Emergent Complexity over Symbolic Simplicity: Inductive Bias and Structural Failure in GANs

We present a mathematically well-founded approach for the synthetic modeling of turbulent flows using generative adversarial networks (GAN). Based on the analysis of chaotic, deterministic systems in terms of ergodicity, we outline a mathematical proof that GAN can actually learn to sample state snapshots from the invariant measure of the chaotic system. Based on this analysis, we study a hierarchy of chaotic systems starting with the Lorenz attractor and then carry on to the modeling of turbulent flows with GAN. As training data, we use fields of velocity fluctuations obtained from large-eddy simulations (LES). Two architectures are investigated in detail: we use a deep, convolutional GAN (DCGAN) to synthesize the turbulent flow around a cylinder. We furthermore simulate the flow around a low-pressure turbine stator using the pix2pixHD architecture for a conditional DCGAN being conditioned on the position of a rotating wake in front of the stator. The settings of adversarial training and the effects of using specific GAN architectures are explained. We thereby show that GAN are efficient in simulating turbulence in technically challenging flow problems on the basis of a moderate amount of training data. GAN training and inference times significantly fall short when compared with classical numerical methods, in particular, LES, while still providing turbulent flows in high resolution. We furthermore analyze the statistical properties of the synthesized and LES flow fields, which agree excellently. We also show the ability of the conditional GAN to generalize over changes of geometry by generating turbulent flow fields for positions of the wake that are not included in the training data.

Early fault detection (EFD) in run-to-failure processes plays a crucial role in the condition monitoring of modern industrial rotating facilities, which entail increasing demands for safety, energy and ecological savings and efficiency. To enable effective protection measures, the evolving faults have to be recognized and identified as early as possible. The major challenge is to distil discriminative features on the basis of only the ‘health’ signal, which is uniquely available from various possible sensors before damage sets in and before the signatures of incipient damage become obvious and well-understood in the signal. Acoustic emission (AE) signals have been frequently reported to be able to deliver early diagnostic information due to their inherently high sensitivity to the incipient fault activities, highlighting the great potential of the AE technique for EFD, which may outperform the traditional vibration-based analysis in many situations. To date, the ‘feature-based’ multivariate analysis dominates the interpretation of AE waveforms. In this way, the decision-making relies heavily on experts’ knowledge and experience, which is often a weak link in the entire EFD chain. With the advent of artificial intelligence, practitioners seek an intelligent method capable of tackling this challenge. In the present paper, we introduce a versatile approach towards intelligent data analysis adapted to AE signals streaming from the sensors used for the continuous monitoring of rotating machinery. A new architecture with a convolutional generative adversarial network (GAN) is designed to extract the deep information embedded in the AE waveforms. In order to improve the robustness of the proposed EFD framework, a novel ensemble technique referred to as ‘history-state ensemble’ (HSE) is introduced and paired with GAN. The primary merits of HSE are twofold: (1) it does not require extra computing time to obtain the base models, and (2) it does not require a special design of the network architecture and can be applied to different networks. To evaluate the proposed method, a durability rolling contact fatigue test was performed with the use of AE monitoring. The experimental results have demonstrated that the proposed ensemble method largely improves the robustness of GAN.

We propose a novel semi -supervised fully convolutional network for Polarimetric synthetic aperture radar (PoISAR) terrain classification. First, by designing a fully convolutional structure, we can perform pixel-based classification tasks. Then, by applying semi -supervised generative adversarial networks (GANs), we utilize both labeled and unlabeled samples and aim to obtain higher classification accuracy. Through a mini-max two-player game, GAN has better performance than other “single-player” classifiers. Finally, we combine the fully convolutional structure with the semi-supervised GAN. Our fully convolutional semi-supervised GAN (FC-SGAN) has excellent spatial feature learning ability and can perform end-to-end pixel-based classification tasks. Experimental results show that compared with existing works, the proposed method has better performances. Even when the training set gets smaller, our method keeps high accuracy.

Speech Enhancement (SE) system deals with improving the perceptual quality and preserving the speech intelligibility of the noisy mixture. The Time-Frequency (T-F) masking-based SE using the supervised learning algorithm, such as a Deep Neural Network (DNN), has outperformed the traditional SE techniques. However, the notable difference observed between the oracle mask and the predicted mask, motivates us to explore different deep learning architectures. In this paper, we propose to use a Convolutional Neural Network (CNN)-based Generative Adversarial Network (GAN) for inherent mask estimation. GAN takes an advantage of the adversarial optimization, an alternative to the other Maximum Likelihood (ML) optimization-based architectures. We also show the need for supervised T-F mask estimation for effective noise suppression. Experimental results demonstrate that the proposed T-F mask-based SE significantly outperforms the recently proposed end-to-end SEGAN and a GAN-based Pix2Pix architecture. The performance evaluation in terms of both the predicted mask and the objective measures, dictates the improvement in the speech quality, while simultaneously reducing the speech distortion observed in the noisy mixture.

