elementnameelementnameelementname -->
Bet365 app Bet365 app download
Bet365 app Bet365 app download Homepage bet365 Play online games building,Party Text

Academic Bet365 app download

Academic Bet365 app download by doctoral students Zhang Jinrui, Wang Dan and Yue Sheng

Source: Click: Bet365 app download: August 23, 2022 20:34

Bet365 app download title 1:MobiPose: Real-Time Multi-Person Pose Estimation on Bet365 app download DevicesBet365 app download Bet365 app download by ACM SenSys’20 and IEEE TMC 2021

Bet365 app download time:2022Year9month1Sunday morning9:00-9:30

Reporter: Bet365 app download

Reporting location: Shenghua back building409 / Tencent Conference 705-467-296

Abstract:

Human pose estimation is a key technique for many vision-based Bet365 app download applications. Yet existing multi-person pose-estimation methods fail to achieve a satisfactory user experience on commodity Bet365 app download devices such as smartphones, due to their long model-inference latency. In this paper, we propose MobiPose, a system designed to enable real-time multi-person pose estimation on Bet365 app download devices through three novel techniques. First, MobiPose takes a motion-vector-based approach to fast locate the human proposals across consecutive frames by fine-grained tracking of joints of human body, rather than running the expensive human-detection model for every frame. Second, MobiPose designs a Bet365 app download-friendly model that uses lightweight multi-stage feature extractions to significantly reduce the latency of pose estimation without compromising the model accuracy. Third, MobiPose leverages the heterogeneous computing resources of both CPU and GPU to execute the pose-estimation model for multiple persons in parallel, which further reduces the total latency.

We have implemented the MobiPose system on off-the-shelf commercial smartphones and conducted comprehensive experiments to evaluate the effectiveness of the proposed techniques. Evaluation results show that MobiPose achieves over 20 frames per second pose estimation with 3 persons per frame, and significantly outperforms the state-of-the-art baseline, with a maximum speedup of up to 4.5× and 2.8× in latency on CPU and GPU, respectively, and an improvement of 5.1% in pose-estimation model accuracy. Furthermore, MobiPose achieves up to 62.5% and 37.9% energy-per-frame saving on average in comparison with the baseline on Bet365 app download CPU and GPU, respectively.


Bet365 app download title 2:MobiDepth: Bet365 app download Using On-Device Dual CamerasBet365 app download Bet365 app download by ACM MobiCom’22

Bet365 app download time:2022Year9month1Sunday morning9:30-10:00

Reporter: Bet365 app download

Reporting location: Shenghua back building409 / Tencent Conference 705-467-296

Abstract:

Real-time depth estimation is critical for the increasingly popular augmented reality and virtual reality applications on Bet365 app download devices. Yet existing solutions are insufficient as they require expensive depth sensors or motion of the device or have a high latency. We propose MobiDepth, a real-time depth estimation system using the widely available on-device dual cameras. While binocular depth estimation is a mature technique, it is challenging to realize the technique on commodity Bet365 app download devices due to the different focal lengths and unsynchronized frame flows of the on-device dual cameras and the heavy stereo-matching algorithm.

To address the challenges, MobiDepth integrates three novel techniques: 1) iterative field-of-view cropping, which crops the field-of-views of the dual cameras to achieve the equivalent focal lengths for accurate epipolar rectification; 2) heterogeneous camera synchronization, which synchronizes the frame flows captured by the dual cameras to avoid the displacement of moving objects across the frames in the same pair; 3) Bet365 app download GPU-friendly stereo matching, which effectively reduces the latency of stereo matching on a Bet365 app download GPU. We implement MobiDepth on multiple commodity Bet365 app download devices and conduct comprehensive evaluations. Experimental results show that MobiDepth achieves real-time depth estimation of 22 frames per second with a significantly reduced depth-estimation error compared with the baselines. Using MobiDepth, we further build an example application of 3D pose estimation, which significantly outperforms the state-of-the-art 3D pose-estimation method, reducing the pose-estimation latency and error by up to 57.1% and 29.5%, respectively.

