SJTU AISIG
SJTU Artificial Intelligence Special Interest Group
研究方向
SJTU AISIG 团队围绕分布式计算与无服务器计算为主要研究方向,致力于构建智能云计算系统。
围绕分布式计算展开联邦学习和并行算法两方面的研究。联邦学习方面研究系统安全性、隐私性、异构性和高效性:安全性主要研究存在拜占庭客户端时,保障系统收敛性;隐私性主要研究如何防御针对模型的隐私攻击;异构性主要研究针对数据异构的个性化联邦学习,和针对设备异构的系统资源调度优化;高效性主要研究模型训练过程中的通信及计算优化。并行算法方面针对于单并行、多并行和自动并行三方面展开,寻找高效的并行方案。
围绕无服务器计算展开应用部署、资源调度和系统安全三方面的研究。应用部署方面研究特定应用在无服务器计算平台的部署和优化,包括机器学习训练与推理、MapReduce大数据处理等。资源调度方面研究如何为无服务器计算提供细粒度的系统资源调度机制,包括空闲资源收集、异构资源等。系统安全研究如何保证轻量级应用的安全执行机制,包括资源隔离、安全监控、认证授权与数据保护等。
新闻速递
No news so far...
代表论文
-
JSAC (CCF-A)SMSS: Stateful Model Serving in Metaverse with Serverless Computing and GPU SharingIEEE Journal on Selected Areas in Communications, 2023
-
SIGKDD (CCF-A)FedCP: Separating Feature Information for Personalized Federated Learning via Conditional PolicyIn Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2023
-
AAAI (CCF-A)Fedala: Adaptive local aggregation for personalized federated learningIn Proceedings of the AAAI Conference on Artificial Intelligence, 2023
-
JSAC (CCF-A)Automatic Pipeline Parallelism: A Parallel Inference Framework for Deep Learning Applications in 6G Mobile Communication SystemsIEEE Journal on Selected Areas in Communications, 2023
-
JSAC (CCF-A)OSTTD: Offloading of splittable tasks with topological dependence in multi-tier computing networksIEEE Journal on Selected Areas in Communications, 2022
-
AAAI (CCF-A)Improving bayesian neural networks by adversarial samplingIn Proceedings of the AAAI Conference on Artificial Intelligence, 2022
-
CVPR (CCF-A)Robust bayesian neural networks by spectral expectation bound regularizationIn Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021
-
ICCV (CCF-A)Self-supervised vessel segmentation via adversarial learningIn Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021
-
AAAI (CCF-A)Reinforcing Neural Network Stability with Attractor DynamicsIn Proceedings of the AAAI Conference on Artificial Intelligence, 2020