We study two important concepts in adversarial deep learning---adversarial training and generative adversarial network (GAN). Adversarial training is the technique used to improve the robustness of discriminator by combining adversarial attacker and discriminator in the training phase. GAN is commonly used for image generation by jointly optimizing discriminator and generator. We show these two concepts are indeed closely related and can be used to strengthen each other---adding a generator to the adversarial training procedure can improve the robustness of discriminators, and adding an adversarial attack to GAN training can improve the convergence speed and lead to better generators. Combining these two insights, we develop a framework called Rob-GAN to jointly optimize generator and discriminator in the presence of adversarial attacks---the generator generates fake images to fool discriminator; the adversarial attacker perturbs real images to fool the discriminator, and the discriminator wants to minimize loss under fake and adversarial images. Through this end-to-end training procedure, we are able to simultaneously improve the convergence speed of GAN training, the quality of synthetic images, and the robustness of discriminator under strong adversarial attacks. Experimental results demonstrate that the obtained classifier is more robust than the state-of-the-art adversarial training approach, and the generator outperforms SN-GAN on ImageNet-143.

A Generative Adversarial Network (GAN) is an artificial intelligence (AI) technique and neural network design that trains the generator and discriminator neural networks simultaneously using adversarial training. The generator aims to generate more realistic data, whereas the discriminator aims to distinguish actual from generated data. This adversarial process makes the generator provide more realistic data. GANs are widely employed in many fields, such as style transfer, data augmentation, image generation, and, more recently, cybersecurity, where they are used to create synthetic data, mimic cyberthreats, and improve security measures. Potential applications of GANs include strengthening intrusion detection systems, enhancing anomaly detection, and simulating a variety of cyberthreats. In contrast, new complications arise when GANs are included in cybersecurity investigations. Adversarial GAN approaches expose privacy and security vulnerabilities and threats in machine learning (ML) models. The adversarial nature of GANs makes AI and ML applications more vulnerable to assaults, thereby compromising security. This dual nature highlights the necessity for a nuanced and complete approach to utilizing GANs for cybersecurity while tackling increasing problems to preserve AI/ML application robustness in the evolving threat landscape.

A Generative Adversarial Network (GAN) is an artifi- cial intelligence model developed specifically to pro- duce synthetic data that resembles real data by training a generative model and a discriminative model simulta- neously using adversarial training. A GAN can be ex- tensively used for generating replicated data, however, it suffers from several issues, one of which is mode col- lapse. Mode collapse takes place when the generator is unable to capture the complete range of diversity in the target data distribution, resulting in the production of limited and repeating variations of samples. Multiple metrics exist to quantify mode collapse in GANs, al- though no individual metric is capable of consistently providing accurate results. This research focuses on the critical need for accurate mode collapse detection tech- niques in GANs, to strengthen the credit card fraud detection systems. In this work, we utilize a GAN to generate numerical data instead of image data. Our ap- proach utilizes a wide range of measures, such as Gener- ator and Discriminator Loss, Wasserstein Distance, pre- cision, recall, and visualization tools, to provide a com- prehensive framework for detecting mode collapse. In addition, we introduce an alert mechanism that identi- fies possible mode collapse at an early stage, allowing for earlier intervention and modifications to the training process. We have further proposed suggestions regard- ing monitoring and analyzing generator and discrimina- tor loss values to identify potential instances of mode collapse to help the developer optimize GAN training and improve the quality of synthetic data.

RAGAN: A Generative Adversarial Network for risk-aware trajectory prediction in multi-ship encounter situations

paper, we focus on exploring the relationship between weave patterns and their mechanical properties in woven fiber composites through Machine Learning. Specifically, we explore the interactions between woven architectures and in-plane stiffness properties through Deep Convolutional Neural Network (DCNN) and Generative Adversarial Network (GAN). Our research is important for exploring how woven composite’s pattern is related to its mechanical properties and accelerating woven composite design as well as optimization. We focus on two tasks: (1) Stiffness prediction: Predicting in-plane stiffness properties for given weave patterns. Our DCNN extracts high-level features through several convolutional and fully connected layers to determine the final predictions. (2) Weave pattern prediction: Predicting weave patterns for target stiffness properties, which can be treated as the reverse task of the first one. Due to many-to-one mapping between weave patterns and the composite properties, we utilize a Decoder Neural Network as our baseline model and compare its performance with GAN and Genetic Algorithm. We represent the weave patterns as 2D checkerboard models and use finite element analysis (FEA) to determine in-plane stiffness properties, which serve as input data for our ML framework. We show that: (1) for stiffness prediction, DCNN can predict stiffness values for a given weave pattern with relatively high accuracy (above 93%); (2) for weave pattern prediction, the GAN model gives the best prediction accuracy (above 92%) while Decoder Neural Network has the best time efficiency. HAOTIAN FENG

