14 June 2022 in IEEE Transactions on Industrial Informatics
Imbalanced data causes deep neural networks to output biased results, and it becomes more serious when facing extremely imbalanced data regarding the outliers with tiny size (the ratio of the outlier size to the image size is around 0.05%). Many data argumentation models are proposed to supplement imbalanced data to alleviate biased results. However, the existing augmentation models cannot synthesize tiny outliers, which makes the generated data unavailable. In this paper, we propose a new augmentation model named Extremely Imbalanced Data Augmentation Generative Adversarial Nets ($EID$-GAN) to address the extremely imbalanced data augmentation problem. First, we design a new penalty function by subtracting the outliers from the cropped region of generated instance to guide the generator to learn the features of outliers. After that, we combine the output value of the penalty function with the generator loss to jointly update the generator’s parameters with back-propagation. Second, we propose a new evaluation approach that adopts two outlier detectors with k-fold cross-validation to assess the availability of generated instances. We conduct extensive experiments to demonstrate the significant performance improvement of $EID$-GAN on two extremely imbalanced datasets: industrial Piston and Fabric datasets; and one general imbalanced dataset: the public DAGM dataset. The experimental results show that our $EID$-GAN outperforms the SOTA augmentation models on different imbalanced datasets.
26 April 2022 in IEEE Transactions on Neural Networks and Learning Systems
The generative adversarial network (GAN) is usually built from the centralized, independent identically distributed (i.i.d.) training data to generate realistic-like instances. In real-world applications, however, the data may be distributed over multiple clients and hard to be gathered due to bandwidth, departmental coordination, or storage concerns. Although existing works, such as federated learning GAN (FL-GAN), adopt different distributed strategies to train GAN models, there are still limitations when data are distributed in a non-i.i.d. manner. These studies suffer from convergence difficulty, producing generated data with low quality. Fortunately, we found that these challenges are often due to the use of a federated averaging strategy to aggregate local GAN models' updates. In this article, we propose an alternative approach to tackling this problem, which learns a globally shared GAN model by aggregating locally trained generators' updates with maximum mean discrepancy (MMD). In this way, we term our approach improved FL-GAN (IFL-GAN). The MMD score helps each local GAN hold different weights, making the global GAN in IFL-GAN getting converged more rapidly than federated averaging. Extensive experiments on MNIST, CIFAR10, and SVHN datasets demonstrate the significant improvement of our IFL-GAN in both achieving the highest inception score and producing high-quality instances.
1 October 2024 in IEEE Transactions on Automation Science and Engineering
19 January 2022 in IEEE Internet of Things Journal
Liquid fraud has plagued people with huge health risks. Liquid fraud detection can help to reduce the risk of liquid hazards. However, existing systems that use biochemical tools or radio frequency signals for liquid sensing are either expensive, intrusive, or inconvenient for public use. In this article, we propose HearLiquid, a low-cost and nonintrusive liquid fraud detection system using commodity acoustic devices. Our insight comes from the fact that acoustic impedance of different liquids results in distinct absorption of the acoustic signal across different frequencies when it travels through the liquid. In specific, we extract the liquid’s acoustic absorption and transmission curve (AATC) over multiple frequencies of the acoustic signal for liquid fraud detection. However, accurately measuring the AATC faces multiple challenges. First, due to the hardware diversity and imperfection, different acoustic devices introduce diverse frequency responses, which brings significant deviations to AATCs of the same liquid. Second, different relative positions between acoustic devices and the liquid container result in variations in the AATC, making the detection result inaccurate. To overcome these challenges, we first calibrate the AATC using a dedicated reference AATC to remove the effect of hardware diversity. To bear the variations in AATCs measured from different relative positions, we apply a well-orchestrated data augmentation technique to automatically generate sufficient AATCs for different positions using a small number of collected data. Finally, AATCs are used to train the liquid detection model. We conduct extensive experiments on many important liquid fraud cases and achieve liquid detection accuracy of 92%–97%.
3 August 2023 in IEEE Transactions on Cybernetics
Mode collapse has been a persisting challenge in generative adversarial networks (GANs), and it directly affects the applications of GAN in many domains. Existing works that attempt to solve this problem have some serious limitations: models using optimal transport (OT) strategies (e.g., Wasserstein distance) lead to vanishing or exploding gradients; increasing the number of generators can cause several generators focusing on the same mode; and approaches that modify the loss also do not satisfactorily resolve mode collapse. In this article, we reduce mode collapse by formulating it as a Monge problem of OT map. We show that the Monge problem can be transformed to the distribution transformation problem in GAN, and a rectified affine neural network can be considered as a measurable function. In this way, we propose Monge GAN that uses this measurable function to transform the generated data distribution into the original data distribution. We utilize the Kantorovich formulation to obtain the OT cost, which is regarded as the OT distance between the two distributions. Finally, we conduct extensive experiments on both image and numerical datasets to validate our Monge GAN in reducing model collapse.
25 May 2023 in IEEE Transactions on Mobile Computing
We study the problem of distributed collision avoidance for mobile robotic systems, where a group of heterogeneous robots with different sizes and motion constraints avoid collisions with each other and static obstacles during the movements from their starting to goal locations. Existing methods mainly consider homogeneous robots and incur a high collision rate in environments with moving robots and static objects. Hence, we propose a distributed c ollision a voidance for heter ogeneous mobile robots ( Heter-CA ), which allows each robot to independently avoid collisions considering the heterogeneity of robots and varying static obstacles. In Heter-CA , each robot predicts the trajectories of neighboring robots and estimates the varying size of static obstacles with the robots' range-finder sensors before motion planning, which enables each robot to avoid obstacles safely. Besides, we prove that Heter-CA can guarantee collision-free movement between heterogeneous robots by satisfying sufficient conditions. We evaluate Heter-CA in numerous simulated and real-world scenarios in which groups of heterogeneous robots perform navigation tasks. The experimental results demonstrate that Heter-CA takes $10\times$ less computation time and achieves $5\%$ less collision rate than baseline algorithms.