Semi-supervised Generative Adversarial Network Research Articles

Development and Clinical Validation of Semi-Supervised Generative Adversarial Networks for Detection of Retinal Disorders in Optical Coherence Tomography Images Using Small Dataset.

To develop and test semi-supervised generative adversarial networks (GANs) that detect retinal disorders on optical coherence tomography (OCT) images using a small-labeled dataset. From a public database, we randomly chose a small supervised dataset with 400 OCT images (100 choroidal neovascularization, 100 diabetic macular edema, 100 drusen, and 100 normal) and assigned all other OCT images to unsupervised dataset (107,912 images without labeling). We adopted a semi-supervised GAN and a supervised deep learning (DL) model for automatically detecting retinal disorders from OCT images. The performance of the 2 models was compared in 3 testing datasets with different OCT devices. The evaluation metrics included accuracy, sensitivity, specificity, and the area under the receiver operating characteristic curves. The local validation dataset included 1000 images with 250 from each category. The independent clinical dataset included 366 OCT images using Cirrus OCT Shanghai Shibei Hospital and 511 OCT images using RTVue OCT from Xinhua Hospital respectively. The semi-supervised GANs classifier achieved better accuracy than supervised DL model (0.91 vs 0.86 for local cell validation dataset, 0.91 vs 0.86 in the Shanghai Shibei Hospital testing dataset, and 0.93 vs 0.92 in Xinhua Hospital testing dataset). For detecting urgent referrals (choroidal neo-vascularization and diabetic macular edema) from nonurgent referrals (drusen and normal) on OCT images, the semi-supervised GANs classifier also achieved better area under the receiver operating characteristic curves than supervised DL model (0.99 vs 0.97, 0.97 vs 0.96, and 0.99 vs 0.99, respectively). A semi-supervised GAN can achieve better performance than that of a supervised DL model when the labeled dataset is limited. The current study offers utility to various research and clinical studies using DL with relatively small datasets. Semi-supervised GANs can detect retinal disorders from OCT images using relatively small dataset.

Leveraging Google Earth Engine and Semi-Supervised Generative Adversarial Networks to Assess Initial Burn Severity in Forest

Mapping and monitoring initial burn severity is a critical aspect of forest management and landscape restoration. Recently, supervised learning has achieved great success in evaluating the post-fire forest condition, but plenty of labeled samples are needed to carry out training for supervised learning. However, due to the limitation of accessible labeled data, it is often arduous to put supervised learning into practice in the remote sensing field. In this paper, a novel semi-unsupervised image classification framework by using generative adversarial networks under the Google Earth Engine (GEE) platform is proposed to undertake the burn severity assessment with limited samples. The generative model can produce additional adversarial samples that look like the real samples; therefore, the discriminative model gains better classification capabilities in burn severity assessment with the extra training data provided by the generator. The detailed evaluation and comparative experiments are done in the context of post-fire assessment on eleven forest fires in northern California in the United States. The experimental results demonstrate that our approach significantly outperforms the two other most well-known methods.

Semi-Supervised Fake Reviews Detection based on AspamGAN

With the popularization of social software and e-business in recent years, more and more consumers like to share their consumption experiences on social networks and refer to other consumers' reviews and opinions when making consumption decisions. Online reviews have become an essential part of browsing on websites such as shopping, and people's reliance on informative reviews have contributed to the rise of fake reviews. The traditional classification method is affected by the label dataset, which is not only time-consuming, laborious, and subjective, but also the extraction of artificial features also affects the classification accuracy. Due to the relative length of the online text, the possibility of the classifier losing important information increases, this weakens the model’s detection capability. To solve this aforementioned problem, a semi-supervised Generative Adversarial Network (AspamGAN) fake reviews detection method incorporating an attention mechanism is proposed. Using labeled and unlabeled data to correctly learn input distributions, the features required for classification are automatically discovered using deep neural networks, providing better prediction accuracy for online reviews. The approach includes attention mechanisms in the classifier to obtain an adequate semantic representation and relies on a limited dataset of labeled data to detect false reviews, and is applied on the TripAdvisor dataset. Experimental results show that the proposed algorithm outperforms state-of-the-art semi-supervised fake review detection techniques when the label dataset is limited.

