothing and projected data smoothing through the use of hybrid graphs. For each smoothing type, the hybrid graph is an adaptive fusion of the two graphs encoding the similarity of the data information and the similarity of the label information. The proposed method leads to an improved discriminant linear transformation. Several experimental results on real image datasets confirm the effectiveness of the proposed method. This work also shows superior performance compared to semi-supervised methods that use simultaneous embedding and inference of labels.

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.

Walking a Tightrope: CEO Emphasis on Digital Transformation and CEO Dismissal

While CEOs’ commitment is crucial to the success of digital transformation, surveys show that CEOs of incumbent firms underemphasize the urgency and challenges of digital transformation. We suggest that career penalties may discourage CEOs from emphasizing digital transformation. We argue that this is driven by governance actors’ attribute substitution, the tendency to unconsciously substitute difficult questions with easier ones. To test these arguments, we apply a semi-supervised machine learning approach to conference calls and create a novel measure for CEOs’ emphasis on digital transformation. We find that CEOs that emphasize digital transformation have a higher probability to be dismissed if financial performance suffers. We find that this relation is stronger if actors within the firm’s governance context possess less digital and more financial expertise, and are busier. Further, we find that firms with CEOs sensitive to career concerns introduce fewer digital products and that this effect is mediated by CEOs’ emphasis on digital transformation. These results suggest that the biased evaluation of CEOs that emphasize digital transformation may impede the transformation of incumbent firms.

Who gets a say in this? Speaking security on social media

Can social media revert the top-down dynamics of securitization? Limited by the notion that ‘security is only articulated in an institutional voice by the elites’, the role of non-elite actors has remained understudied. Only recently has it been proposed that lay actors can become influential security agents through their online activity. However, social media’s capacity to revert the top-down dynamics of securitization remains contended. To explore this puzzle and seeking to update the theory of securitization to the modern context of political communication, this study employs a semi-supervised machine learning approach to analyse a novel dataset of over 10 million Twitter messages by five elite and non-elite actor groups discussing the Amazon rainforest fires in 2019. Finally, the study uses vector autoregression (VAR) models to explore who leads and who echoes the securitization process. The results show that both elite and lay actors behave as security agents and demonstrate the methodological contribution offered by the text-as-data approach developed in this analysis.

Identifying infected patients using semi-supervised and transfer learning.

Early identification of infection improves outcomes, but developing models for early identification requires determining infection status with manual chart review, limiting sample size. Therefore, we aimed to compare semi-supervised and transfer learning algorithms with algorithms based solely on manual chart review for identifying infection in hospitalized patients. This multicenter retrospective study of admissions to 6 hospitals included "gold-standard" labels of infection from manual chart review and "silver-standard" labels from nonchart-reviewed patients using the Sepsis-3 infection criteria based on antibiotic and culture orders. "Gold-standard" labeled admissions were randomly allocated to training (70%) and testing (30%) datasets. Using patient characteristics, vital signs, and laboratory data from the first 24 hours of admission, we derived deep learning and non-deep learning models using transfer learning and semi-supervised methods. Performance was compared in the gold-standard test set using discrimination and calibration metrics. The study comprised 432 965 admissions, of which 2724 underwent chart review. In the test set, deep learning and non-deep learning approaches had similar discrimination (area under the receiver operating characteristic curve of 0.82). Semi-supervised and transfer learning approaches did not improve discrimination over models fit using only silver- or gold-standard data. Transfer learning had the best calibration (unreliability index P value: .997, Brier score: 0.173), followed by self-learning gradient boosted machine (P value: .67, Brier score: 0.170). Deep learning and non-deep learning models performed similarly for identifying infection, as did models developed using Sepsis-3 and manual chart review labels. In a multicenter study of almost 3000 chart-reviewed patients, semi-supervised and transfer learning models showed similar performance for model discrimination as baseline XGBoost, while transfer learning improved calibration.

