ABOUT

Hi, I'm Rwiddhi, currently a PhD research fellow with Prof. Michael Kampffmeyer at the Department of Physics and Technology, Arctic University of Norway. In the past, I have been a Master's student in Informatics at the University of Lugano (Switzerland), and more recently, a visiting researcher at Fernando De la Torre's Human Sensing Lab at the Robotics Institute, Carnegie Mellon University. My research interests include learning from limited labeled data—semi-supervised learning, unsupervised learning, few-shot learning, and multimodal learning.

WORK

1. Paper at NeurIPS 2024! We propose a new framework for model debiasing using concept graphs that encode co-occurrence-based biases in visual datasets. More details here.
2. Paper accepted at CVPR 2024! We leverage explainability heatmaps to improve worst group robustness to spurious correlations. More details here.
3. Paper accepted at CVPR 2023! We propose hyperspherical embeddings to reduce hubness and improve the state of the art in transductive few shot learning.
4. A Review and Refinement of Surprise Adequacy, published at the DeepTest workshop of the International Conference on Software Engineering (ICSE), 2021.
5. Contributing to the ML Reproducibility Challenge 2020.

OTHER

I write (I want to say) frequently on the literature I'm reading in artificial general intelligence, the philosophy of AI, and the nature of computation here. Some of my deep learning tutorials can be found here. If I'm really feeling like it, I could also probably write on what music I'm listening to (or playing on my Alhambra), or what Pasolini film I'm currently hyping to my friends.