Modern wafer inspection systems in Integrated Circuit (IC) manufacturing utilize deep neural networks. The training of such networks requires the availability of a very large number of defective or faulty die patterns...
详细信息
ISBN:
(纸本)9781510673878;9781510673861
Modern wafer inspection systems in Integrated Circuit (IC) manufacturing utilize deep neural networks. The training of such networks requires the availability of a very large number of defective or faulty die patterns on a wafer called wafer maps. The number of defective wafer maps on a production line is often limited. In order to have a very large number of defective wafer maps for the training of deep neural networks, generative models can be utilized to generate realistic synthesized defective wafer maps. This paper compares the following three generative models that are commonly used for generating synthesized images: Generative Adversarial Network (GAN), variational Auto-Encoder (VAE), and CycleGAN which is a variant of GAN. The comparison is carried out based on the public domain wafer map dataset WM-811K. The quality aspect of the generated wafer map images is evaluated by computing the five metrics of peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM), inception score (IS), Frechet inception distance (FID), and kernel inception distance (KID). Furthermore, the computational efficiency of these generative networks is examined in terms of their deployment in a real-time inspection system.
This study proposes a novel parameter and topology optimization method for an interior permanent magnet synchronous motor using a multimodal neural network. The optimized shape obtained by the proposed method improves...
详细信息
ISBN:
(纸本)9798350348958;9798350348965
This study proposes a novel parameter and topology optimization method for an interior permanent magnet synchronous motor using a multimodal neural network. The optimized shape obtained by the proposed method improves the manufacturability compared to the conventional method. In addition, the computational time of the optimization using the proposed method is reduced by about 99.9% compared to the optimization using only the finite element method.
Audio and Visual are two important visual modalities in video content understanding. However, the absence of one modality may be observed in practical applications due to the real environmental factors, which leads to...
详细信息
ISBN:
(纸本)9798350330991;9798350331004
Audio and Visual are two important visual modalities in video content understanding. However, the absence of one modality may be observed in practical applications due to the real environmental factors, which leads to the information loss. Therefore, audio and visual fusion is focused on using the shared and complementary information between modalities to recover the missing modalities from the available data modalities. In this paper, an Adversarial Hierarchical variational autoencoder (Adv-HVAE) model is proposed to solve this problem of modality data loss. A multimodal representation is first learned using a hierarchical variational autoencoder (VAE) model that enables the generation of missing modal data under any subset of available modalities. Also to obtain a more robust multimodal representation, a feature generation network is utilized to approximate the latent distribution of missing modalities. Finally, the adversarial training network is shown to be effective in improving the data quality generated through the Adv-HVAE framework. Experimental results demonstrate that Adv-HVAE achieves best generation results on two benchmark datasets, avMNIST and Sub-URMP.
For the simulation-based test and evaluation of connected and automated vehicles (CAVs), the trajectory of the background vehicle has a direct effect on the performance of CAVs and experiment outcomes. The collected r...
详细信息
For the simulation-based test and evaluation of connected and automated vehicles (CAVs), the trajectory of the background vehicle has a direct effect on the performance of CAVs and experiment outcomes. The collected real trajectory data are limited by the sample size and diversity, and may exclude critical attribute combinations that are of vital importance for CAVs' tests. Consequently, it is indispensable to increase the richness of accessible trajectory data. In this study, we developed the Wasserstein generative adversarial network with gradient penalty (WGAN-GP) and a hybrid model of variational autoencoder and generative adversarial network (VAE-GAN) for trajectory data generation. These models are capable of learning a compressed representation of the observed data space, and generating data by sampling in the latent space and then mapping back to the original space. The real data and the generated data are applied in the car-following model of CAVs with cooperative adaptive cruise control (CACC) to evaluate safety performance using the time-to-collision (TTC) index. The results indicate that the generated data of the two generative models have reasonable differences while maintaining a certain similarity with the real samples. When real and generated trajectory data are applied to the car-following model of CAVs, the generated trajectory data increases the number of new critical fragments whose TTC is smaller than the threshold. The WGAN-GP model performs better than the VAE-GAN model according to the ratio of critical fragments. Findings of this study provide useful insights for CAVs' tests and safety performance improvement.
The inference of gene regulatory networks (GRNs) from single-cell RNA sequencing (scRNA-seq) data enables describing the regulatory relationships among genes from a cellular perspective, and revealing the essence of v...
详细信息
ISBN:
(纸本)9789819751273;9789819751280
The inference of gene regulatory networks (GRNs) from single-cell RNA sequencing (scRNA-seq) data enables describing the regulatory relationships among genes from a cellular perspective, and revealing the essence of various life phenomena. However, the high sparsity of scRNA-seq data poses new challenges to the inference of GRNs. In this study, an introspective adversarial gene regulatory network unsupervised inference model, called IntroGRN, is designed based on variational autoencoder and structural equation model. IntroGRN attempts to produce better reconstruction samples via adversarial training, and introduce the structural equation model into the process of adversarial. Compared with eight state-of-the-art methods, the proposed IntroGRN method can infer the best gene regulatory networks in most benchmark datasets, which has been verified through extensive experiments. The source code of method IntroGRN can be downloaded from https://***/lryup/IntroGRN.
This paper presents a data-driven model that, by exploring the correlation between the data coming from two sensors (GPS and camera), allows us to explain the odometry anomalies by analyzing video data. Our approach u...
