Advanced Topics in AI for Networking (Summer 2022): Difference between revisions

From NET Wiki
Jump to navigation Jump to search
No edit summary
 
(34 intermediate revisions by the same user not shown)
Line 3: Line 3:
{{CourseDetails
{{CourseDetails
|credits=5 ECTS
|credits=5 ECTS
|module=M.Inf.1222.Mp: Specialization Computer Networks Module Description ''-or-'' 3.10: Advanced Topics in Internet Research (II)(ITIS); M.Inf.1223 (new Regulations)
|module=M.Inf.1123 (new Regulations)
|lecturer=[http://134.76.18.81/?q=people/prof-dr-xiaoming-fu Prof. Xiaoming Fu];  [http://www.net.informatik.uni-goettingen.de/?q=people/dr-tingting-yuan, Dr. Tingting Yuan]
|lecturer=[http://134.76.18.81/?q=people/prof-dr-xiaoming-fu Prof. Xiaoming Fu];  [http://www.net.informatik.uni-goettingen.de/?q=people/dr-tingting-yuan, Dr. Tingting Yuan]
|ta=[NA]
|ta=[NA]
|time=Wednesday 14:00-16:00
|time=Wednesday 14:00-16:00
|place=IfI 0.101
|place=IfI 0.101
|univz=[https://univz.uni-goettingen.de/qisserver/rds?state=verpublish&status=init&vmfile=no&publishid=296747&moduleCall=webInfo&publishConfFile=webInfo&publishSubDir=veranstaltung]
}}
}}


Line 16: Line 15:
==Course Overview==
==Course Overview==
The purpose of this seminar is to discuss some advanced topics in '''computer networks'''. This course is a '''theory-oriented''' research seminar (5 ECTS, 2 SWS), held on a weekly base and comprises the following components:
The purpose of this seminar is to discuss some advanced topics in '''computer networks'''. This course is a '''theory-oriented''' research seminar (5 ECTS, 2 SWS), held on a weekly base and comprises the following components:
* Weekly Presentation + Weekly Paper Reading and Discussion 50%
* Weekly Presentation + Weekly Paper Reading and Discussion 40%
* Final Presentation 25%
* Final Presentation 30%
* Final Report 25%
* Final Report 30%


The material in the seminar is mainly drawn from the research literature in top journals/conferences, like ToN,TMC, TPDS, SIGCOMM, SIGMETRICS, INFOCOM, MOBICOM, MOBIHOC, WWW, CoNEXT.
The material in the seminar is mainly drawn from the research literature in top journals/conferences, like ToN,TMC, TPDS, SIGCOMM, SIGMETRICS, INFOCOM, MOBICOM, MOBIHOC, WWW, CoNEXT.
Line 26: Line 25:
** Summary of the paper
** Summary of the paper
** Pros and cons of the paper (your conclusion)  
** Pros and cons of the paper (your conclusion)  
** '''NOTE!! Every participant should provide the paper review BEFORE the seminar (23:59 Wednesday). => the review form is available at [[http://user.informatik.uni-goettingen.de/~fu/Paper_Review_Form_ATCN_WS201112.doc Paper_Review_Form_ATCN_WS201112.doc]]
** '''NOTE!! Every participant should provide the paper review BEFORE the seminar (23:59 Tuesday). => the review form is available at [[http://user.informatik.uni-goettingen.de/~fu/Paper_Review_Form_ATCN_WS201112.doc Paper_Review_Form_ATCN_WS201112.doc]]
* During each weekly seminar, one participant is assigned for presenting the paper (each presentation lasts for ~20 minutes) and the list of pros and cons are discussed by all the participants.
* During each weekly seminar, one participant is assigned for presenting the paper (each presentation lasts for ~20 minutes) and the list of pros and cons are discussed by all the participants.
* In the middle of the semester, everyone is requested to prepare:
* In the middle of the semester, everyone is requested to prepare:
Line 33: Line 32:


==List of Papers ==
==List of Papers ==
1. Lightweight and Robust Representation of Economic Scales from Satellite Imagery [https://ojs.aaai.org/index.php/AAAI/article/download/5379/5235]


2. AutoML for Video Analytics with Edge Computing [https://ieeexplore.ieee.org/abstract/document/9488704]
Demo: Pollux: Co-adaptive Cluster Scheduling for Goodput-Optimized Deep Learning [https://www.usenix.org/conference/osdi21/presentation/qiao][A]


3. Source Compression with Bounded DNN Perception Loss for IoT Edge Computer Vision [https://dl.acm.org/doi/10.1145/3300061.3345448]
1. Lightweight and Robust Representation of Economic Scales from Satellite Imagery [https://ojs.aaai.org/index.php/AAAI/article/download/5379/5235] [L]


