Embedded Files
Tutorials
Tutorials
Title: Coding-theoretic ideas for straggler mitigation in large-scale distributed matrix computation and parameter learning
Title: Coding-theoretic ideas for straggler mitigation in large-scale distributed matrix computation and parameter learning
Speaker: Aditya Ramamoorthy, Iowa State University, USA
Speaker: Aditya Ramamoorthy, Iowa State University, USA
Abstract: In modern computing environments, the growing demand for processing and storage has made distributed computing the technology of choice across many domains. However, large-scale distributed clusters, particularly in cloud settings, often suffer from the problem of stragglers—slow or failed workers that delay computation. Coded distributed computation addresses this challenge using ideas rooted in coding theory, introducing carefully designed redundancy across worker tasks so that the desired computation can be completed despite the presence of stragglers.
In the context of distributed matrix computations and distributed parameter learning, this framework has produced fundamental results characterizing the trade-off between redundancy and the level of straggler resilience that can be achieved. In this tutorial, we will provide an overview of the key ideas underlying coded computation for these applications. We will then discuss important practical considerations, including numerical stability during recovery, leveraging partial work from slow (but non-failed) workers, and handling scenarios with sparse inputs, highlighting representative works addressing these challenges. Our discussion will conclude with an overview of open problems and promising future research directions in this area.
Bio: Aditya Ramamoorthy is a Professor of Electrical and Computer Engineering, the John Ryder Professor of Engineering and (by courtesy) of Mathematics at Iowa State University. He received his B. Tech. degree in Electrical Engineering from the Indian Institute of Technology, Delhi and the M.S. and Ph.D. degrees from the University of California, Los Angeles (UCLA). His research interests are in the areas of classical/quantum information theory and coding techniques with applications to distributed computation, content distribution networks and machine learning.
Dr. Ramamoorthy currently serves as an editor for the IEEE Transactions on Information Theory (previous term from 2016 — 2019) and the IEEE Transactions on Communications from 2011 — 2015. He is the recipient of the Northrop Grumman professorship (2022 – 204), the 2020 Mid-Career Achievement in Research Award, the 2019 Boast-Nilsson Educational Impact Award and the 2012 Early Career Engineering Faculty Research Award from Iowa State University, the 2012 NSF CAREER award, and the Harpole-Pentair professorship in 2009-2010.
Title: “Optimize Now, Regret Later” - Online Convex Optimization: Standard, Constrained, and Multi-Objective Formulations
Title: “Optimize Now, Regret Later” - Online Convex Optimization: Standard, Constrained, and Multi-Objective Formulations
Speaker: Rahul Vaze, TIFR, India
Speaker: Rahul Vaze, TIFR, India
Abstract: In this tutorial, we will provide a gentle introduction to the area of online convex optimization (OCO), a framework for making decisions sequentially in an uncertain and possibly adversarial environment. With OCO, at each round, the learner chooses an action and suffers a loss driven by the evaluation of the action over a convex cost function that is revealed after committing to the action. The goal of the learner is not to be optimal in hindsight at every step (which is impossible), but to perform nearly as well as the best fixed decision chosen in hindsight, as measured by regret. Minimizing regret is challenging since cost functions can change arbitrarily over time and are revealed after an action has been chosen at each round.
OCO provides a unifying language for online learning, adaptive control, and real-time optimization with applications in spam filtering, portfolio optimization, learning with expert-advice, recommendation systems, etc.
We will consider three OCO formulations in the tutorial:
i) standard/unconstrained, where the objective is to minimize regret with respect to time varying cost functions,
ii) constrained, where in addition to cost functions, there are time varying constraints and an algorithm has to simultaneously minimize the regret and the constraint violation, and
iii) multi-objective, where the objective is to simultaneously minimize regret with respect to multiple sequences of time varying cost functions.
For all three themes, we will discuss optimal algorithms with surprisingly simple analysis, suitable for first time learners.
Bio: Rahul Vaze obtained his Ph.D. from The University of Texas at Austin in 2009. Currently he is an Associate Professor at the School of Technology and Computer Science, Tata Institute of Fundamental Research, Mumbai, India. His research interests are in communication networks, combinatorial resource allocation, online algorithms. He is the author of "Random Wireless Networks", Cambridge University Press, 2015. He is a co-recipient of the EURASIP best paper award for year 2010 for the Journal of Wireless Communication and Networking, the best paper award WiOpt 2020, and the best paper award Performance 2020.
Title: An introduction to lattice-based cryptography
Speaker: Amin Sakzad, Monash University, Australia
Title: An introduction to lattice-based cryptography
Speaker: Amin Sakzad, Monash University, Australia
Abstract: In this tutorial, I will cover the foundation of Euclidean lattices, their geometric properties, as well as the hard problems related to them. I will introduce LWE and SIS and their more structured variants like RLWE, RSIS, MLWE, MSIS, MP-LWE, and Hint-MLWE. I will then construct a few cryptographic primitives (including a public-key encryption scheme as well as a digital signature) and analyse their security and performances. I will end the tutorial with more advanced primitives, like IBE, THE, etc.
Bio: Dr. Amin Sakzad is an Associate Professor and Director of Research in the Department of Software Systems and Cybersecurity at Monash University. With a career spanning cutting-edge post-quantum cryptography, lattice coding, and secure wireless communications, Amin has become a global leader in shaping the cryptographic foundations that will safeguard the digital world against quantum threats.
His research on Euclidean lattices and post-quantum cryptography (PQC) has not only advanced theory but also translated into real-world adoption. Notably, his work on the FACCT sampler was adopted in Falcon, a NIST PQC standard. Beyond academia, his open-source cryptographic implementations have reached extraordinary global impact, with over 30 million downloads via the Legion of the Bouncy Castle platform.
Amin has attracted over $8M in competitive industry and government funding, including a landmark USD $1M project with the U.S. Department of State to advance PQC adoption across 11 Indo-Pacific nations.
As a mentor, Amin has guided PhD students who have gone on to careers at CSIRO, Amazon, ANU, and ONI, ensuring his impact extends across academia, government, and industry
Amin is also a decorated educator, having received multiple awards including the Monash Vice-Chancellor’s Teaching Excellence Award and recognition as a Senior Fellow of the Higher Education Academy (SFHEA).
Page updated
Report abuse