Dlsa: Semi-supervised partial label learning via dependence-maximized label set assignment

Semi-supervised partial label learning is an emerging weakly supervised learning paradigm dealing with partially labeled data and unlabeled data simultaneously. The supervision information is weak and even missing for training instances. Most existing researches simply propagate labels to unlabeled data and focus on inducing a weighted classifier model, ignoring the importance of acquiring reliable label confidences for training instances. In this paper, by label set assignment, the valid label supervision information can be preserved and propagated to unlabeled instances to a greater extent. Then, the training instances are projected into a lower-dimensional feature space. Reliable label confidence can be obtained by maximizing the dependence between the feature description and the associated label information. After that, the predictive model can be induced by preserving the intrinsic manifold structure in both feature space and label space. Accordingly, a novel semi-supervised partial label learning approach via dependence-maximized label set assignment (Dlsa) is proposed, where an accurate and robust model can be induced based on reliable label confidences obtained by label set assignment and dependence-maximized dimensionality reduction. Extensive experiments on five real-world data sets validate the effectiveness and superiority of the proposed method.

FIEMA, a system of fuzzy inference and emission analytics for sustainability-oriented chemical process design

In the quest to achieve sustainable development goals, developments in sustainability-oriented chemical process design are key to innovation in the chemical industry, especially important for processes aiming for sustainable fuels. One of the greatest challenges is the difficulty of modeling the highly complex interactions among the design variables, such as catalyst technology attributes, and greenhouse gas emissions. Most of the computational aids crucial to deal with the complexity of chemical processes require data that is either unavailable or uncertain at an early stage of design. The multistage integrated system for sustainable design proposed in this paper boosts these computational aids by applying data science techniques to allow uncertainty to be handled more efficiently, thereby facilitating the modeling of the interactions between the properties of new materials or processes and sustainability indicators. In this system, current data connectivity methods are used to find paths of correlation among catalysts properties and greenhouse gas emissions. The key feature of the proposed system relies on the integration through multiple stages of Fuzzy Inference systems and a data-driven technique for Emissions Analytics, FIEMA.11FIEMA: Fuzzy Inference and Emission Analytics. The algorithm in FIEMA provides a semi-supervised learning approach to emission analytics: it determines data clusters by a C-means algorithm and subsequently builds fuzzy sets for multiple stages of input–output inference. The proposed FIEMA system was demonstrated in an effort to determine the optimal configurations of the properties of catalysts to minimize the probability of associated greenhouse gas emissions for a methanol production process. The results showed the potential of this approach to reduce the search space of catalyst material designs by suggesting promising configurations for oxygen storage capacity, mechanical strength, lifetime, size, and poisoning level. The research impacts of this study contribute to the development of clean fuels by a computationally-efficient system for early design, and by the determination of catalysts development paths that assure an actual reduction of the life-cycle emissions.

On robust risk-based active-learning algorithms for enhanced decision support

Classification models are a fundamental component of physical-asset management technologies such as structural health monitoring (SHM) systems and digital twins. Previous work introduced risk-based active learning, an online approach for the development of statistical classifiers that takes into account the decision-support context in which they are applied. Decision-making is considered by preferentially querying data labels according to expected value of perfect information (EVPI). Although several benefits are gained by adopting a risk-based active learning approach, including improved decision-making performance, the algorithms suffer from issues relating to sampling bias as a result of the guided querying process. This sampling bias ultimately manifests as a decline in decision-making performance during the later stages of active learning, which in turn corresponds to lost resource/utility.The current paper proposes two novel approaches to counteract the effects of sampling bias: semi-supervised learning, and discriminative classification models. These approaches are first visualised using a synthetic dataset, then subsequently applied to an experimental case study, specifically, the Z24 Bridge dataset. The semi-supervised learning approach is shown to have variable performance; with robustness to sampling bias dependent on the suitability of the generative distributions selected for the model with respect to each dataset. In contrast, the discriminative classifiers are shown to have excellent robustness to the effects of sampling bias. Moreover, it was found that the number of inspections made during a monitoring campaign, and therefore resource expenditure, could be reduced with the careful selection of the statistical classifiers used within a decision-supporting monitoring system.

