Gugan Chandrashekhar Mallika Thoppe

Eminent Speaker

Short CV: Gugan Thoppe has been an Assistant Professor at the Indian Institute of Science (IISc), Computer Science and Automation department, since 2019. Concurrently, he also serves as an Associate Researcher at the Robert Bosch Centre, IIT Madras. He received his Ph.D. in 2016 from the Tata Institute of Fundamental Research (TIFR), Mumbai. Following this, he did postdoctoral research at the Technion Institute of Technology, Israel (2015-17) and Duke University, USA (2017-19). Among his accolades are the Pratiksha Trust's Young Investigator Award, the IISc Award for Excellence in Teaching, and the TIFR award for the best Ph.D. thesis. His research spans a broad array of subjects including reinforcement learning, online learning, stochastic approximation, and random topology. His contributions are recognized for pioneering numerous innovative methodologies for the analysis of stochastic optimization algorithms, especially those in reinforcement learning.

Title of Talk 1: Exploring the Evolution of Reinforcement Learning: From Dynamic Programming to Deep Q-Networks

Synopsis: The field of Reinforcement Learning (RL) attempts to answer the question "Can a machine train itself to solve a complex task the way infants learn to sit, crawl and walk?” A notable milestone here has been the Deep Q-Network (DQN), an innovative algorithm that has surpassed human capabilities in playing sophisticated video games by learning directly from visual. In this talk, we will embark on a journey through the transformative landscape of RL, from its foundational principles to the development of DQN. We will begin with experimental demonstrations that showcase DQN's remarkable abilities. To understand the principles behind its success, we will then play an interactive game called "But Who's Counting." This game will introduce dynamic programming, laying the groundwork for understanding the Bellman equation and Q-learning. We will then discuss function approximation with neural networks and the other pivotal developments that led to creation of DQN. We will end with some exciting possibilities that lie ahead in the realm of artificial intelligence.

Title of Talk 2: Mastering Uncertainty: The Art and Science of Stochastic Optimization Algorithms

Synopsis: Stochastic optimization has been paramount across a myriad of applications—from machine learning to operations research. In this talk, we will explore the application of stochastic optimization through the lens of network tomography, a sophisticated technique designed for diagnosing and analysing the performance of computer networks. Our focus will be on the estimation of latent delays in network tomography, which presents a unique set of challenges. We will employ stochastic approximation, a method pioneered by Robbins and Monro, to develop and motivate an online algorithm to solve the above problem. We will also examine the algorithm's convergence through the Ordinary Differential Equation (ODE) method, analyze its convergence rates, and investigate acceleration techniques. We will conclude by discussing the future challenges in stochastic optimization, including scalability, efficiency, and privacy. Title of

Talk 3: Empowering Privacy and Efficiency: A Dive into Federated Learning

Synopsis: Federated learning has emerged as a revolutionary approach in collaborative machine learning. This talk introduces the essentials of federated learning, underscoring its ability to perform model training across diverse devices while maintaining data privacy. We explore the benefits enabled by distributed computation, which not only expedites the learning process but also enhances model robustness through a wealth of heterogeneous data. We will also address the pivotal challenges of security, focusing on designing algorithms resilient to adversarial threats. We will conclude with a discussion on some of the ongoing challenges and future directions in federated learning.

Gugan Chandrashekhar Mallika Thoppe

Qualifications: PhD. (Systems Sciences)

Title: Assistant Professor

Affiliation: Indian Institute of Science (IISc), Bengaluru

LinkedIn:

Twitter/X:

Facebook:

Instagram:

Email:

About the speaker: