Position Overview

I am working as a Project Associate (Ad Hoc) in the Department of Mechanical Engineering at the Indian Institute of Technology Palakkad, under the mentorship of Prof. Arijit Hazra.

My research focuses on developing deep learning-based methodologies for solving Bayesian inverse problems, with particular emphasis on efficient posterior approximation and uncertainty quantification in computational engineering problems.

Research Focus

The primary objective of this work is to investigate how deep generative and variational learning approaches can be used to approximate complex posterior probability distributions arising in Bayesian inverse problems.

The research focuses on combining probabilistic inference with deep learning to develop computationally efficient alternatives to conventional sampling-based Bayesian methods.

Research Activities

• Developing deep learning frameworks for Bayesian inverse problems.
• Investigating Variational Autoencoders (VAEs) for posterior approximation.
• Comparing variational posterior estimates with Markov Chain Monte Carlo (MCMC) reference solutions.
• Investigating uncertainty quantification and posterior covariance estimation.
• Studying the effect of ill-conditioning on posterior inference and learning behavior.
• Analyzing posterior geometry using the Fisher Information Matrix.
• Investigating information-theoretic regularization strategies for improving posterior approximation.

Methods

• Bayesian Inference
• Variational Autoencoders
• Variational Inference
• Markov Chain Monte Carlo
• Fisher Information Matrix
• Information Geometry
• Uncertainty Quantification
• Deep Generative Models

Technical Tools

• Python
• TensorFlow
• NumPy
• SciPy
• Optuna
• Scientific Computing

Research Direction

This work aims to develop efficient and uncertainty-aware deep learning methodologies for Bayesian inverse problems, with broader applications in computational mechanics, scientific machine learning, and engineering uncertainty quantification.