YZ
Y. Zhu
info
Please Note
<p>This page displays the records of the person named above and is not linked to a unique person identifier. This record may need to be merged to a profile.</p>
8 records found
1
This study presents a novel image-based machine learning (ML) method for automating I–V parameter extraction in gallium nitride (GaN) devices. Using Ampleon’s GEAR model, a dataset of 100000 simulated I–V curves are converted into I–V images through specifically designed transfer functions to train a convolutional neural network. The proposed method outperforms the existing ML method based on a fully connected neural network, particularly for I–V curves in the subthreshold region. Validation with measured pulse I–V data shows its superior accuracy, achieving a normalized mean square error (NMSE) of −30 dB compared with −24 dB with the existing ML method. The proposed method demonstrates a strong potential to accelerate the extraction and enhance the accuracy of GaN device modeling.
...
This study presents a novel image-based machine learning (ML) method for automating I–V parameter extraction in gallium nitride (GaN) devices. Using Ampleon’s GEAR model, a dataset of 100000 simulated I–V curves are converted into I–V images through specifically designed transfer functions to train a convolutional neural network. The proposed method outperforms the existing ML method based on a fully connected neural network, particularly for I–V curves in the subthreshold region. Validation with measured pulse I–V data shows its superior accuracy, achieving a normalized mean square error (NMSE) of −30 dB compared with −24 dB with the existing ML method. The proposed method demonstrates a strong potential to accelerate the extraction and enhance the accuracy of GaN device modeling.
Conference paper
(2024)
-
Yang Lou, Yi Zhu, Qun Song, Rui Tan, Chunming Qiao, Wei Bin Lee, Jianping Wang
Trajectory prediction forecasts nearby agents' moves based on their historical trajectories. Accurate trajectory prediction (or prediction in short) is crucial for autonomous vehicles (AVs). Existing attacks compromise the prediction model of a victim AV by directly manipulating the historical trajectory of an attacker AV, which has limited real-world applicability. This paper, for the first time, explores an indirect attack approach that induces prediction errors via attacks against the perception module of a victim AV. Although it has been shown that physically realizable attacks against LiDAR-based perception are possible by placing a few objects at strategic locations, it is still an open challenge to find an object location from the vast search space in order to launch effective attacks against prediction under varying victim AV velocities. Through analysis, we observe that a prediction model is prone to an attack focusing on a single point in the scene. Consequently, we propose a novel two-stage attack framework to realize the single-point attack. The first stage of prediction-side attack efficiently identifies, guided by the distribution of detection results under object-based attacks against perception, the state perturbations for the prediction model that are effective and velocity-insensitive. In the second stage of location matching, we match the feasible object locations with the found state perturbations. Our evaluation using a public autonomous driving dataset shows that our attack causes a collision rate of up to 63% and various hazardous responses of the victim AV. The effectiveness of our attack is also demonstrated on a real testbed car 1. To the best of our knowledge, this study is the first security analysis spanning from LiDAR-based perception to prediction in autonomous driving, leading to a realistic attack on prediction. To counteract the proposed attack, potential defenses are discussed.
...
Trajectory prediction forecasts nearby agents' moves based on their historical trajectories. Accurate trajectory prediction (or prediction in short) is crucial for autonomous vehicles (AVs). Existing attacks compromise the prediction model of a victim AV by directly manipulating the historical trajectory of an attacker AV, which has limited real-world applicability. This paper, for the first time, explores an indirect attack approach that induces prediction errors via attacks against the perception module of a victim AV. Although it has been shown that physically realizable attacks against LiDAR-based perception are possible by placing a few objects at strategic locations, it is still an open challenge to find an object location from the vast search space in order to launch effective attacks against prediction under varying victim AV velocities. Through analysis, we observe that a prediction model is prone to an attack focusing on a single point in the scene. Consequently, we propose a novel two-stage attack framework to realize the single-point attack. The first stage of prediction-side attack efficiently identifies, guided by the distribution of detection results under object-based attacks against perception, the state perturbations for the prediction model that are effective and velocity-insensitive. In the second stage of location matching, we match the feasible object locations with the found state perturbations. Our evaluation using a public autonomous driving dataset shows that our attack causes a collision rate of up to 63% and various hazardous responses of the victim AV. The effectiveness of our attack is also demonstrated on a real testbed car 1. To the best of our knowledge, this study is the first security analysis spanning from LiDAR-based perception to prediction in autonomous driving, leading to a realistic attack on prediction. To counteract the proposed attack, potential defenses are discussed.