A novel semi-supervised generative adversarial network based on the actor-critic algorithm for compound fault recognition

A novel semi-supervised generative adversarial network based on the actor-critic algorithm for compound fault recognition

Semi-supervised generative adversarial networks for anomaly detection

Advancements in security have provided ways of recording anomalies of daily life through video surveillance. For the present investigation, a semi-supervised generative adversarial network model to detect and classify different types of crimes on videos. Additionally, we intend to tackle one of the most recurring difficulties of anomaly detection: illumination. For this, we propose a light augmentation algorithm based on gamma correction to help the semi-supervised generative adversarial networks on its classification task. The proposed process performs slightly better than other proposed models.

Diagnosis of Scoliosis Using Chest Radiographs with a Semi-Supervised Generative Adversarial Network.

To develop and validate a deep learning-based screening tool for the early diagnosis of scoliosis using chest radiographs with a semi-supervised generative adversarial network (GAN). Using a semi-supervised learning framework with a GAN, a screening tool for diagnosing scoliosis was developed and validated through the chest PA radiographs of patients at two different tertiary hospitals. Our proposed method used training GAN with mild to severe scoliosis only in a semi-supervised manner, as an upstream task to learn scoliosis representations and a downstream task to perform simple classification for differentiating between normal and scoliosis states sensitively. The area under the receiver operating characteristic curve, negative predictive value (NPV), positive predictive value, sensitivity, and specificity were 0.856, 0.950, 0.579, 0.985, and 0.285, respectively. Our deep learning-based artificial intelligence software in a semi-supervised manner achieved excellent performance in diagnosing scoliosis using the chest PA radiographs of young individuals; thus, it could be used as a screening tool with high NPV and sensitivity and reduce the burden on radiologists for diagnosing scoliosis through health screening chest radiographs.

Localizing Microseismic Events Using Semi-Supervised Generative Adversarial Networks

The performance of the microseismic monitoring technique depends greatly on the accuracy of microseismic event localization. Recently, machine learning (ML) methods have been extensively implemented for the localization of microseismic events. These neural networks are typically trained using numerous microseismic events labeled with known source locations. Obtaining enough microseismic events with good source locations can be difficult and costly. To overcome this shortcoming, we present a microseismic events localization method using semi-supervised generative adversarial networks (GANs). We utilize limited labeled seismograms and large amounts of unlabeled seismograms to train the semi-supervised GANs, thus improving the prediction ability of the networks. Finally, we evaluate the performance of the proposed method using synthetic microseismic data and field data. Comparison with the supervised learning methods on the same microseismic data shows that the proposed method can significantly improve the accuracy of locating microseismic sources in the lack of sufficient source labels.

IKnight–Guarding IoT Infrastructure Using Generative Adversarial Networks

With the phenomenal growth of IoT devices and their exponentially increasing applications at the edge of the network, it has become imminent to provide secure, flexible, and programmable network services to support user privacy, security, and quality of service. However, such services can only be enabled and effectively applied when the edge systems automatically identify these devices. Existing IoT classification systems are based on supervised training methods that require labeled data and manual feature extraction. Such approaches suffer from many challenges such as privacy, labeling efforts, and adaptability to new devices. This paper develops a multistage multi-class classifier based on semi-supervised generative adversarial networks that perform automatic feature extraction with minimal labeled data. The classifier can identify devices with 96% accuracy when only 3% of the training data is labelled. Moreover, the classifier can infer the device type (IoT, Non-IoT, and anomaly) of any new device correctly with 90% accuracy. We also show how our model can support various features such as noise robustness, continual learning ability, and novelty detection, such as zero-day malware attacks. Finally, we integrate our classifier with a software-defined network to enable smart IoT network services at the edge.