Semi-automatic data annotation based on feature-space projection and local quality metrics: An application to cerebral emboli characterization.

We propose a semi-supervised learning approach to annotate a dataset with reduced requirements for manual annotation and with controlled annotation error. The method is based on feature-space projection and label propagation using local quality metrics. First, an auto-encoder extracts the features of the samples in an unsupervised manner. Then, the extracted features are projected by a t-distributed stochastic neighbor embedding algorithm into a two-dimensional (2D) space. A selection of the best 2D projection is introduced based on the silhouette score. The expert annotator uses the obtained 2D representation to manually label samples. Finally, the labels of the labeled samples are propagated to the unlabeled samples using a K-nearest neighbor strategy and local quality metrics. We compare our method against semi-supervised optimum-path forest and K-nearest neighbor label propagation (without considering local quality metrics). Our method achieves state-of-the-art results on three different datasets by labeling more than 96% of the samples with an annotation error from 7% to 17%. Additionally, our method allows to control the trade-off between annotation error and number of labeled samples. Moreover, we combine our method with robust loss functions to compensate for the label noise introduced by automatic label propagation. Our method allows to achieve similar, and even better, classification performances compared to those obtained using a fully manually labeled dataset, with up to 6% in terms of classification accuracy.

A semi-supervised learning approach for variance reduction in life insurance

A semi-supervised learning approach for variance reduction in life insurance

A Semi-supervised Learning Approach with Two Teachers to Improve Breakdown Identification in Dialogues

Identifying breakdowns in ongoing dialogues helps to improve communication effectiveness. Most prior work on this topic relies on human annotated data and data augmentation to learn a classification model. While quality labeled dialogue data requires human annotation and is usually expensive to obtain, unlabeled data is easier to collect from various sources. In this paper, we propose a novel semi-supervised teacher-student learning framework to tackle this task. We introduce two teachers which are trained on labeled data and perturbed labeled data respectively. We leverage unlabeled data to improve classification in student training where we employ two teachers to refine the labeling of unlabeled data through teacher-student learning in a bootstrapping manner. Through our proposed training approach, the student can achieve improvements over single-teacher performance. Experimental results on the Dialogue Breakdown Detection Challenge dataset DBDC5 and Learning to Identify Follow-Up Questions dataset LIF show that our approach outperforms all previous published approaches as well as other supervised and semi-supervised baseline methods.

A semi-supervised learning approach for COVID-19 detection from chest CT scans

COVID-19 has spread rapidly all over the world and has infected more than 200 countries and regions. Early screening of suspected infected patients is essential for preventing and combating COVID-19. Computed Tomography (CT) is a fast and efficient tool which can quickly provide chest scan results. To reduce the burden on doctors of reading CTs, in this article, a high precision diagnosis algorithm of COVID-19 from chest CTs is designed for intelligent diagnosis. A semi-supervised learning approach is developed to solve the problem when only small amount of labelled data is available. While following the MixMatch rules to conduct sophisticated data augmentation, we introduce a model training technique to reduce the risk of model over-fitting. At the same time, a new data enhancement method is proposed to modify the regularization term in MixMatch. To further enhance the generalization of the model, a convolutional neural network based on an attention mechanism is then developed that enables to extract multi-scale features on CT scans. The proposed algorithm is evaluated on an independent CT dataset of the chest from COVID-19 and achieves the area under the receiver operating characteristic curve (AUC) value of 0.932, accuracy of 90.1%, sensitivity of 91.4%, specificity of 88.9%, and F1-score of 89.9%. The results show that the proposed algorithm can accurately diagnose whether a chest CT belongs to a positive or negative indication of COVID-19, and can help doctors to diagnose rapidly in the early stages of a COVID-19 outbreak.

