Computer Networking A Top-down Approach Pdf Github
Sourabh Doherty
Fei Dou is an Assistant Professor in the School of Computing at the University of Georgia. She obtained her Ph.D. degree from the Department of Computer Science and Engineering at the University of Connecticut inLaboratory of Machine Learning & Health Informatics, working on Machine Learning (ML) / Artificial Intelligence (AI) in the Internet of Things (IoT) / Cyber-Physical Systems (CPS)supervised by Prof. Jinbo Bi. Previously, she was working on Underwater Acoustic Sensor Networks(UWASN) in Tianjin University supervised by Prof. Zhigang Jin.
Fei's research lies in analyzing and resolving challenges associated with Machine Intelligence of UbiquitousComputing in Distributed Systems, including:
- inefficiency and low scalability of trained models, especially when thesolution space is large;
- security and privacy concerns of user data;
- data heterogeneity across devices and imbalanceddata distribution on individual devices;
- communication bottlenecks and high computational costs in distributed systems.
Latent Orthonormal Contrastive Learning in Disaster Damage Assessment Using Paired Remote Sensing Imagery
Fei Dou, Cameron Cianci, Jinbo Bi
Submitted
A latent orthonormal contrastive learning approach is proposed to handle the high-resolutionsatellite/aerial imagery.
On-Device Indoor Positioning: A Federated Reinforcement Learning Approach with Heterogeneous Devices
Fei Dou, Jin Lu, Tan Zhu, Jinbo Bi
IOT-J 2023 [paper]
A personalized federated learning (FL) for reinforcement learning (RL) is proposed to automatically learnenvironmental dynamics by client-environment interactions via RL and cope with the diversity of client devicesand their non-identical data distributions via personalized FL.
A Bisection Reinforcement Learning Approach to 3-D Indoor Localization
Fei Dou, Jin Lu, Tingyang Xu, Chun-Hsi Huang, and Jinbo Bi
IOT-J 2021 [paper]
A bisection reinforcement learning method is proposed to bisect the search space in a hierarchy from the entirebuilding down to a prespecified distance scale to the object position, and a unified framework for single-floor,multifloor, and 3-D indoor localization is proposed.
Top-Down Indoor Localization with Wi-Fi Fingerprints using Deep Q-Network
Fei Dou, Jin Lu, Zigeng Wang, Xia Xiao, Jinbo Bi, Chun-Hsi Huang
MASS 2018 [paper]
A top-down searching method using a deep Q-network agent is proposed to tackle environment dynamics inindoor positioning with Wi-Fi fingerprints, by formulating the indoor localizationproblem as a Markov Decision Process rather than a typical classification or regression problem.
A Smart Narrow Down Approach based on Machine Learning for Indoor localization
Sahibzada Umair, Fei Dou, Tughrul Arslan
Under Review by IEEE Internet of Things Journal (IOT-J)
A Narrow down approach has been presented for indoor localization that involves in coarse andaccurate positioning phases, where training points selection, area division and overlapping strategies have beenpresented to reduce the uncertainty.
On-demand Pipelined MAC for Multi-hop Underwater Wireless Sensor Networks
Fei Dou, Zheng Peng
WuWNet 2015,[paper]
An on-demand pipeline is established to enable data transmission from the source to the destination over multiplehops in a short time, where a special acknowledgement mechanism is designed to guarantee the reliability ofthe communication with low overheads.
The Multi-channel MAC Protocol for High Performance Underwater Sensor Networks
Fei Dou, Zhigang Jin, Yao Zhang, Yishan Su
Journal of Harbin Engineering University 2015, CWSN 2013, [paper]
A multi-channel MAC protocol is proposed to tackle the spatial-temporal uncertainty by constructing the reservation model of the control channel using Markov Chain.
Motion Prediction Based MAC for Underwater Wireless Sensor Networks
Yishan Su, Zhigang Jin, Zixin Liu, Fei Dou*
Journal of Electronics & Information Technology 2013, [paper]
The motion model of underwater nodes is established to address the multiple access problem in underwater mobile networks,with the aid of an AutoRegression (AR) mobility prediction algorithm.
