Publications

Research contributions in Graph Neural Networks, focusing on expressivity, efficiency, and practical applications

'Si'multaneous 'S'patial-'T'emporal Message Passing for Dynamic Graph Representation Learning

Shubhajit Roy, Anirban Dasgupta • 2026

ArXiv
We introduce SiST-GNN, a dynamic graph neural network that integrates spatial and temporal reasoning within a unified message-passing framework. Instead of sequentially chaining spatial and temporal components, SiST-GNN maintains recurrent node states representing history, pairs them with current features across cross-time edges, and applies standard graph convolution simultaneously. Our approach achieves 109–277% improvement over prior methods on link prediction benchmarks in fixed-split settings and 68–194% in live-update regimes.
Dynamic Graphs Temporal Graph Learning Message Passing Link Prediction

FIT-GNN: Faster Inference Time for GNNs that 'FIT' in Memory Using Coarsening

Shubhajit Roy, Hrriday Ruparel, Kishan Ved, Anirban Dasgupta • 2026

TMLR
This paper addresses the computational efficiency challenges in Graph Neural Networks by proposing a novel graph coarsening technique. Our approach significantly reduces inference time and memory while maintaining model performance, making GNNs more practical for real-world applications with large-scale graphs.
Graph Neural Networks Inference Optimization Graph Coarsening Performance

Local Fragments, Global Gains: Subgraph Counting using Graph Neural Networks

Shubhajit Roy, Shrutimoy Das, Binita Maity, Anant Kumar, Anirban Dasgupta • 2025

DiffCoAlg @ NeurIPS 2025 Web&Graph @ WSDM 2026
We propose a novel approach to enhance the expressivity of Graph Neural Networks beyond the traditional Weisfeiler-Leman hierarchy limitations. Our localization-based method improves the ability of GNNs to distinguish between different graph structures while maintaining computational efficiency.
Graph Neural Networks Expressivity Weisfeiler-Leman Localization

Research Interests

My research focuses on advancing the theoretical understanding and practical applications of Graph Neural Networks. I am particularly interested in:

  • Expressivity Enhancement: Developing methods to improve GNN expressivity beyond current theoretical limitations
  • Inference Optimization: Creating efficient algorithms for faster GNN inference in large-scale applications
  • Dynamic Graph Learning: Unified spatial-temporal message passing for dynamic and temporal graph representation
  • Streaming Graph Learning: Addressing challenges in dynamic graph scenarios and catastrophic forgetting
  • Applications: Weather prediction, social network analysis, and knowledge graph construction