The 18th International Symposium on Modeling and Optimization in Mobile, Ad Hoc and Wireless Networks (WiOpt 2020)
Session Keynote-2
Keynote 2
Conference
5:30 PM
—
6:45 PM EEST
Local
Jun 17 Wed, 7:30 AM
—
8:45 AM PDT
Learning and Resource Allocation in Networks: Finite Time Bounds and Insights
Sanjay Shakkottai (University of Texas at Austin)
0
In this talk, we discuss classical resource allocation problems viewed through a finite time learning (regret) perspective. First, we look at (queueing+learning) in a multi-armed bandit setting, where queues need to be matched to servers (aka arms of a bandit, e.g. channels) whose service rates are unknown. We study algorithms that minimize queue-regret: the expected difference between the queue-lengths (backlogs) obtained by the algorithm, and those obtained by a genie-aided matching algorithm that knows exact service rates. A naive view of this problem would suggest that queue-regret could grow logarithmically (since queue-regret cannot be larger than classical regret). Our work shows surprisingly more complex behavior — specifically, the optimal queue-regret decreases with time and scales as O(1/t). Next, we consider a (networks+learning) bandit, where N agents collaborate by exchanging only arm recommendations through pairwise gossip to determine the best resource (aka arm). We establish that even with very limited communications, the regret per agent is a factor of order N smaller compared to the case of no collaborations. Furthermore, we show that the communication constraints only have a second order effect on regret. In both these problems, we discuss the new insights for resource allocation that emerge from a bandit and learning perspective.
Session Chair
Vijay Subramanian (University of Michigan)
Session Keynote-3
Keynote 3
Conference
7:00 PM
—
8:15 PM EEST
Local
Jun 17 Wed, 9:00 AM
—
10:15 AM PDT
Scaling Blockchains to Physical Limits
David Tse (Stanford University)
0
Bitcoin is the world’s first large-scale decentralized trust system. It has excellent security, robust to any attacker with less than 50% of the system’s resource. However, its throughput is 7 transactions per second, and its confirmation latency is hours. This talk describes how the performance of Bitcoin can be scaled up to the physical limits of the underlying communication network while maintaining its strong security. A live demo of an implemented system with 10,000X the throughput of Bitcoin will be presented.
Session Chair
Vijay Subramanian (University of Michigan)
Made with in Toronto · Privacy Policy · © 2022 Duetone Corp.