检索结果-内蒙古大学图书馆

Exploiting the Semantic Knowledge of Pre-trained Text-Encoders for Continual Learning

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Yu, Lu Tao, Zhe Yao, Hantao Van de Weijer, Joost Xu, Changsheng School of Computer Science and Engineering Tianjin University of Technology 300384 China State Key Laboratory of Multimodal Artificial Intelligence Systems Institute of Automation Chinese Academy of Sciences Beijing100190 China Computer Vision Center Universitat Au-tonoma de Barcelona Barcelona08193 Spain School of Artificial Intelligence University of the Chinese Academy of Sciences Beijing100049 China

Deep neural networks (DNNs) excel on fixed datasets but struggle with incremental and shifting data in real-world scenarios. Continual learning addresses this challenge by allowing models to learn from new data while retaining previously learned knowledge. Existing methods mainly rely on visual features, often neglecting the rich semantic information encoded in text. The semantic knowledge available in the label information of the images, offers important semantic information that can be related with previously acquired knowledge of semantic classes. Consequently, effectively leveraging this information throughout continual learning is expected to be beneficial. To address this, we propose integrating semantic guidance within and across tasks by capturing semantic similarity using text embeddings. We start from a pre-trained CLIP model, employ the Semantically-guided Representation Learning (SG-RL) module for a soft-assignment towards all current task classes, and use the Semantically-guided Knowledge Distillation (SG-KD) module for enhanced knowledge transfer. Experimental results demonstrate the superiority of our method on general and fine-grained datasets. Our code can be found in https://***/aprilsveryown/semantically-guided-continual-learning. © 2024, CC BY-NC-ND.

关键词： Contrastive Learning

Simple Image-level Classification Improves Open-vocabulary Object Detection

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Fang, Ruohuan Pang, Guansong Bai, Xiao School of Computer Science and Engineering Beihang University China School of Computing and Information Systems Singapore Management University Singapore State Key Laboratory of Software Development Environment Jiangxi Research Institute Beihang University China

Open-Vocabulary Object Detection (OVOD) aims to detect novel objects beyond a given set of base categories on which the detection model is trained. Recent OVOD methods focus on adapting the image-level pre-trained vision-language models (VLMs), such as CLIP, to a region-level object detection task via, e.g., region-level knowledge distillation, regional prompt learning, or region-text pre-training, to expand the detection vocabulary. These methods have demonstrated remarkable performance in recognizing regional visual concepts, but they are weak in exploiting the VLMs' powerful global scene understanding ability learned from the billion-scale image-level text descriptions. This limits their capability in detecting hard objects of small, blurred, or occluded appearance from novel/base categories, whose detection heavily relies on contextual information. To address this, we propose a novel approach, namely Simple Image-level Classification for Context-Aware Detection Scoring (SIC-CADS), to leverage the superior global knowledge yielded from CLIP for complementing the current OVOD models from a global perspective. The core of SIC-CADS is a multi-modal multi-label recognition (MLR) module that learns the object co-occurrence-based contextual information from CLIP to recognize all possible object categories in the scene. These image-level MLR scores can then be utilized to refine the instance-level detection scores of the current OVOD models in detecting those hard objects. This is verified by extensive empirical results on two popular benchmarks, OV-LVIS and OV-COCO, which show that SIC-CADS achieves significant and consistent improvement when combined with different types of OVOD models. Further, SIC-CADS also improves the cross-dataset generalization ability on Objects365 and OpenImages. Code is available at https://***/mala-lab/SIC-CADS. Copyright © 2023, The Authors. All rights reserved.

关键词： Object detection

Optimal Inter-area Oscillation Damping Control: A Transfer Deep Reinforcement Learning Approach with Switching Control Strategy

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Liang, Siyuan Huo, Long Chen, Xin Sun, Peiyuan The Department of Computer Science and Engineering The Chinese University of Hong Kong Shatin999077 Hong Kong The School of Electrical Engineering The Center of Nanomaterials for Renewable Energy State Key Laboratory of Electrical Insulation and Power Equipment Xi’an Jiaotong University Shaanxi Xi'An710049 China The School of Electrical Engineering Xi’an Jiaotong University Shaanxi Xi'An710049 China

—Wide-area damping control for inter-area oscillation (IAO) is critical to modern power systems. The recent breakthroughs in deep learning and the broad deployment of phasor measurement units (PMU) promote the development of data-driven IAO damping controllers. In this paper, the damping control of IAOs is modeled as a Markov Decision Process (MDP) and solved by the proposed Deep Deterministic Policy Gradient (DDPG) based deep reinforcement learning (DRL) approach. The proposed approach optimizes the eigenvalue distribution of the system, which determines the IAO modes in nature. The eigenvalues are evaluated by the data-driven method called dynamic mode decomposition. For a given power system, only a subset of generators selected by participation factors needs to be controlled, alleviating the control and computing burdens. A Switching Control Strategy (SCS) is introduced to improve the transient response of IAOs. Numerical simulations of the IEEE-39 New England power grid model validate the effectiveness and advanced performance of the proposed approach as well as its robustness against communication delays. In addition, we demonstrate the transfer ability of the DRL model trained on the linearized power grid model to provide effective IAO damping control in the non-linear power grid model environment. Copyright © 2023, The Authors. All rights reserved.