Semi-supervised and personalized federated activity recognition based on active learning and label propagation

One of the major open problems in sensor-based Human Activity Recognition (HAR) is the scarcity of labeled data. Among the many solutions to address this challenge, semi-supervised learning approaches represent a promising direction. However, their centralized architecture incurs in the scalability and privacy problems that arise when the process involves a large number of users. Federated learning (FL) is a promising paradigm to address these problems. However, the FL methods that have been proposed for HAR assume that the participating users can always obtain labels to train their local models (i.e., they assume a fully supervised setting). In this work, we propose FedAR: a novel hybrid method for HAR that combines semi-supervised and federated learning to take advantage of the strengths of both approaches. FedAR combines active learning and label propagation to semi-automatically annotate the local streams of unlabeled sensor data, and it relies on FL to build a global activity model in a scalable and privacy-aware fashion. FedAR also includes a transfer learning strategy to fine-tune the global model on each user. We evaluated our method on two public datasets, showing that FedAR reaches recognition rates and personalization capabilities similar to state-of-the-art FL supervised approaches. As a major advantage, FedAR only requires a very limited number of annotated data to populate a pre-trained model and a small number of active learning questions that quickly decrease while using the system, leading to an effective and scalable solution for the data scarcity problem of HAR.

Comprehensive study of semi-supervised learning for DNA methylation-based supervised classification of central nervous system tumors.

BackgroundPrecision medicine for cancer treatment relies on an accurate pathological diagnosis. The number of known tumor classes has increased rapidly, and reliance on traditional methods of histopathologic classification alone has become unfeasible. To help reduce variability, validation costs, and standardize the histopathological diagnostic process, supervised machine learning models using DNA-methylation data have been developed for tumor classification. These methods require large labeled training data sets to obtain clinically acceptable classification accuracy. While there is abundant unlabeled epigenetic data across multiple databases, labeling pathology data for machine learning models is time-consuming and resource-intensive, especially for rare tumor types. Semi-supervised learning (SSL) approaches have been used to maximize the utility of labeled and unlabeled data for classification tasks and are effectively applied in genomics. SSL methods have not yet been explored with epigenetic data nor demonstrated beneficial to central nervous system (CNS) tumor classification.ResultsThis paper explores the application of semi-supervised machine learning on methylation data to improve the accuracy of supervised learning models in classifying CNS tumors. We comprehensively evaluated 11 SSL methods and developed a novel combination approach that included a self-training with editing using support vector machine (SETRED-SVM) model and an L2-penalized, multinomial logistic regression model to obtain high confidence labels from a few labeled instances. Results across eight random forest and neural net models show that the pseudo-labels derived from our SSL method can significantly increase prediction accuracy for 82 CNS tumors and 9 normal controls.ConclusionsThe proposed combination of semi-supervised technique and multinomial logistic regression holds the potential to leverage the abundant publicly available unlabeled methylation data effectively. Such an approach is highly beneficial in providing additional training examples, especially for scarce tumor types, to boost the prediction accuracy of supervised models.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12859-022-04764-1.

A semi-supervised learning approach for bladder cancer grading

Recent advances in semi-supervised learning algorithms (SSL) have made great strides in reducing the training dependency on labeled datasets and requiring that only a subset of the data be labeled. The presented work explores a class of semi-supervised learning algorithms that uses consistency regularization and self-ensembling to leverage the unlabeled portion of the dataset. Labeling medical image datasets are time-consuming and prohibitively expensive, requiring hundreds of hours of effort from expert diagnosticians. This research presents an approach for building and training a deep learning model to grade medical images while requiring only a minimal number of labels. Consistency regularization has been used in SSL to great success in datasets of natural images but not for more complex images such as pathology slides where the dataset consists of cell patterns. This research successfully proposes and applies an SSL algorithm based on the VGG-16 neural network, which combines techniques introduced by the Π model and FixMatch algorithms to a cell pattern-based pathology image dataset. The results presented in this research show that using the proposed approach, it is possible to label only 3% of the samples in a dataset, use the remaining 97% of samples as unlabeled data, and achieve a 19% increase over the baseline accuracy. The second contribution of this research shows a ratio of labeled vs. unlabeled images in a dataset beyond which continuing to label the data increases the cost but offers little performance gains.