<p>High data collection costs and complicated data access regulations increase the demand for synthetic data. Generative Adversarial Networks (GANs) are a novel generative framework with great potential for high quality synthetic data generation. GANs formulate the true distribution of data implicitly, and the success of GANs are often measured based on the similarity of generated data to this true distribution. GANs were originally designed to work with continuous data. However, many important real-world datasets such as medical images involve discontinuous distributions. GAN training for discontinuous distributions is relatively more challenging, as the training procedure often suffers from instability and mode collapse issues. This dissertation focuses on designing novel GAN architectures to generate representative synthetic image data, and proposes new structures to alleviate GANs' mode collapse issue. As part of this thesis, novel applications of image data generation with GANs have been also investigated for important problems arising in the telecommunication industry and medical domain. Specifically, we first explore various GAN structures to generate engineered electromagnetic surfaces. We consider the continuous approximation of the data and explore the capabilities of feed-forward and convolutional GANs for synthetic data generation. Next, we introduce a novel GAN architecture to address the problem of mode collapse in GAN training. The proposed structure incorporates a third network that penalizes the generator for generating low diversity samples. Lastly, we study the challenging problem of object generation in 3D space using GANs, and we propose extensions to existing 3D GAN structures to generate connected 3D volumes. Additionally, we explore a more challenging version of this 3D volume generation problem by generating connected volumes packed with spheres. This research has applications in radiosurgery treatment planning, and the proposed 3D GAN structure can help generate rare, unseen 3D tumor volumes and information on how to treat these tumors. Accordingly, our analysis contributes to overcoming data scarcity issues (e.g., due to privacy considerations) for an important practical problem in the medical domain.</p>

Data-driven fault diagnosis is essential for the reliability and safety of industry equipment. However, the lack of real labeled fault data make the machine learning-based diagnosis methods difficult to carry out. To solve this problem, this article proposes a new fault diagnosis framework called multilabel one-dimensional (1-D) generation adversarial network (ML1-D-GAN). In our method, Auxiliary Classifier GAN is utilized first for real damage data generation. Then the generated and real damage data are both used to train the fault classifier. Experimental results reveal that the generated data is applicable, and ML1-D-GAN improves the diagnosing accuracy for real bearing faults from 95% to 98% when trained with the generated data. The scalability of the learning model is also proven in the experiment.

Investigating object compositionality in Generative Adversarial Networks

Steganography is an art to hide information in the carriers to prevent from being detected, while steganalysis is the opposite art to detect the presence of the hidden information. With the development of deep learning, several state-of-the-art steganography and steganalysis based on deep learning techniques have been proposed to improve hiding or detection capabilities. Generative Adversarial Networks (GANs) based steganography directly uses the minimax game between the generator and discriminator, to automatically generate steganography algorithms resisting being detected by powerful steganalysis. The steganography without embedding (SWE) based on GANs, where the generated cover images themselves are stego ones carrying secret information has shown its state-of-the-art steganography performance. However, SWE based on GANs has serious weaknesses, such as low information recovery accuracy, low steganography capacity and poor natural showing. To solve these problems, this paper proposes a new SWE based on adversarial training, with carefully designed generator, discriminator and extractor, as well as their loss functions and optimized training mode. The proposed method can achieve a very high information recovery accuracy (100% in some cases), and at the same time improve the steganography capacity and image quality.

In this work we utilize generative adversarial networks (GANs) to synthesize realistic transformations for remote sensing imagery in the multispectral domain. Despite the apparent perceptual realism of the transformed images at a first glance, we show that a deep learning classifier can very easily be trained to differentiate between real and GAN-generated images, likely due to subtle but pervasive artifacts introduced by the GAN during the synthesis process. We also show that a very low-amplitude adversarial attack can easily fool the aforementioned deep learning classifier, although these types of attacks can be partially mitigated via adversarial training. Finally, we explore the features utilized by the classifier to differentiate real images from GAN-generated ones, and how adversarial training causes the classifier to focus on different, lower-frequency features.

Recently, Generative Adversarial Networks (GANs) have received enormous progress, which makes them able to learn complex data distributions in particular faces. More and more efficient GAN architectures have been designed and proposed to learn the different variations of faces, such as cross pose, age, expression and style. These GAN based approaches need to be reviewed, discussed, and categorized in terms of architectures, applications, and metrics. Several reviews that focus on the use and advances of GAN in general have been proposed. However, the GAN models applied to the face, that we call facial GANs, have never been addressed. In this article, we review facial GANs and their different applications. We mainly focus on architectures, problems and performance evaluation with respect to each application and used datasets. More precisely, we reviewed the progress of architectures and we discussed the contributions and limits of each. Then, we exposed the encountered problems of facial GANs and proposed solutions to handle them. Additionally, as GANs evaluation has become a notable current defiance, we investigate the state of the art quantitative and qualitative evaluation metrics and their applications. We concluded the article with a discussion on the face generation challenges and proposed open research issues.