International Workshop on Federated Learning for User Privacy and Data Confidentiality
in Conjunction with ICML 2020 (FL-ICML'20)
Workshop Date: July 18, 2020
How to join the workshop: Please go to https://icml.cc/virtual/2020/workshop/5730 and click on "Join Zoom" at the top of the page (requires ICML registration).
Workshop Program (Eastern Daylight Time (EDT))
Time | Activity |
---|---|
08:45 – 09:00 | Presenters to connect and test the system |
09:00 – 09:10 | Opening Address |
09:10 – 09:35 | Keynote Session 1: Balancing Efficiency and Security in Federated Learning, by Qiang Yang (WeBank) |
09:35 – 10:25 | Technical Talks Session 1 (4 talks, 12 mins each)
|
10:25 – 10:40 | Break (Presenters should connect and test the system) |
10:40 – 11:05 | Keynote Session 2: Federated Learning in Enterprise Settings, by Rania Khalaf (IBM Research) |
11:05 – 11:35 | Lightning Talks Session 1 (9 talks, 3 mins each)
|
11:35 – 12:05 | Poster Session 1 (lightning talk presenters) |
12:05 – 13:20 | Lunch (Presenters to re-connect at 13:15 PM and test the system) |
13:20 – 13:45 | Keynote Session 3: Federated Learning Applications in Alexa, by Shiv Vitaladevuni (Amazon Alexa) |
13:45 – 15:10 | Technical Talks Session 2 (7 talks, 12 mins each)
|
15:10 – 15:25 | Break (Presenters should connect and test the system) |
15:25 – 15:50 | Keynote Session 4: The Shuffle Model and Federated Learning, by Ilya Mironov (Facebook) |
15:50 – 16:15 | Lightning Talks Session 2 (8 talks, 3 mins each)
|
16:15 – 16:45 | Poster Session 2 (lightning talk presenters) |
16:45 – 17:00 | Break |
17:00 – 17:25 | Keynote Session 5: Advances and Open Problems in Federated Learning, by Brendan McMahan (Google) |
17:25 – 17:35 | Closing Remarks |
Keynote Abstracts
Abstract: Federated learning systems need to balance the efficiency and security of machine learning algorithms while maintaining model accuracy. In this talk we discuss this trade-off in two settings. One is when two collaborating organisations wish to transfer the knowledge from one to another via a federated learning framework. We present a federated transfer learning algorithm to both improve the security and the performance while preserving privacy. Another case is when one exploits differential privacy in a federated learning framework to ensure efficiency, but this may cause security degradation. To solve the problem, we employ a dual-headed network architecture that guarantees training data privacy by exerting secret gradient perturbations to original gradients, while maintaining high performance of the global shared model. We find that the combination of secret-public networks provides a preferable alternative to DP-based mechanisms in federated learning applications.
Biography: Qiang Yang is Chief Artificial Intelligence Officer of WeBank and Chair Professor of CSE Department of Hong Kong Univ. of Sci. and Tech. He is the Conference Chair of AAAI-21, President of Hong Kong Society of Artificial Intelligence and Robotics(HKSAIR) and a former President of IJCAI (2017-2019). He is a fellow of AAAI, ACM, IEEE and AAAS. His research interests include transfer learning and federated learning. He is the founding EiC of two journals: IEEE Transactions on Big Data and ACM Transactions on Intelligent Systems and Technology.
Abstract: Federated learning in consumer scenarios has garnered a lot of interest. However, its application in large enterprises brings to bear additional needs and guarantees. In this talk, I will highlight key drivers for federated learning in enterprises, illustrate representative uses cases, and summarize the requirements for a platform that can support it. I will then present the newly released IBM Federated Learning framework (git, white paper) and show how it can be used and extended by researchers. Finally, I will highlight recent advances in federated learning and privacy from IBM Research.