关键词： Dynamic mode decomposition

Two-Timescale Design for Active STAR-RIS Aided Massive MIMO systems

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Papazafeiropoulos, Anastasios Ge, Hanxiao Kourtessis, Pandelis Ratnarajah, Tharmalingam Chatzinotas, Symeon Papavassiliou, Symeon The Communications and Intelligent Systems Research Group University of Hertfordshire HatfieldAL10 9AB United Kingdom SnT The University of Luxembourg Luxembourg The Institute for Digital Communications The University of Edinburgh EdinburghEH9 3FG United Kingdom The Institute of Communication and Computer Systems School of Electrical and Computer Engineering National Technical University of Athens Zografou15780 Greece

Simultaneously transmitting and reflecting reconfigurable intelligent surface (STAR-RIS) is a promising implementation of RIS-assisted systems that enables full-space coverage. However, STAR-RIS as well as conventional RIS suffer from the double-fading effect. Thus, in this paper, we propose the marriage of active RIS and STAR-RIS, denoted as ASTARS for massive multiple-input multiple-output (mMIMO) systems, and we focus on the energy splitting (ES) and mode switching (MS) protocols. Compared to prior literature, we consider the impact of correlated fading, and we rely our analysis on the two timescale protocol, being dependent on statistical channel state information (CSI). On this ground, we propose a channel estimation method for ASTARS with reduced overhead that accounts for its architecture. Next, we derive a closed-form expression for the achievable sum-rate for both types of users in the transmission and reflection regions in a unified approach with significant practical advantages such as reduced complexity and overhead, which result in a lower number of required iterations for convergence compared to an alternating optimization (AO) approach. Notably, we maximize simultaneously the amplitudes, the phase shifts, and the active amplifying coefficients of the ASTARS by applying the projected gradient ascent method (PGAM). Remarkably, the proposed optimization can be executed at every several coherence intervals that reduces the processing burden considerably. Simulations corroborate the analytical results, provide insight into the effects of fundamental variables on the sum achievable SE, and present the superiority of ASTARS compared to passive STAR-RIS for a practical number of surface elements. © 2024, CC BY.

关键词： MIMO systems

A photonic crystal fiber for supporting 30 orbital angular momentum modes with low dispersion

学校读者我要写书评

暂无评论

Optoelectronics Letters 2020年第1期16卷 34-39页

作者： HUANG Wei YOU Yong SONG Bin-bin CHEN Sheng-yong Key Laboratory of Computer Vision and Systems(Ministry of Education) School of Computer Science and EngineeringTianjin University of TechnologyTianjin 300384China Institute of Modern Optics Nankai UniversityTianjin 300350China

This paper proposes a novel photonic crystal optical fiber which can support 30 orbital angular momentum(OAM)modes transmission and possesses relatively flat and low *** OAM modes can be well-separated due to the large effective refractive index difference(above 10^-4)between the *** only material of the designed fiber is *** dispersion of each OAM mode is controlled in the range of 50-100 ps·nm^-1·km^-1 and the total dispersion variation is below 10 ps·nm^-1·km^-1 from 1500 nm to 1600 ***,the confinement loss of each OAM mode is below 8.17×10^-10 dB/m at 1550 nm,and the nonlinear coefficients is less than 0.71 W^-1/km for all modes at 1550 *** all these good features,this proposed optical fiber is promising to be applied in fiberbased OAM communication systems.

关键词： fiber OAM dispersion

Faster Convergence of Local SGD for Over-Parameterized Models

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Qin, Tiancheng Rasoul Etesami, S. Uribe, Cesár A. Department of Industrial and Systems Engineering Coordinated Science Laboratory University of Illinois at Urbana-Champaign United States Department of Electrical and Computer Engineering Rice University United States

Modern machine learning architectures are often highly expressive. They are usually overparameterized and can interpolate the data by driving the empirical loss close to zero. We analyze the convergence of Local SGD (or FedAvg) for such over-parameterized models in the heterogeneous data setting and improve upon the existing literature by establishing the following convergence rates. For general convex loss functions, we establish an error bound of O(1/T) under a mild data similarity assumption and an error bound of O(K/T) otherwise, where K is the number of local steps and T is the total number of iterations. For non-convex loss functions we prove an _error bound of O(K/T). These bounds improve upon the best previous bound of O(1/√nT) in both cases, where n is the number of nodes, when no assumption on the model being over-parameterized is made. We complete our results by providing problem instances in which our established convergence rates are tight to a constant factor with a reasonably small stepsize. Finally, we validate our theoretical results by performing large-scale numerical experiments that reveal the convergence behavior of Local SGD for practical over-parameterized deep learning models, in which the O(1/T) convergence rate of Local SGD is clearly shown. Copyright © 2022, The Authors. All rights reserved.