Introduction to Dr. Bet365 app download:

Zhang Jinrui, Bet365 app download2018 PhD, supervisor: Academician Zhang Yao, co-supervisor: Associate Professor Zhang Deyu。The research direction is mobile deep Bet365 app download system optimization、Edge Computing、Mobile terminal depth vision system optimization、AR et al. everYu2019-2022At Peking University and Beijing Bet365 app downloadMicrosoft Research Asia, Tsinghua University Intelligent Industry Research Institute (AIR) exchange Bet365 app download.Yu2021YearReceived scholarship Bet365 app download China Scholarship Council,About to go to Seoul National University in South Korea for joint doctoral training。atPhD period,He has Bet365 app download in school、National Scholarship,And has been rated as an outstanding student and outstanding student cadre at the school level for many times,Yu2021yearRated as a model for outstanding graduate cadres of Bet365 app download.He is hereACM MobiCom, SenSys, 4 papers published in top international conferences and journals Bet365 app download IEEE TMC,whereIndependentOne Bet365 app download3 articles, 3 patent Bet365 app download, first prize in provincial competition1 Bet365 app download


Bet365 app download title three: Dual privacy protection and quality-aware incentive mechanism for federated learning

Bet365 app download time:2022Year9month1Sunday morning10:00-10:30

Reporter: Bet365 app download

Reporting location: Shenghua back building409 / Tencent Conference 705-467-296


Abstract: Bet365 app download protection and incentive mechanism are two basic issues in federated learning,The purpose is to protect the Bet365 app download of data owners and motivate them to share more resources respectively。Recent research proposed based on differential Bet365 app download(DP)'s Bet365 app download protection incentive mechanism to solve these two problems at the same time.However,Almost all studies use Bet365 app download level as the only incentive item,without considering other factors such as the quantity and quality of data。In addition,Untrusted servers can further infer sensitive information from bids that reflect the true cost to the data owner。In order to solve these problems,This article proposes a dual Bet365 app download protection and quality-aware federated learning incentive mechanism,PrivAimSpecifically,We use differential Bet365 app download to protect local models and true costs from leaking Bet365 app download to untrusted parameter servers,And carefully designed an incentive mechanism based on multi-dimensional reverse auction,Let the central server incentivize high quality without knowing the real quotation、Low-cost data owner participationFLWe proved Bet365 app download theoreticallyPrivAimSatisfiedΔb-Authenticity, individual rationality, computational efficiency and differential Bet365 app download.A large number of experiments Bet365 app download thatPrivAimAble to effectively protect bidding Bet365 app download,Compared to the current state-of-the-art technology,The social welfare and model accuracy have been improved by at least21%and6%


Bet365 app download Topic 4: Streaming Data Truth Exploration for Differential Privacy Protection in Crowdsourcing SystemsSO

Bet365 app download time:2022Year9month1Sunday morning10:30-11:00

Reporter: Bet365 app download

Reporting location: Shenghua back building409 / Tencent Conference 705-467-296


Abstract: Truth discovery is possible when user quality is unknown,Estimate user quality and task true value simultaneously,Gradually becoming one of the mainstream data aggregation algorithms。Differential Bet365 app download(DP)Due to its strong Bet365 app download protection, inBet365 app download value discoveryThe field has received more and more attention.But the existing truth discovery algorithms based on differential Bet365 app download are not suitable for streaming data,Because they ignore the correlation of the true value of the data in time。In addition,Existing differential Bet365 app download algorithms for streaming data are not suitable for truth discovery,Because they don’t consider the user when perturbing the dataReliabilityDifference,Therefore, directly applying existing algorithms to streaming crowdsourcing data will lead to poor data aggregation effects。To solve this problem,This paper proposes a Bet365 app download-preserving truth discovery mechanism based on edge computing for streaming crowdsourced data,PrivSTD, in Bet365 app downloadBet365 app downloadSimultaneous realization of Bet365 app downloadHigh-precision aggregated Bet365 app download valueSpecifically,We introduce edge servers between untrusted cloud servers and users,Implemented secure calculation of local truth values ​​and user reliability。For each edge server,Propose a budget recovery mechanism based on truth value,Adaptively determine the differential Bet365 app download perturbation timestamp and allocate the Bet365 app download budget according to the changing rules of the local truth value。In addition,A reliability-based perturbation mechanism is also proposed,Further reduce the size of disturbance noise based on user reliability。We are rightPrivSTD’s Bet365 app download utility and computational cost were theoretically analyzed,ProvedPrivSTDCan satisfyw-EventDifferential Bet365 app download。Extensive experimental results on synthetic and real data sets show that,PrivSTDAchieved ratioCurrently the mostBetter effectiveness of advanced Bet365 app download.


