Graduate Students Seminar

Speaker 1: Gargi Chaudhuri

Title

Controllability of linear systems

Abstract

Controllability is an essential property of a dynamic system.  It plays an important role in many control problems such as stabilization of unstable systems by feedback or optimal control. It helps to determine whether a system's state can be steered to reach a desired state or not. It focuses on steering from any initial state to the origin in linear systems.  In this presentation we will discuss the controllabillity matrix, gramian and whether controllability depends on time or not for a linear system in continuous time.

Speaker 2: Soumadeep Bhowmick

Title

A Brief Introduction to Conformal Prediction

Abstract

Black-box machine learning models are increasingly deployed in high-risk applications, such as medical diagnostics, where uncertainty quantification is essential to prevent costly errors. Conformal Prediction (also known as Conformal Inference), pioneered by Vovk et al., offers a user-friendly and statistically rigorous framework for constructing uncertainty sets or prediction intervals. A key strength of this approach is its distribution-free validity; it provides explicit, non-asymptotic guarantees on coverage without relying on strong distributional or model assumptions. This flexibility allows conformal prediction to be seamlessly integrated with any pre-trained model, including neural networks, to produce predictions that contain the actual outcome with a user-specified probability. Its ease of implementation and broad applicability make it a valuable tool across diverse fields.