HyperViTGAN: Semisupervised Generative Adversarial Network With Transformer for Hyperspectral Image Classification

Generative adversarial networks (GANs) have achieved many excellent results in hyperspectral image (HSI) classification in recent years, as GANs can effectively solve the dilemma of limited training samples in HSI classification. However, due to the class imbalance problem of HSI data, GANs always associate minority-class samples with fake label. To address this issue, we first propose a semi-supervised generative adversarial network incorporating a transformer, called HyperViTGAN. The proposed HyperViTGAN is designed with an external semi-supervised classifier to avoid self-contradiction when the discriminator performs both classification and discrimination tasks. The generator and discriminator with skip connection are utilized to generate HSI patches by adversarial learning. The proposed HyperViTGAN captures semantic context and low-level textures to reduce the loss of critical information. In addition, the generalization ability of HyperViTGAN is improved through the use of data augmentation. Experimental results on three well-known HSI data sets, Houston 2013, Indian Pines 2010, and Xuzhou, show that the proposed model achieves competitive HSI classification performance in comparison with the current state-of-the-art classification models.

A Semi-Supervised Generative Adversarial Network for Amodal Instance Segmentation of Piglets in Farrowing Pens

A Semi-Supervised Generative Adversarial Network for Amodal Instance Segmentation of Piglets in Farrowing Pens

Hyper-Parameter Optimization of Semi-Supervised GANs Based-Sine Cosine Algorithm for Multimedia Datasets

Generative Adversarial Networks (GANs) are neural networks that allow models to learn deep representations without requiring a large amount of training data. Semi-Supervised GAN Classifiers are a recent innovation in GANs, where GANs are used to classify generated images into real and fake and multiple classes, similar to a general multi-class classifier. However, GANs have a sophisticated design that can be challenging to train. This is because obtaining the proper set of parameters for all models-generator, discriminator, and classifier is complex. As a result, training a single GAN model for different datasets may not produce satisfactory results. Therefore, this study proposes an SGAN model (Semi-Supervised GAN Classifier). First, a baseline model was constructed. The model was then enhanced by leveraging the Sine-Cosine Algorithm and Synthetic Minority Oversampling Technique (SMOTE). SMOTE was used to address class imbalances in the dataset, while Sine Cosine Algorithm (SCA) was used to optimize the weights of the classifier models. The optimal set of hyperparameters (learning rate and batch size) were obtained using grid manual search. Four well-known benchmark datasets and a set of evaluation measures were used to validate the proposed model. The proposed method was then compared against existing models, and the results on each dataset were recorded and demonstrated the effectiveness of the proposed model. The proposed model successfully showed improved test accuracy scores of 1%, 2%, 15%, and 5% on benchmarking multimedia datasets; Modified National Institute of Standards and Technology (MNIST) digits, Fashion MNIST, Pneumonia Chest X-ray, and Facial Emotion Detection Dataset, respectively.

Semi-supervised learning for optical fiber sensor road intrusion signal detection.

This paper proposes a road intrusion detection model based on distributed optical fiber vibration sensors signals. Considering that the existing unsupervised classification method often has a high false alarm rate when meeting the new non-intrusion samples, we propose a one-dimensional semi-supervised generative adversarial network (1D-SSGAN) model for intrusion signal recognition. The 1D-SSGAN is composed of a generator and a discriminator. The output layer of the discriminator is mapped to N+1 classes, and the generator and discriminator are trained on the N class dataset. During the learning process of the generator against the discriminator, many new samples are generated based on a small number of samples, which effectively expands the datasets and assists the training of the discriminator. Experimental result analysis demonstrates the effectiveness of the proposed model.

A semi-supervised GAN method for RUL prediction using failure and suspension histories

Deep learning methods have shown great potential to provide reliable remaining useful life (RUL) predictions in Prognostics and Health Management applications. However, deep learning models, particularly supervised learning methods, are strongly dependent on a large number of failure histories. In practice, engineering assets are generally replaced by new ones before failure during planned maintenance, resulting in a small number of failure histories and often times more than twice as many suspension histories. In this paper, a semi-supervised generative adversarial network (GAN) regression model is developed to consider both failure and suspension histories for RUL predictions. The proposed GAN model utilizes conditional multi-task objective functions to capture useful information from suspension histories to improve prediction accuracy, instead of simply treating them as unlabeled data. The method will not directly predict the failure times of suspension histories, but match the statistical information between similar failure and suspension histories to the greatest extent for model training. As a result, the failure information of suspension histories will not only rely on the failure histories but also on the generated data, thereby improving the model generalization, especially when the amount of data is limited. In addition, a robustness evaluation method is proposed to assess the uncertainty of the prognostic model caused by the scarce failure data. The accuracy and credibility of the proposed approach are validated by using two case studies.