4. NN-Meter: Towards Accurate Latency Prediction of Deep-Learning Model Inference on Diverse Edge Devices [https://dl.acm.org/doi/10.1145/3458864.3467882]
2. AutoML for Video Analytics with Edge Computing [https://ieeexplore.ieee.org/abstract/document/9488704][A]


5. Reducto: On-Camera Filtering for Resource-Efficient Real-Time Video Analytics [https://dl.acm.org/doi/10.1145/3387514.3405874]
3. Source Compression with Bounded DNN Perception Loss for IoT Edge Computer Vision [https://dl.acm.org/doi/10.1145/3300061.3345448][A]


6. Deep Interest Network for Click-Through Rate Prediction [https://arxiv.org/pdf/1706.06978.pdf]
4. NN-Meter: Towards Accurate Latency Prediction of Deep-Learning Model Inference on Diverse Edge Devices [https://dl.acm.org/doi/10.1145/3458864.3467882] [B]


7. Energy-Efficient 3D Vehicular Crowdsourcing For Disaster Response by Distributed Deep Reinforcement Learning [https://dl.acm.org/doi/pdf/10.1145/3447548.3467070]
5. Reducto: On-Camera Filtering for Resource-Efficient Real-Time Video Analytics [https://dl.acm.org/doi/10.1145/3387514.3405874] [B]
 
6. Deep Interest Network for Click-Through Rate Prediction [https://arxiv.org/pdf/1706.06978.pdf] [L]
 
7. Energy-Efficient 3D Vehicular Crowdsourcing For Disaster Response by Distributed Deep Reinforcement Learning [https://dl.acm.org/doi/pdf/10.1145/3447548.3467070] [B]


8. Reducing the Service Function Chain Backup Cost over the Edge and Cloud by a Self-Adapting Scheme [https://ieeexplore.ieee.org/document/9312434]
8. Reducing the Service Function Chain Backup Cost over the Edge and Cloud by a Self-Adapting Scheme [https://ieeexplore.ieee.org/document/9312434]


9. Routing on Multiple Optimality Criteria[https://www.lx.it.pt/~jls/publications_ficheiros/RoutingMultipleOptimalCriteria.pdf]
9. Routing on Multiple Optimality Criteria[https://dl.acm.org/doi/pdf/10.1145/3387514.3405864]


10. Understanding, Detecting and Localizing Partial Failures in Large System Software [https://www.cs.jhu.edu/~huang/paper/omegagen-nsdi20-preprint.pdf]
10. Understanding, Detecting and Localizing Partial Failures in Large System Software [https://www.cs.jhu.edu/~huang/paper/omegagen-nsdi20-preprint.pdf]


11. Understanding Lifecycle Management Complexity of Datacenter Topologies [https://www.cs.jhu.edu/~huang/paper/omegagen-nsdi20-preprint.pdf]
11. Understanding Lifecycle Management Complexity of Datacenter Topologies [https://www.cs.jhu.edu/~huang/paper/omegagen-nsdi20-preprint.pdf] [L]


12. ACC: Automatic ECN Tuning for High-Speed Datacenter Networks [https://dl.acm.org/doi/pdf/10.1145/3452296.3472927]
12. ACC: Automatic ECN Tuning for High-Speed Datacenter Networks [https://dl.acm.org/doi/pdf/10.1145/3452296.3472927][A]


13. Seven Years in the Life of Hypergiants’ Off-Nets [https://dl.acm.org/doi/pdf/10.1145/3452296.3472928]
13. Seven Years in the Life of Hypergiants’ Off-Nets [https://dl.acm.org/doi/pdf/10.1145/3452296.3472928]


==Schedule==
14. ATP: In-network Aggregation for Multi-tenant Learning [https://www.usenix.org/system/files/nsdi21-lao.pdf]
W1: Open Talk


W2: Assignment Topics
15. Segcache: a memory-efficient and scalable in-memory key-value cache for small objects[https://www.usenix.org/system/files/nsdi21-yang.pdf]


W3: Paper ID:
16. MAGE: Nearly Zero-Cost Virtual Memory for Secure Computation [https://www.usenix.org/system/files/osdi21-kumar.pdf]


W4: Paper ID:
==Schedule==
W1: Open Talk


W5: Paper ID:
W2: Assignment Topics and demo paper reading


W6: Paper ID:
W3: Paper ID: 1 (04.05)


W7: Paper ID:
W5: Paper ID: 3, 4 (18.05)


W8: Paper ID:
W7: Paper ID: 5, 12 (01.06)


W9: Paper ID:
'''!! 25.06 deadline for registration on Flexnow'''


W10: Paper ID:
W9: Rehearsal: 7, demo (15.06)


