Research
I am currently advised by Profs. Alexandre Reiffers-Masson and Abdeldjalil Aïssa-El-Bey at IMTA. I like to theoretically study the limits of machine learning problems under adversarially corrupted data. My interests have thus far led me to the lands of linear algebra, probability, and optimization. My goal is to develop a theory of learnability and approximation algorithms to learn from corrupted data when the underlying model is highly nonlinear, à la neural networks.
Papers
Published
- Vishal Halder, Alexandre Reiffers-Masson, Abdeldjalil Aïssa-El-Bey, Gugan Thoppe. “What Can Be Recovered Under Sparse Adversarial Corruption? Assumption-Free Theory for Linear Measurements.” EUSIPCO 2026. arXiv
Preprints
- Vishal Halder, Alexandre Reiffers-Masson, Abdeldjalil Aïssa-El-Bey, Gugan Thoppe. “Robustness to Sparse Adversarial Corruption in Arbitrary Linear Measurements: Beyond Exact Recovery.” 2026. arXiv
- Nibedita Roy, Vishal Halder, Gugan Thoppe, Alexandre Reiffers-Masson, Mihir Dhanakshirur, Naman, Alexandre Azor. “Tight Convergence Rates for Online Distributed Linear Estimation with Adversarial Measurements.” 2026. arXiv
Background
I have done my Master's in AI from the Indian Institute of Science, wherein I worked with Profs. Rajesh Sundaresan and Gugan Thoppe, the latter of whom I am actively collaborating with. Before this, I have done my Bachelor's from the Indian Institute of Technology Bombay. I grew up in Hyderabad, India.
Contact
vh [at] firstnamelastname [dot] com