Out-of-distribution (OOD) detection is vital for the safe application of intelligent systems in real-world scenarios. This paper proposes an enhancement to OOD detection by leveraging the consistency in cognition betw...
详细信息
ISBN:
(数字)9798350390155
ISBN:
(纸本)9798350390162
Out-of-distribution (OOD) detection is vital for the safe application of intelligent systems in real-world scenarios. This paper proposes an enhancement to OOD detection by leveraging the consistency in cognition between two models, both pretrained on in-distribution (ID) data. Specifically, for a given test sample, we first apply Principal Component Analysis (PCA)-based projection on the feature vectors from each model. These obtained feature vectors (with correlation between dimensions decoupled by PCA projection) are then aligned using a multiple linear mapping, which is fitted using the least squares method on the training data. We hypothesize that the regression error for OOD data will be larger than that for ID data, making it a useful metric for OOD detection. Our experimental results demonstrate the effectiveness of this method. When combined with existing robust baselines, our approach achieves state-of-the-art performance in OOD detection.
Multimodal Sentiment Analysis (MSA) is an attractive research that aims to integrate sentiment expressed in textual, visual, and acoustic signals. There are two main problems in the existing methods: 1) the dominant r...
详细信息
Re-identification (ReID) of individuals across different cameras is a challenging task due to the high-quality large-scale datasets required for the model. Intra-camera supervised (ICS) person re-identification has be...
详细信息
ISBN:
(数字)9798350394948
ISBN:
(纸本)9798350394955
Re-identification (ReID) of individuals across different cameras is a challenging task due to the high-quality large-scale datasets required for the model. Intra-camera supervised (ICS) person re-identification has been proposed to address the high cost of annotating large-scale datasets, but reducing the gap between camera domains remains a major challenge. Current approaches use intra- and inter-camera learning with contrastive learning performed separately in both phases. However, the effect of features from the same person under different cameras on the model during inter-camera learning, and the fine-grained characteristic where the same person can be classified into multiple classes based on the camera labels, still require further research. To address this issue, we propose a Camera-Based Contrastive Learning (CBCL) method that moves features away from their respective cameras and closer to other cameras to reduce domain gaps. We also introduce an Intra-Person Camera Adversarial (IPCA) loss that effectively utilizes fine-grained characteristics of person re-identification and improve IPCA by introducing camera labels to obtain IPCA_2 which achieves better model recognition performance than IPCA alone. Extensive experiments on multiple datasets demonstrate that our method outperforms existing methods and is comparable to fully-supervised methods.
Users usually focus on the application-level requirements which are quite friendly and direct to them. However, there are no existing tools automating the application-level requirements to infrastructure provisioning ...
Users usually focus on the application-level requirements which are quite friendly and direct to them. However, there are no existing tools automating the application-level requirements to infrastructure provisioning and application deployment. Although some security issues have been solved during the development phase, the undiscovered vulnerabilities remain hidden threats to the application's security. Cyberspace mimic defense(CMD) technologies can help to enhance the application's security despite the existence of the vulnerability. In this paper,the concept of SECurity-as-a-Service(SECaaS) is proposed with CMD technologies in cloud environments. The experiment on it was implemented. It is found that the application's security is greatly improved to meet the user's security and performance requirements within budgets through SECaaS. The experimental results show that SECaaS can help the users to focus on application-level requirements(monetary costs,required security level, etc.) and automate the process of application orchestration.
Unstable zeros are widely known to dramatically curb the performance of controllers and the direct applications of several control algorithms. Time delay is a crucial research topic due to the prevalence of communicat...
详细信息
Unstable zeros are widely known to dramatically curb the performance of controllers and the direct applications of several control algorithms. Time delay is a crucial research topic due to the prevalence of communication and computational delays in the control process. In particular, time delay exacerbates the possibility of continuous-time systems with stable zeros transforming into discrete-time systems with unstable zeros. For continuous-time systems with a relative degree greater than or equal to two, at least one zero of the corresponding discrete-time system converges to either marginally stable or unstable locations as the sampling time approaches zero. This study investigates the sampling zeros of discrete-time systems in the case of a generalized sample hold function (GSHF) with time delay. Moreover, it presents a novel modeling method and stability conditions of the limiting zeros for the discrete-time system. The results demonstrate that the stable properties of zeros for the time delay systems can be preserved in the discretization process when GSHF is applied for signal reconstruction when zero-order hold fails to do so. Finally, numerical examples are provided to illustrate the effectiveness of the proposed method in this study.
Visible light positioning (VLP) is a promising positioning technique, which, however, typically requires multiple luminaires to achieve accurate positioning. This paper proposes a novel visual odometry (VO) assisted v...
Visible light positioning (VLP) is a promising positioning technique, which, however, typically requires multiple luminaires to achieve accurate positioning. This paper proposes a novel visual odometry (VO) assisted visible light positioning algorithm (VO-VLP) in achieving positioning with only a single luminaire. In the considered model, a user equipped with a camera jointly uses geometric features in the captured images and coordinates information obtained via visible light communication (VLC) for positioning. The proposed VLP algorithm does not rely on any extra inertial measurement unit and relaxes the tilted angle limitation at the user. In particular, VO-VLP first uses the circle feature of a luminaire to obtain dual normal vectors of the luminaire. Then, the basic principle of VO is used to eliminate the wrong normal vector by exploiting the geometric features in two consecutive images captured when the user moves. Finally, the pose and location of the user are obtained by using an artificially marked point on the luminaire's contour. VO-VLP can achieve accurate positioning with only a single luminaire and a camera. Simulation results show that the proposed indoor positioning algorithm can achieve a 97th-percentile positioning accuracy of around 10 cm.
The smart grid contains lots of intelligent electronic devices (IEDs). IEDs are considered high-priority targets of cyber-attacks. Assessing the security of IEDs effectively can help operators take timely measures to ...
详细信息
The Industrial Internet of Things (IIoT) leverages Federated Learning (FL) for distributed model training while preserving data privacy, and meta-computing enhances FL by optimizing and integrating distributed computi...
详细信息
The mystery of aesthetics attracts scientists from various research fields. The topic of aesthetics, in combination with other disciplines such as neuroscience and computer science, has brought out the burgeoning fiel...
详细信息
In this paper, the problem of maximizing the sum rate of mobile users in a multi-base station (BS) cooperative millimeter-wave (mmWave) multicast communication system is studied. In the considered model, due to the re...
详细信息
ISBN:
(数字)9798350304053
ISBN:
(纸本)9798350304060
In this paper, the problem of maximizing the sum rate of mobile users in a multi-base station (BS) cooperative millimeter-wave (mmWave) multicast communication system is studied. In the considered model, due to the real-time mobility of users, the users being served by a given BS and beamforming of BSs and users are dynamic. Multiple BSs must cooperate to serve dynamic requests of multiple mobile users. This problem is posed as an optimization framework whose goal is to maximize the sum rate of all mobile users by jointly optimizing the number of users served by all BSs and beamforming matrices of both BSs and users. To solve this non-convex optimization problem, we first introduce a value decomposition based reinforcement learning (VD- RL) algorithm to determine the users to be served by each BS. Then, we use the block diagonalization method to obtain the fully digital transmit beamforming matrices of all BSs as well as the receive beamforming matrices of the users. Finally, a fast optimization algorithm is used to optimize the hybrid beamforming matrices of both BSs and users. Simulation results show that, the proposed algorithm can achieve up to 51 % gain in terms of the sum rate of all mobile users compared to baseline multi-agent algorithms.
暂无评论