W11: Paper ID:
Final Presentation (29.06 maybe)
*'''Paper Title''': 7
*'''Paper Title''': demo


Final Presentation
Report deadline (30.07)
*'''Paper Title''':
*'''Paper Title''':
*'''Paper Title''':


==Final Presentations & Report==
==Final Presentations & Report==

Latest revision as of 10:03, 20 December 2022

Details

Workload/ECTS Credits: 5 ECTS
Module: M.Inf.1123 (new Regulations)
Lecturer: Prof. Xiaoming Fu; Dr. Tingting Yuan
Teaching assistant: [NA]
Time: Wednesday 14:00-16:00
Place: IfI 0.101


Announcements

Please contact me by email:tingting.yuan@cs.uni-goettingen.de if you have any questions.

Course Overview

The purpose of this seminar is to discuss some advanced topics in computer networks. This course is a theory-oriented research seminar (5 ECTS, 2 SWS), held on a weekly base and comprises the following components:

  • Weekly Presentation + Weekly Paper Reading and Discussion 40%
  • Final Presentation 30%
  • Final Report 30%

The material in the seminar is mainly drawn from the research literature in top journals/conferences, like ToN,TMC, TPDS, SIGCOMM, SIGMETRICS, INFOCOM, MOBICOM, MOBIHOC, WWW, CoNEXT.

Requirements

  • Each participant is required to read the assigned paper before the seminar and prepare the review of the paper, which should include the following parts:
    • Summary of the paper
    • Pros and cons of the paper (your conclusion)
    • NOTE!! Every participant should provide the paper review BEFORE the seminar (23:59 Tuesday). => the review form is available at [Paper_Review_Form_ATCN_WS201112.doc]
  • During each weekly seminar, one participant is assigned for presenting the paper (each presentation lasts for ~20 minutes) and the list of pros and cons are discussed by all the participants.
  • In the middle of the semester, everyone is requested to prepare:
    • Final report: Essay (5~6 pages, double columns, IEEE format) for your chosen research topic, which contains a comprehensive literature survey + a detailed discussion of some key enabling technologies
    • Final presentation: each presentation lasts for ~20 minutes, plus ~10 minutes Q&A

List of Papers

Demo: Pollux: Co-adaptive Cluster Scheduling for Goodput-Optimized Deep Learning [1][A]

1. Lightweight and Robust Representation of Economic Scales from Satellite Imagery [2] [L]

2. AutoML for Video Analytics with Edge Computing [3][A]

3. Source Compression with Bounded DNN Perception Loss for IoT Edge Computer Vision [4][A]

4. NN-Meter: Towards Accurate Latency Prediction of Deep-Learning Model Inference on Diverse Edge Devices [5] [B]

5. Reducto: On-Camera Filtering for Resource-Efficient Real-Time Video Analytics [6] [B]

6. Deep Interest Network for Click-Through Rate Prediction [7] [L]

7. Energy-Efficient 3D Vehicular Crowdsourcing For Disaster Response by Distributed Deep Reinforcement Learning [8] [B]

8. Reducing the Service Function Chain Backup Cost over the Edge and Cloud by a Self-Adapting Scheme [9]

9. Routing on Multiple Optimality Criteria[10]

10. Understanding, Detecting and Localizing Partial Failures in Large System Software [11]

11. Understanding Lifecycle Management Complexity of Datacenter Topologies [12] [L]

12. ACC: Automatic ECN Tuning for High-Speed Datacenter Networks [13][A]

13. Seven Years in the Life of Hypergiants’ Off-Nets [14]

14. ATP: In-network Aggregation for Multi-tenant Learning [15]

15. Segcache: a memory-efficient and scalable in-memory key-value cache for small objects[16]

16. MAGE: Nearly Zero-Cost Virtual Memory for Secure Computation [17]

Schedule

W1: Open Talk

W2: Assignment Topics and demo paper reading

W3: Paper ID: 1 (04.05)

W5: Paper ID: 3, 4 (18.05)

W7: Paper ID: 5, 12 (01.06)

!! 25.06 deadline for registration on Flexnow

W9: Rehearsal: 7, demo (15.06)

Final Presentation (29.06 maybe)

  • Paper Title: 7
  • Paper Title: demo

Report deadline (30.07)

Final Presentations & Report

  • Final Registration in FlexNow: To Be Announced (TBA).


  • Final Presentation:
    • Each for ~20 minutes, plus ~10 minutes Q&A


  • Final Presentation Slots:
    • To Be Announced (TBA)


Retrieved from "https://wiki.net.informatik.uni-goettingen.de/index.php?title=Advanced_Topics_in_AI_for_Networking_(Summer_2022)&oldid=7929"