Biography: Rania Khalaf is the Director of AI Platforms and Runtimes at IBM Research where she leads teams pushing the envelope in AI platforms to make creating AI models and applications easy, fast, and safe for data scientists and developers. Her multi-disciplinary teams tackle key problems at the intersection of core AI, distributed systems, human computer interaction and cloud computing. Prior to this role, Rania was Director of Cloud Platform, Programming Models and Runtimes. Rania serves as a Judge for the MIT Solve AI for Humanity Prize, on the Leadership Challenge Group for MIT Solve's Learning for Girls and Women Challenge and on the Advisory Board of the Hariri Institute for Computing at Boston University. She has received several Outstanding Technical Innovation awards for major impact to the field of computer science and was a finalist for the 2019 MassTLC CTO of the Year award.
Abstract: Alexa is a virtual assistant AI technology launched by Amazon in 2014. One of key enabling technologies is wakeword, which allows users to interact with Alexa devices hands-free via voice. We present some of the unique ML challengesposed in wakeword, and how Federated Learning can be used to address them. We also present some considerations when bringing Federated Learning to consumer grade, embedded applications.
Biography: Shiv Vitaladevuni is a Senior Manager in Machine Learning at Amazon Alexa, focusing on R&D for Alexa family of devices such as Echo, Dot, FireTV, etc. At Amazon, Shiv leads a team of scientists and engineers inventing embedded speechand ML products used by millions of Alexa customers across all Alexa devices, around the globe. His team conducts research in areas such as Federated ML, Large scale semi/unsupervised learning, User diversity and fairness in ML, Speaker adaptation and personalization,memory efficient deep learning models, etc. Prior to Amazon, Shiv worked on video and text document analysis at Raytheon BBN Technologies, and bio-medical image analysis at Howard Hughes Medical Institute.
Abstract: The shuffle model of computation, also known as the Encode-Shuffle-Analyze (ESA) architecture, is a recently introduced powerful approach towards combining anonymization channels and differentially private distributed computations. We present general results about amplification-by-shuffling unlocked by ESA, as well as more specialized theoretical and empirical findings. We discuss challenges of instantiating the shuffle model in practice.
Biography: Ilya Mironov obtained his Ph.D. in cryptography from Stanford in 2003. In 2003-2014 he was a member of Microsoft Research-Silicon Valley Campus, where he contributed to early works on differential privacy. In 2015-2019 he worked in Google Brain. Since 2019 he has been part of Facebook AI working on privacy-preserving machine learning.
Abstract: Motivated by the explosive growth in federated learning research, 22 Google researchers and 36 academics from 24 institutions collaborated on a paper titled Advances and Open Problems in Federated Learning. In this talk, I will survey some of the main themes from the paper, particularly the defining characteristics and challenges of different FL settings. I will then briefly discuss some of the ways FL increasingly powers Google products, and also highlight several exciting FL research results from Google.
Biography: Brendan McMahan is a research scientist at Google, where he leads efforts on decentralized and privacy-preserving machine learning. His team pioneered the concept of federated learning, and continues to push the boundaries of what is possible when working with decentralized data using privacy-preserving techniques. Previously, he has worked in the fields of online learning, large-scale convex optimization, and reinforcement learning. Brendan received his Ph.D. in computer science from Carnegie Mellon University.
Awards
Accepted Full Papers
Accepted Short Papers
Call for Papers
Despite the advantages of federated learning, and its successful application in certain industry-based cases, this field is still in its infancy due to new challenges that are imposed by limited visibility of the training data, potential lack of trust among participants training a single model, potential privacy inferences, and in some cases, limited or unreliable connectivity.
The goal of this workshop is to bring together researchers and practitioners interested in FL. This day-long event will facilitate interaction among students, scholars, and industry professionals from around the world to understand the topic, identify technical challenges, and discuss potential solutions. This will lead to an overall advancement of FL and its impact in the community. Topics of interest include, but are not limited to, the following:
Proceedings and Dual Submission Policy
Submission Instructions
Easychair submission link: https://easychair.org/conferences/?conf=flicml20
Submission Due (Final): May 17, 2020 June 10, 2020 (23:59 UTC-12)
Notification Due: May 31, 2020 June 30, 2020 (23:59 UTC-12)
If you have any enquiries, please email us at: flworkshop.icml.2020@gmail.com
Organizing Committee
Program Committee
Organized by