Welcome to the 3rd edition of the KuVS Fachgespräch "Machine Learning & Networking (Malene)", October 6 & 7, 2022, at Hasso-Plattner-Institut in Potsdam. It is jointly supported by GI and ITG/VDE.
|Day||Time||Session / Paper title ||Authors||Folien|
|Thursday, Oct. 6th||10:00 - 12:00|
Joint working session: Tools, Frameworks, Data sets
| || |
| ||12:00 - 13:00 ||Lunch break|| || |
| ||13:00 - 13:30 ||Registration || || |
| ||13:30 - 15:30||Session 1: Basics of ML in and for networks|| || |
| ||13:30 - 14:00 ||Survey: Data-parallel Architectures for Distributed Machine Learning||Paeleke, Karl||PDF|
| ||14:00 - 14:30 ||Analyzing Sparsity of Deep Neural Networks in an Edge AI Split Computing Scenario||Haberer, Landsiedel||PDF|
| ||14:30 - 15:00 ||Resource-Aware Distributed Machine Learning on Heterogeneous IoT Devices||Samikwa, Braun||PDF|
| ||15:00 - 15:30||Preliminary Rate of Convergence Analysis of Constant Step-Size Distributed Stochastic Gradient Descent||Redder||PDF|
| ||15:30 - 16:00||Break|| || |
| ||16:00 - 17:30||Session 2: Reinforcement learning, multi-agent approaches|| || |
| ||16:00 - 16:30||Multi-Agent Reinforcement Learning for Coordination of Device-to-Device communication||Pochaba, Dorfinger, Herlich, Kwitt, Hirländer||PDF|
| ||16:30 - 17:00||Multi-Agent Distributed Reinforcement Learning for Making Decentralized Decisions||Tan, Khalili, Hecker, Karl|| |
| ||17:00 - 17:30||Towards Deep Learning Multiple Access (DLMA) through Multi-Agent Reinforcement Learning||Lindner|| |
| ||19:00||Joint dinner|| || |
|Friday, Oct. 7th||9:00 - 10:30||Session 3: Distributed learning|| || |
| ||9:00 - 9:30||Ensuring Fairness in Renewable-Aware Scheduling for Federated Learning||Agrawal, Wiesner|| |
| ||9:30 - 10:00||Distributed Optimal-Transport Clustering for Malicious User Rejection in Federated-Learning VANETs||Pacheco, Braun||PDF|
| ||10:00 - 10:30||Towards Autonomous Edge Cloud Maintenance using Machine-Learned Digital Twin Networks||Boltres||PDF|
| ||10:30 - 11:00||Break|| || |
| ||11:00 - 12:00||Session 4: Networking protocols|| || |
| ||11:00 - 11:30||Machine-learning based Multipath Wireless Access on an Infrastructure for Latency Critical AI-Applications||Poorzare, Wippel, Waldhorst, Zirpins||PDF|
| ||11:30 - 12:00||End-to-End is Not Enough: Towards a Coordinated Congestion Control (C3)||König, Zitterbart||PDF|
| ||12:00 - 13:00||Lunch break || || |
| ||13:00 - 14:30||Session 5: Network structure|| || |
| ||13:00 - 13:30||Passive Investigation of Networks and Classification of the discovered Assets according to the IT Basic Protection of the BSI||Schwinger, Meyer, Krille||PDF|
| ||13:30 - 14:00||How to Predict Flow-Level Network Traffic Evolution||Gholam Zadeh|| |
| ||14:00 - 14:30||WAN-GAN: Synthesizing Wide Area Network Topologies via Generative Adversarial Networks||Dietz, Seufert, Hoßfeld|| |
Work session: Tools, Frameworks, Data sets
To kickstart the discussions, we will start with a joint working session without a dedicated speaker. The idea is to collect sets of commonly used tools, data sets, or frameworks to ease entering into research into this field. The outcome of this session should at least be turned into a living document hosted by KuVS, possibly, a publication is conceivable (similar to Dagstuhl result white papers) as a survey or a tutorial. Everybody is cordially invited to attend.
To register, please send an email to email@example.com. There is no registration fee; participation is free of charge. Please do register nevertheless so that we can plan accordingly.
Procedure and Dates
- Registration deadline: Friday, September 30
- Location: At Hasso-Plattner-Institute, Griebnitzsee (travel hints) at building L, groud floor ("Eventfläche L")
What is a Fachgespräch?
Fachgespräche are a low-key version of workshops, organized (among others) by the Communication and Distributed Systems group of the Gesellschaft für Informatik. They foster community building by organising meetings on topics of current interest, with a low entry barrier. Typical target audience are people in their later Master or early PhD phase, but of course everybody is welcome. Participants are invited to present early ideas, work in progress, or also talk about work that has already been published - we deliberately do NOT require original research at a Fachgespräch. For PhD students new to the field of machine learning and networks, please also have a look at the summer school on AI in Networking Summer School.
What is Malene?
Malene is one such Fachgespräch, dedicated to the interaction of machine learning and networking. You can find earlier editions here: 2020a, 2020b, 2021 (co-located with NetSys conference).
What is special about the 3rd edition of Malene?
This time, we want to focus the discussion on how networking can assist machine learning systems and, in particular, distributed ML systems. E.g., how to ensure that distributed training is provided with the right data and runs at the right place in a network, taking all kinds of resource requirements into account.
- Andreas Blenk, TU Munich
- Holger Karl, HPI
- Olaf Landsiedel, Uni Kiel
- Stefan Schneider, Uni Paderborn
- Michael Seufert, Uni Würzburg