详细信息
ISBN:
(纸本)9798350374520;9798350374513
This paper presents a data-driven model that, by exploring the correlation between the data coming from two sensors (GPS and camera), allows us to explain the odometry anomalies by analyzing video data. Our approach uses a Markov Jump Particle Filter (MJPF) to process the odometry data of a vehicle, extracting its features and identifying instant anomalies. Simultaneously, the object causing the anomaly is detected and tracked in the video data. After that, the correlation between the odometric trajectories and the object trajectories is determined in both normal and abnormal cases. Another correlation coefficient is then employed to calculate the distance between the computed correlations. The proposed method is evaluated using multi-modal data collected from a vehicle operating in a closed environment, where pedestrians represent anomalies. We show that our system is able to distinguish which video anomaly better explains the odometry anomaly.
Hybrid neural networks are able to capture the time-dependency of electroencephalography (EEG) signals, and can therefore effectively perform pattern recognition of motor imagery EEG signals. However, a sufficiently l...
详细信息
ISBN:
(纸本)9798350351040;9798350351033
Hybrid neural networks are able to capture the time-dependency of electroencephalography (EEG) signals, and can therefore effectively perform pattern recognition of motor imagery EEG signals. However, a sufficiently large amount of training data is required to achieve optimal results, so it is necessary to use data augmentation methods to increase the amount of data. Therefore, we propose a data augmentation method based on the Attention Convolutional Variation autoencoder (ACVAE) and design a Convolutional :Neural Network-Long Short -Term Memory (CNN-LSTN1) network for pattern recognition. The BCI Competition IV dataset 2a is used for experimental validation. The results show that the ACVAE method produces higher quality data, with the highest recognition accuracy of 97.16% for a single subject in the four-classification task. Compared to other methods, we proposed method shows excellent performance.
Online process monitoring is essential to detect failures and respond promptly in automated industrial processes such as injection molding. Traditional systems rely on experienced operators manually defining operation...
详细信息
ISBN:
(纸本)9798350309300;9798350309294
Online process monitoring is essential to detect failures and respond promptly in automated industrial processes such as injection molding. Traditional systems rely on experienced operators manually defining operational boundaries around a reference signal. We propose a data-driven representation that auto-tunes the sensitivity to a pre-set specificity threshold and automatically detects anomalies alongside interpretable indices that help identify root causes. Our automated system achieved an average AUC of 0.998 and detected 100 percent of the anomalies with the proposed dynamic calibration of the data-driven embedding method. The dynamic calibration, which accounted for drift, boosts the average specificity from 0.362 to 0.869. The outputs also indicate the direction and relative magnitude of characteristic deviations caused by machine parameters, including holding pressure, mold temperature, and injection speed. The AI-derived process boundaries are superior to manual annotation in tested real-world production environments.
Vertical Federated Learning (VFL) is a prevalent paradigm designed to facilitate collaboration between multiple entities possessing distinct feature sets yet sharing a common user base for model training. However, cha...
详细信息
ISBN:
(数字)9789819754984
ISBN:
(纸本)9789819754977;9789819754984
Vertical Federated Learning (VFL) is a prevalent paradigm designed to facilitate collaboration between multiple entities possessing distinct feature sets yet sharing a common user base for model training. However, challenges emerge when generating predictions for users with partial data, such as the scenario where a new user registers with only one participating company and submits a limited number of features. The typical method of addressing missing data involves padding the absent values with zeros or mean figures, which can cause an out-of-distribution issue, thereby precipitating a notable deterioration in model accuracy. To address this issue, we introduce DiVerFed, a distribution-aware VFL framework for missing information. DiVerFed's primary objective is to maintain robust model performance even when confronted with incomplete user information. Within this framework, we treat the VFL's top model as an encoder responsible for capturing the underlying data distribution. To make this distribution effectively, we incorporate a reconstruction component. Furthermore, we simulate the scenario of missing information by integrating a Feature-wise Dropout module during the training phase. To enhance the framework's efficacy in classification tasks, we also incorporate label information within the loss function to leverage a class-aware distribution to bolster the model's accuracy. Our experimental analyses confirm that DiVerFed significantly outperforms conventional approaches in classification tasks when information from only one party is accessible.
The millimeter wave bands are being increasingly considered for wireless communication to unmanned aerial vehicles (UAVs). Critical to this undertaking are statistical channel models that describe the distribution of ...
详细信息
The millimeter wave bands are being increasingly considered for wireless communication to unmanned aerial vehicles (UAVs). Critical to this undertaking are statistical channel models that describe the distribution of constituent parameters in scenarios of interest. This paper presents a general modeling methodology based on data-training a generative neural network. The proposed generative model has a two-stage structure that first predicts the link state (line-of-sight, non-line-of-sight, or outage), and subsequently feeds this state into a conditional variational autoencoder (VAE) that generates the path losses, delays, and angles of arrival and departure for all the propagation paths. The methodology is demonstrated for 28GHz air-to-ground channels between UAVs and a cellular system in representative urban environments, with training datasets produced through ray tracing. The demonstration extends to both standard base stations (installed at street level and downtilted) as well as dedicated base stations (mounted on rooftops and uptilted). The proposed approach is able to capture complex statistical relations in the data and it significantly outperforms standard 3GPP models, even after refitting the parameters of those models to the data.
暂无评论