My teaching experience include but not limit to conducting lab sessions, grading homework, holding office hours, answering questions from students, conducting review sessions, helping creating and proctoring the written exams, invigilating exams in a separate room that meets the special needs of students with disabilities or medical conditions, etc..
In recognition of her ability to motivate and effectively teach students, Fei was among the only few students who received the "Provost Recognition of Teaching Excellence" award from UCONN in 2017.
Giving lectures on basics of Markov Decision Process and Reinforcement Learning,including Bellman Equation, Value Iteration, Policy Iteration, Sampling Policy, Exploitation and Exploration,Model-based Learning, Model-free Approach, Monte Carlo Method, Temporal Difference, off-policy Q-Learning, on-policy SARSA, etc.Giving talks on Applications of Reinforcement Learning in real-world scenarios.
Fei Dou's research interests lie in analyzing and resolving challenges associated with Machine Intelligence of UbiquitousComputing in Distributed Systems, including:
- inefficiency and low scalability of trained models, especially when thesolution space is large;
- security and privacy concerns of user data;
- data heterogeneity across devices and imbalanceddata distribution on individual devices;
- communication bottlenecks and high computational costs in distributed systems.
In Fall of 2019 I started a Slack channel on AI for Conservation, to provide a shared, interdisciplinary space for researchers who work across the fields of computer vision, machine learning, and AI for conservation and sustainability applications to share opportunities, discuss best practices, and find collaborators. Now our community is over 1000 strong, with researchers from all over the globe. If you'd like to join us, just email (aiforcon...@gmail.com)
Nunc lacinia ante nunc ac lobortis. Interdum adipiscing gravida odio porttitor sem non mi integer non faucibus ornare mi ut ante amet placerat aliquet. Volutpat eu sed ante lacinia sapien lorem accumsan varius montes viverra nibh in adipiscing blandit tempus accumsan.
Fringilla nisl. Donec accumsan interdum nisi, quis tincidunt felis sagittis eget. tempus euismod. Vestibulum ante ipsum primis in faucibus vestibulum. Blandit adipiscing eu felis iaculis volutpat ac adipiscing accumsan eu faucibus. Integer ac pellentesque praesent tincidunt felis sagittis eget. tempus euismod. Vestibulum ante ipsum primis in faucibus vestibulum. Blandit adipiscing eu felis iaculis volutpat ac adipiscing accumsan eu faucibus. Integer ac pellentesque praesent. Donec accumsan interdum nisi, quis tincidunt felis sagittis eget. tempus euismod. Vestibulum ante ipsum primis in faucibus vestibulum. Blandit adipiscing eu felis iaculis volutpat ac adipiscing accumsan eu faucibus. Integer ac pellentesque praesent tincidunt felis sagittis eget. tempus euismod. Vestibulum ante ipsum primis in faucibus vestibulum. Blandit adipiscing eu felis iaculis volutpat ac adipiscing accumsan eu faucibus. Integer ac pellentesque praesent.
We explore building generative neural network models of popular reinforcement learning environments. Our world model can be trained quickly in an unsupervised manner to learn a compressed spatial and temporal representation of the environment. By using features extracted from the world model as inputs to an agent, we can train a very compact and simple policy that can solve the required task. We can even train our agent entirely inside of its own dream environment generated by its world model, and transfer this policy back into the actual environment.
Humans develop a mental model of the world based on what they are able to perceive with their limited senses. The decisions and actions we make are based on this internal model. Jay Wright Forrester, the father of system dynamics, described a mental model as:
One way of understanding the predictive model inside of our brains is that it might not be about just predicting the future in general, but predicting future sensory data given our current motor actions . We are able to instinctively act on this predictive model and perform fast reflexive behaviours when we face danger , without the need to consciously plan out a course of action.
c80f0f1006