On predicting individual video viewing experience
The value of user information
Experience prediction is one key component in today’s multimedia delivery. Knowing user’s viewing experience allows online video service providers (e.g., Netflix, YouTube) to create value for their customers by providing personalized content and service. However, individual experience prediction is a challenging problem since viewing experience (defined as Quality of Experience in this thesis) is a multifaceted quantity and it is rather personal and subjective. The existing methods for quantifying Quality of Experience (QoE) target at estimating how the video quality is perceived by users, neglecting the hedonic part of experience (the degree of enjoyment of a user watching a video). Quite naturally, these methods consider only factors related to video perceptual quality (purely from video), which is insufficient to properly assess viewing experience. The research reported in this thesis attempts for the first time at shifting the paradigm for perceptual quality modeling, towards measuring and predicting the level of enjoyable viewing experience a user has with a video. In particular, it focuses on exploiting the potential value of user factors (information from users) and investigate their influences on QoE prediction.
The goal of this thesis is to develop a feasible method for predicting the individual viewing experiences in terms of perceptual quality and enjoyment by taking multiple influencing factors into account. Here, the influencing factors are taken from both video (e.g., related to perceptual quality) and user (user factors, e.g., interest. personality). We take three major steps to accomplish this goal. We first deploy a subjective experiment to understand the relationship between perceptual quality and enjoyment, and how their influencing factors form the final viewing experience. With a set of identified influencing factors, we then propose a new QoE prediction model which processes both user and video information to predict individual experience (i.e., either perceptual quality or enjoyment). We show that combining information from video and user enables better prediction performance as compared to only considering information from video related to perceptual quality. Our third step tackles the problem of reliable data collection for the individual QoE research. We developed an open-sourced Facebook application, named YouQ, as an experimental platform for automatic user information collection from social media while performing an online QoE subjective experiment. We show that YouQ can produce reliable results as compared to a controlled laboratory experiment, both in terms of QoE and of quantification of user factors and traits. As a result, a complete, feasible method for individual QoE prediction is presented in this thesis.
Based on the findings presented in this thesis, we reflect on the contribution and make recommendations for future research directions, which we think are substantial and promising for individual QoE prediction. ...
The goal of this thesis is to develop a feasible method for predicting the individual viewing experiences in terms of perceptual quality and enjoyment by taking multiple influencing factors into account. Here, the influencing factors are taken from both video (e.g., related to perceptual quality) and user (user factors, e.g., interest. personality). We take three major steps to accomplish this goal. We first deploy a subjective experiment to understand the relationship between perceptual quality and enjoyment, and how their influencing factors form the final viewing experience. With a set of identified influencing factors, we then propose a new QoE prediction model which processes both user and video information to predict individual experience (i.e., either perceptual quality or enjoyment). We show that combining information from video and user enables better prediction performance as compared to only considering information from video related to perceptual quality. Our third step tackles the problem of reliable data collection for the individual QoE research. We developed an open-sourced Facebook application, named YouQ, as an experimental platform for automatic user information collection from social media while performing an online QoE subjective experiment. We show that YouQ can produce reliable results as compared to a controlled laboratory experiment, both in terms of QoE and of quantification of user factors and traits. As a result, a complete, feasible method for individual QoE prediction is presented in this thesis.
Based on the findings presented in this thesis, we reflect on the contribution and make recommendations for future research directions, which we think are substantial and promising for individual QoE prediction. ...
Experience prediction is one key component in today’s multimedia delivery. Knowing user’s viewing experience allows online video service providers (e.g., Netflix, YouTube) to create value for their customers by providing personalized content and service. However, individual experience prediction is a challenging problem since viewing experience (defined as Quality of Experience in this thesis) is a multifaceted quantity and it is rather personal and subjective. The existing methods for quantifying Quality of Experience (QoE) target at estimating how the video quality is perceived by users, neglecting the hedonic part of experience (the degree of enjoyment of a user watching a video). Quite naturally, these methods consider only factors related to video perceptual quality (purely from video), which is insufficient to properly assess viewing experience. The research reported in this thesis attempts for the first time at shifting the paradigm for perceptual quality modeling, towards measuring and predicting the level of enjoyable viewing experience a user has with a video. In particular, it focuses on exploiting the potential value of user factors (information from users) and investigate their influences on QoE prediction.