关键词： Parameterization

Use of smart thermostat data to investigate harmful exposure to heat in Calgary, Canada

学校读者我要写书评

暂无评论

Use of smart thermostat data to investigate harmful exposure...

International Workshop on Secure Internet of Things (SIoT)

作者： Júlia Perassolli de Lázari Arlene Oetomo Paula Dornhofer Paro Costa Plinio Pelegrini Morita School of Electrical and Computer Engineering (FEEC) State University of Campinas (UNICAMP) Campinas Brazil School of Public Health and Health Systems University of Waterloo Waterloo Canada

With the increase in temperatures and extreme heat events, developing a better strategy to assess and quantify heat exposure indoors is crucial. Although people spend most of their time indoors, most studies and alert systems focus on outdoor temperatures. The growing adoption of smart thermostats in home heating, ventilation, and air conditioning (HVAC) systems offers an opportunity to study the relationship between indoor and outdoor temperatures. The present study uses indoor smart thermostats and weather station data to investigate heat exposure metrics for indoor and outdoor temperatures. We also analyzed the percentage of time indoor temperatures are within the ASHARE acceptability limit. We found that houses spend around 5% of the time above the ASHRAE limit and that indoor heat exposure metrics are higher than outdoor metrics, meaning greater harmful exposure to human health during extreme events like heatwaves.

关键词： Heating systems Measurement Air conditioning Temperature distribution HVAC Ventilation Social Internet of Things

Enhancing Privacy in Federated Learning through Local Training

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Bastianello, Nicola Liu, Changxin Johansson, Karl H. School of Electrical Engineering and Computer Science and Digital Futures KTH Royal Institute of Technology Sweden Key Laboratory of Smart Manufacturing in Energy Chemical Process Ministry of Education East China University of Science and Technology Shanghai200237 China

In this paper we propose the federated learning algorithm Fed-PLT to overcome the challenges of (i) expensive communications and (ii) privacy preservation. We address (i) by allowing for both partial participation and local training, which significantly reduce the number of communication rounds between the central coordinator and computing agents. The algorithm matches the state of the art in the sense that the use of local training demonstrably does not impact accuracy. Additionally, agents have the flexibility to choose from various local training solvers, such as (stochastic) gradient descent and accelerated gradient descent. Further, we investigate how employing local training can enhance privacy, addressing point (ii). In particular, we derive differential privacy bounds and highlight their dependence on the number of local training epochs. We assess the effectiveness of the proposed algorithm by comparing it to alternative techniques, considering both theoretical analysis and numerical results from a classification task. Copyright © 2024, The Authors. All rights reserved.

关键词： Federated learning

Evaluation of the Improvement in Hierarchical Lossless Video Compression

学校读者我要写书评

暂无评论

Evaluation of the Improvement in Hierarchical Lossless Video...

International Conference on Control, Decision and Information Technologies (CoDIT)

作者： Donia Ammous Amina Kessentini Naziha Khlif Fahmi Kammoun Nouri Masmoudi Electrical engineering Department Laboratory of Electronics and Information Technologies National Engineering School of Sfax Sfax University Sfax Tunisia Electronics and telecommunications department Higher Institute of Computer and Multimedia of Gabes Gabes university Gabes Tunisia

The digital medical video is widely used in several healthcare institutions and medical centers. Therefore, a considerable storage space and a high transmission bandwidth are necessary. Thus, efficient compression tools are required to reduce the storage space and fit bandwidth capacities. Indeed, several lossy and lossless compression techniques are applied. Lossy techniques ensure a good performance. But they lead to erroneous medical data, which can negatively affect the doctor's decisions about the patient's health conditions. As a result, lossless compression guarantees an optimal solution for medical applications. Among available lossless techniques, H.264/AVC (Advanced Video Coding) presents high coding performance in lossless compression. H.264/AVC was chosen since it simplicity in terms of computational complexity. It is suitable for mobile devices that require high resolution. In this paper, several techniques used with the H.264/AVC are discussed. Moreover, a new approach is applied to enhance the lossless techniques. Its results show an important compression ratio compared to the previous works. The enhanced method results in an increase in compression ratio from 8,429 to 16,159. Besides, the gain performance reaches from 32,625 to 42.074 %.

关键词：