Research

My research interests broadly encompass geometric deep learning and graph representation learning. I am driven by questions such as

• Can we design deep learning models that can naturally encode symmetries without explicit guidance?

• How can we introduce non-Euclidean spaces as geometrical inductive biases for learning efficient representations of the data, especially for graphs?

• I am also interested in finding interesting applications for modeling molecular data and biological networks. I am currently working on improving the generalization ability of GNNs by tackling problems like over-squashing and over-smoothing.

Papers

• SoLAR : Surrogate Label Aware Rewiring for Graph-Task Alignment.
  Celia Rubio-Madrigal*, Adarsh Jamadandi*, Rebekka Burkholz. Pre-print, 2024.

• Spectral Pruning Against Over-Squashing and Over-Smoothing.
  Adarsh Jamadandi*, Celia Rubio-Madrigal*, Rebekka Burkholz. NeurIPS, 2024.

• Graph of Thrones: Adversarial Perturbations dismantle Aristocracy in Graphs.
  Adarsh Jamadandi, Uma Mudenagudi. AAAI Student Poster, 2021. Extended Version in Differential Geometry meets Deep Learning - DiffGeo4DL (NeurIPS), 2020.

• Probabilistic Word Embeddings in Kinematic Space.
  Adarsh Jamadandi, Rishabh Tigadoli, Ramesh Tabib, Uma Mudenagudi. International Conference on Pattern Recognition (ICPR), 2020.

• Exemplar-based Underwater Image Enhancement Augmented by Wavelet Corrected Transforms.
  Adarsh Jamadandi, Uma Mudenagudi. Computer Vision and Pattern Recognition (CVPR Workshop, Oral), 2019.

• Learning Hierarchical Representations in Kinematic Space.
  Adarsh Jamadandi, Uma Mudenagudi. Graph Representation Learning Workshop, Neural Information Processing Systems (NeurIPS), 2019.

• PredGAN: a deep multi-scale video prediction framework for detecting anomalies in videos.
  Adarsh Jamadandi, Sunidhi Kotturshettar, Uma Mudenagudi. 11th Indian Conference on Computer Vision, Graphics and Image Processing (ICVGIP ‘18), 2018.