The goal of this thesis is to develop a feasible method for predicting the individual viewing experiences in terms of perceptual quality and enjoyment by taking multiple influencing factors into account. Here, the influencing factors are taken from both video (e.g., related to perceptual quality) and user (user factors, e.g., interest. personality). We take three major steps to accomplish this goal. We first deploy a subjective experiment to understand the relationship between perceptual quality and enjoyment, and how their influencing factors form the final viewing experience. With a set of identified influencing factors, we then propose a new QoE prediction model which processes both user and video information to predict individual experience (i.e., either perceptual quality or enjoyment). We show that combining information from video and user enables better prediction performance as compared to only considering information from video related to perceptual quality. Our third step tackles the problem of reliable data collection for the individual QoE research. We developed an open-sourced Facebook application, named YouQ, as an experimental platform for automatic user information collection from social media while performing an online QoE subjective experiment. We show that YouQ can produce reliable results as compared to a controlled laboratory experiment, both in terms of QoE and of quantification of user factors and traits. As a result, a complete, feasible method for individual QoE prediction is presented in this thesis.
Based on the findings presented in this thesis, we reflect on the contribution and make recommendations for future research directions, which we think are substantial and promising for individual QoE prediction.
The goal of this thesis is to develop a feasible method for predicting the individual viewing experiences in terms of perceptual quality and enjoyment by taking multiple influencing factors into account. Here, the influencing factors are taken from both video (e.g., related to perceptual quality) and user (user factors, e.g., interest. personality). We take three major steps to accomplish this goal. We first deploy a subjective experiment to understand the relationship between perceptual quality and enjoyment, and how their influencing factors form the final viewing experience. With a set of identified influencing factors, we then propose a new QoE prediction model which processes both user and video information to predict individual experience (i.e., either perceptual quality or enjoyment). We show that combining information from video and user enables better prediction performance as compared to only considering information from video related to perceptual quality. Our third step tackles the problem of reliable data collection for the individual QoE research. We developed an open-sourced Facebook application, named YouQ, as an experimental platform for automatic user information collection from social media while performing an online QoE subjective experiment. We show that YouQ can produce reliable results as compared to a controlled laboratory experiment, both in terms of QoE and of quantification of user factors and traits. As a result, a complete, feasible method for individual QoE prediction is presented in this thesis.
Based on the findings presented in this thesis, we reflect on the contribution and make recommendations for future research directions, which we think are substantial and promising for individual QoE prediction.
Most automatic Quality of Experience (QoE) assessment models have so far aimed at predicting the QoE of a video as experienced by an average user, and solely based on perceptual characteristics of the video being viewed. The importance of other characteristics, such as those related to the video content being watched, or those related to an individual user have been largely neglected. This is suboptimal in view of the fact that video viewing experience is individual and multifaceted, considering the perceived quality (related to coding or network-induced artifacts), but also other -- more hedonic - aspects, like enjoyment. In this paper, we propose an expanded model which aims to assess QoE of a given video, not only in terms of perceived quality but also of enjoyment, as experienced by a specific user. To do so, we feed the model not only with information extracted from the video (related to both perceived quality and content), but also with individual user characteristics, such as interest, personality and gender. We assess our expanded QoE model based on two publicly available QoE datasets, namely i_QoE and CP-QAE-I. The results show that combining various types of characteristics enables better QoE prediction performance as compared to only considering perceptual characteristics of the video, both when targeting perceived quality and enjoyment.
...
Most automatic Quality of Experience (QoE) assessment models have so far aimed at predicting the QoE of a video as experienced by an average user, and solely based on perceptual characteristics of the video being viewed. The importance of other characteristics, such as those related to the video content being watched, or those related to an individual user have been largely neglected. This is suboptimal in view of the fact that video viewing experience is individual and multifaceted, considering the perceived quality (related to coding or network-induced artifacts), but also other -- more hedonic - aspects, like enjoyment. In this paper, we propose an expanded model which aims to assess QoE of a given video, not only in terms of perceived quality but also of enjoyment, as experienced by a specific user. To do so, we feed the model not only with information extracted from the video (related to both perceived quality and content), but also with individual user characteristics, such as interest, personality and gender. We assess our expanded QoE model based on two publicly available QoE datasets, namely i_QoE and CP-QAE-I. The results show that combining various types of characteristics enables better QoE prediction performance as compared to only considering perceptual characteristics of the video, both when targeting perceived quality and enjoyment.