Introduction to Dr. Bet365 app download:

Wang Dan, Bet365 app download2018 PhD,Computer Science and Technology Major,Instructor Academician Zhang Yao,Instructor Professor Ren Ju,Research direction is edge computing、Data security and Bet365 app download protection、Deep learning, etc.。At HPCC,IEEE TMC、 3 papers published in conferences and journals such as Information Fusion。


Bet365 app download title 5: Research on cooperative learning and task migration algorithms for edge intelligence

Bet365 app download time:2022Year9month1Sunday morning11:00-11:30

Reporter: Bet365 app download

Reporting location: Tencent Conference 705-467-296

Abstract:With mobile Internet、Internet of Things、The booming development of artificial intelligence technology,And the rapid rise of new computing paradigms represented by mobile Bet365 app download computing,The concept of “Bet365 app download intelligence” has received widespread attention and research in recent years。Bet365 app download intelligence aims to train models、Migrating reasoning and decision-making from the traditional cloud to the Bet365 app download,make itTaking full advantage of network Bet365 app download devicesBet365 app download, storage and computingResource,To provide security、Low consumption、In time、Reliable smart service。However,The development of edge intelligence is still in its infancy,Its effective deployment faces many challenges,For example, the terminal device has limited data、How computing power supports effective model training,Edge network and edge server dynamics、How to efficiently operate and manage heterogeneous communication and computing resources, etc.。In this Bet365 app download,The speaker will introduce his recent work in these directions,Includes fast convergence、Personalized cooperative learning method with low communication overhead,Based onADMM’s edge-side continuous cooperative Bet365 app download solution,And distributed efficient task migration strategy。


Bet365 app download topic six:Model-based offline inverse reinforcement Bet365 app download algorithm design

Bet365 app download time:2022Year9month1Sunday morning11:30-12:00

Reporter: Bet365 app download

Reporting location: Tencent Conference 705-467-296

Abstract:Offline inverse reinforcement learning aims to learn reward functions and policies from fixed Bet365 app download sets (such as demo Bet365 app download),And no reward information and online exploration process are required。Despite the recent good progress,However, existing methods have poor generalization capabilities in unknown environments,Hindering its effective deployment in real environments。To solve this problem,This article designs a model-based offline inverse reinforcement learning algorithm,NamedCLARE,Aims to use the simulation Bet365 app download generated by the model to enhance the generalization ability of the algorithm。Through carefully designed training objective function,CLAREExpert intent and conservatism can be combined (conservatism) concisely integrated into the learned reward function。The reward function can then be guided by conservative offline Bet365 app download exploration“Safe” imitation Bet365 app download。We theoretically quantify exploration by limiting the cumulative return gap to experts-Develop the impact of trade-offs and Bet365 app download demonstrate CLARE The right trade-off can be made by minimizing the corresponding gap。Our experiments confirmed Bet365 app download CLARE Significant performance improvement over existing state-of-the-art algorithms on offline reinforcement Bet365 app download benchmarks。


Introduction to Dr. Bet365 app download:

Yue Sheng is a doctoral candidate at the School of Computer Science, Bet365 app download.2017Bachelor of Science degree from Bet365 app download,In the same year, he was recommended to study for a doctoral degree at the School of Computer Science, Bet365 app download。Yu2018Served at the Institute of Cross-Bet365 app download, Tsinghua University for a period of 34736_347534Month visiting doctoral student, worked at2020Year2Month Solstice2022Year2Going to Arizona State University for a period of study sponsored by China Scholarship Council2years of joint training。During my Ph.D. studies, I won many awards from the school、National Scholarship。His research direction mainly focuses on edge intelligence、Distributed optimization、Reinforcement Bet365 app download。While studying for PhD,He is hereIEEE JSACIEEE TPDSand other Bet365 app download andACM MobiHocPublished in other Bet365 app download5papers。Served in many top international publications、Reviewer Bet365 app download conference,includesIEEE JSACIEEE INFOCOM 2021/2022IEEE MobiHoc 2021AISTATS 2022Wait.


Address: Information Building of Bet365 app download, Yuelu District, Changsha City, Hunan Province Contact: 0731-88836659

Copyright ® 2017-2019 School of Computer Science, Bet365 app download All Rights Reserved.