GAN-based reactive motion synthesis with class-aware discriminators for human–human interaction

Creating realistic characters that can react to the users’ or another character’s movement can benefit computer graphics, games and virtual reality hugely. However, synthesizing such reactive motions in human–human interactions is a challenging task due to the many different ways two humans can interact. While there are a number of successful researches in adapting the generative adversarial network (GAN) in synthesizing single human actions, there are very few on modeling human–human interactions. In this paper, we propose a semi-supervised GAN system that synthesizes the reactive motion of a character given the active motion from another character. Our key insights are two-fold. First, to effectively encode the complicated spatial–temporal information of a human motion, we empower the generator with a part-based long short-term memory (LSTM) module, such that the temporal movement of different limbs can be effectively modeled. We further include an attention module such that the temporal significance of the interaction can be learned, which enhances the temporal alignment of the active–reactive motion pair. Second, as the reactive motion of different types of interactions can be significantly different, we introduce a discriminator that not only tells if the generated movement is realistic or not, but also tells the class label of the interaction. This allows the use of such labels in supervising the training of the generator. We experiment with the SBU, the HHOI and the 2C datasets. The high quality of the synthetic motion demonstrates the effective design of our generator, and the discriminability of the synthesis also demonstrates the strength of our discriminator.

Quantum semi-supervised generative adversarial network for enhanced data classification

In this paper, we propose the quantum semi-supervised generative adversarial network (qSGAN). The system is composed of a quantum generator and a classical discriminator/classifier (D/C). The goal is to train both the generator and the D/C, so that the latter may get a high classification accuracy for a given dataset. Hence the qSGAN needs neither any data loading nor to generate a pure quantum state, implying that qSGAN is much easier to implement than many existing quantum algorithms. Also the generator can serve as a stronger adversary than a classical one thanks to its rich expressibility, and it is expected to be robust against noise. These advantages are demonstrated in a numerical simulation.

GAN-based method for generative design of visual comfort in underground space

Existing methods for the generative design of architecture mainly focus on layout planning, and the genetic algorithm is popular in related research. However, the genetic algorithm is not available for non-quantitative problems (e.g., visual comfort) and inefficient for complex problems. In response, recent studies used the generative adversarial network (GAN) as the generative model to realize efficient generation. However, existing GAN-based methods for generative design are not effective enough because the output of their models is a two-dimensional image, which must be reconstructed to a three-dimensional (3D) model in practical engineering applications. In this study, a visual comfort generative network (VCGN) based on parametric modeling and a semi-supervised generative adversarial network (SGAN) is proposed for a more effective generation. The VCGN contains three parts, namely the parametric and teacher models and the SGAN. The parametric model is designed to generate a 3D underground space model by controlling the key parameters. The teacher model rates the visual comfort level of the generated model and trains the SGAN and could be a neural network or another criterion. The SGAN is designed to learn parameter distribution and generate parameters. Different from the SGAN mentioned by Odena, the SGAN used herein comprises four parts: two generators, a discriminator, and a classifier. The first generator is designed to generate parameters, while the second is used to simulate the parametric modeling process. The discriminator is supposed to distinguish simulated results from real results. The classifier is designed to classify the visual comfort levels. Moreover, a spatial color-generation task in underground space is applied as a case study to validate the effectiveness of the VCGN. A comparison of the random sampling method and the existing GAN-based method shows that the VCGN could generate more high-comfort underground space models than other methods for the same amount of time. In addition, a comparison of the underground space models under different comfort levels generated by the VCGN illustrates that the VCGN has the potential to generate an underground space model with a specified comfort level.

