CS Katha Barta | ସଂଗଣକ ବିଜ୍ଞାନ କଥା ବାର୍ତା
Hosted by
Subhankar Mishra's Lab
People -> Rucha Bhalchandra Joshi, Subhankar Mishra
CS Katha Barta 2026
Upcoming Talks
Past Talks
- Dr. Konduri Aditya
Assistant Professor, IISc Bangalore
- Date: March 20 2026, 15:30 hours IST
- Title: Scalable methods for massively parallel flow solvers: application to combustion
-
Abstract
High-fidelity direct numerical simulations (DNS) of turbulent combustion are often
performed to gain fundamental insights into flow–chemistry interactions under conditions
relevant to practical engines. These simulations solve highly nonlinear partial
differential equations, which require massive computations on large supercomputers. A
key challenge is to perform the simulations efficiently and in a scalable manner. In
this talk, we introduce two methods that can significantly improve the scalability of
DNS solvers: First, an asynchronous computing method that significantly minimizes the
data movement costs at extreme scales. Second, a low-dimensional manifold is used to
reduce the chemistry computation costs.
Current state-of-the-art direct numerical simulations are routinely performed on
hundreds of thousands of processing elements (PEs). At an extreme scale, it was observed
that data movement and its synchronization pose a bottleneck to the scalability of
solvers. We introduce an asynchronous computing method that relaxes communication
synchronization at the mathematical level and has shown significant promise in improving
the scalability of PDE solvers. In this method, communication synchronization between
PEs owing to halo exchanges is relaxed, and computations proceed regardless of the
communication status. It was shown that the numerical accuracy of standard schemes, such
as finite differences implemented with relaxed communication synchronization, is
significantly affected. Subsequently, new asynchrony-tolerant schemes have been
developed to compute accurate solutions and demonstrate good scalability. This section
presents an overview of the status of asynchronous computing methods for PDE solvers and
their applicability to exascale simulations.
Identifying low-dimensional manifolds (LDMs) to represent the thermochemical state in
reacting flows is crucial for significantly reducing the computational costs. Widely
used principal component analysis (PCA) achieves this by obtaining an eigenvector basis
for the LDM through eigenvalue decomposition of the data covariance matrix. However,
this may not effectively capture the stiff chemical dynamics when the reaction zones are
localized in space and time. Alternatively, we propose a co-kurtosis PCA (CoK-PCA),
wherein the principal components are obtained from the singular value decomposition
(SVD) of the matricized co-kurtosis tensor. The efficacy of the CoK-PCA-based reduced
manifold was assessed by simulating spontaneous ignition in a homogeneous reactor. The
time-evolved profiles of the PCs and reconstructed thermochemical scalars demonstrate
the robustness of the CoK-PCA-based low-dimensional manifold in accurately capturing the
ignition process. The results of this study show the potential of CoK-PCA-based
manifolds to be implemented in massively parallel reacting flow solvers
- Dr. Adarsh Barik
Assistant Professor, IIT Delhi
- Date: March 13 2026, 15:00 hours IST
- Title: Sequential Decision Making Under Uncertainty: Parameter-free Algorithms for the
Stochastically Extended Adversarial Model
[Youtube]
-
Abstract
I will discuss the problem of sequential decision making under
uncertainty. In particular, I will propose the first parameter-free algorithms and their
regret bounds for the Stochastically Extended Adversarial (SEA) model, a framework that
bridges adversarial and stochastic online convex optimization. Existing approaches for
the SEA model require prior knowledge of problem-specific parameters, such as the
diameter of the domain and the Lipschitz constant of the loss functions, which limits
their practical applicability. Addressing this, we have developed parameter-free methods
by leveraging the Optimistic Online Newton Step (OONS) algorithm to eliminate the need
for these parameters. We first establish a comparator-adaptive algorithm for the
scenario with unknown domain diameter but known Lipschitz constant, and then extend this
to the more general setting where both the diameter and the Lipschitz constant are
unknown, attaining the comparator-and Lipschitz-adaptive algorithm.
Reference: Parameter-free Algorithms for the Stochastically Extended Adversarial
Model.
Authors: Shuche Wang, Adarsh Barik, Peng Zhao, Vincent Y. F. Tan (Appearing in
NeurIPS
2025)
- Dr. Yogesh Simmhan
Associate Professor, IISc Bangalore
- Date: January 19 2026, 10:00 hours IST
- Title: From Accelerated Edge to Emergent Ethics: Optimizing ML Systems and Navigating Agentic
AI’s Impact
-
Abstract
Abstract - The rapid evolution of Machine Learning and Large Language Models is driving
unprecedented demand for accelerated computing at the edge. This talk explores systems
optimizations for Deep Neural Networks and LLMs on heterogeneous edge platforms,
focusing on energy efficiency, latency reduction and scalability. Looking beyond, we
also examine emerging platforms for agentic workflows, leveraging Function-as-a-Service
(FaaS) and Model Context Protocol (MCP), that orchestrate autonomous, goal-driven AI
agents to perform complex tasks across distributed cloud and edge environments. Lastly,
as these agentic systems mature, we explore their ability to supplant human
decision-making, with the consequent ethical and societal challenges. We will discuss
challenges in balancing innovation with responsibility, which is particularly amplified
for the developing world with resource inequities.