Adversarial Sample Detection with Gaussian Mixture Conditional Generative Adversarial Networks

It is important to detect adversarial samples in the physical world that are far away from the training data distribution. Some adversarial samples can make a machine learning model generate a highly overconfident distribution in the testing stage. Thus, we proposed a mechanism for detecting adversarial samples based on semisupervised generative adversarial networks (GANs) with an encoder-decoder structure; this mechanism can be applied to any pretrained neural network without changing the network’s structure. The semisupervised GANs also give us insight into the behavior of adversarial samples and their flow through the layers of a deep neural network. In the supervised scenario, the latent feature (or the discriminator’s output score information) of the semi-supervised GAN and the target network's logit information are used as the input of logistic regression classifier to detect the adversarial samples. In the unsupervised scenario, first, we proposed a one-class classier based on the semisupervised Gaussian mixture conditional generative adversarial network (GM-CGAN) to fit the joint feature information of the normal data, and then, we used a discriminator network to detect normal data and adversarial samples. In both supervised scenarios and unsupervised scenarios, experimental results show that our method outperforms latest methods.

Improving quality prediction in radial-axial ring rolling using a semi-supervised approach and generative adversarial networks for synthetic data generation

As artificial intelligence and especially machine learning gained a lot of attention during the last few years, methods and models have been improving and are becoming easily applicable. This possibility was used to develop a quality prediction system using supervised machine learning methods in form of time series classification models to predict ovality in radial-axial ring rolling. Different preprocessing steps and model implementations have been used to improve quality prediction. A semi-supervised approach is used to improve the prediction and analyze, to what extend it can improve current research in machine learning for quality prediciton. Moreover, first research steps are taken towards a synthetic data generation within the radial-axial ring rolling domain using generative adversarial networks.

CCS-GAN: a semi-supervised generative adversarial network for image classification

CCS-GAN: a semi-supervised generative adversarial network for image classification

Semi-supervised generative adversarial networks for closed-angle detection on anterior segment optical coherence tomography images: an empirical study with a small training dataset.

BackgroundSemi-supervised learning algorithms can leverage an unlabeled dataset when labeling is limited or expensive to obtain. In the current study, we developed and evaluated a semi-supervised generative adversarial networks (GANs) model that detects closed-angle on anterior segment optical coherence tomography (AS-OCT) images using a small labeled dataset.MethodsIn this cross-sectional study, a semi-supervised GANs model was developed for automatic closed-angle detection training on a small labeled and large unsupervised training dataset collected from the Joint Shantou International Eye Center of Shantou University and the Chinese University of Hong Kong (JSIEC). The closed-angle was defined as iris-trabecular contact beyond the scleral spur in AS-OCT images. We further developed two supervised deep learning (DL) models training on the same supervised dataset and the whole dataset separately. The semi-supervised GANs model and supervised DL models’ performance were compared on two independent testing datasets from JSIEC (515 images) and the Department of Ophthalmology (84 images), National University Health System, respectively. The diagnostic performance was assessed by evaluation matrices, including the accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC).ResultsFor closed-angle detection using clinician grading of AS-OCT imaging as the reference standard, the semi-supervised GANs model showed comparable performance, with AUCs of 0.97 (95% CI, 0.96–0.99) and 0.98 (95% CI, 0.94–1.00), compared with the supervised DL model (using the whole dataset) [AUCs of 0.97 (95% CI, 0.96–0.99), and 0.97 (95% CI, 0.94–1.00)]. When training on the same small supervised dataset, the semi-supervised GANs achieved performance at least as well as, if not better than, the supervised DL model [AUCs of 0.90 (95% CI: 0.84–0.96), and 0.92 (95% CI, 0.86–0.97)].ConclusionsThe semi-supervised GANs method achieves diagnostic performance at least as good as a supervised DL model when trained on small labeled datasets. Further development of semi-supervised learning methods could be useful within clinical and research settings.Trial registration numberChiCTR2000037892.