Siddhartha Shankar Das

My research connects scalable graph learning, AI for system design, cybersecurity, multimodal document intelligence, and evolutionary optimization.

I develop graph sampling and sparsification methods that preserve useful structure while reducing computation, and apply learning methods to hardware/software partitioning, scheduling, and security analysis.

My earlier work in multi-objective and many-objective optimization studies adaptive fuzzy dominance, reference-point guidance, Pareto convergence, solution diversity, and competing objectives in computing clusters.

The highlights below pair short research summaries with figures excerpted from the linked papers. Explore all research themes and current projects.

SGS-GNN: A Supervised Graph Sparsifier for Graph Neural Networks

KDD 2026

Learn which graph edges matter for a downstream prediction task. SGS-GNN trains an edge probability model and a GNN together to sample compact, task-relevant subgraphs, reducing memory requirements while preserving predictive signal.

Paper ↗ · arXiv ↗ · PDF ↗ · Code ↗

Figure 1 · SGS-GNN architecture (preprint) — excerpt from the linked paper.

Scaffold: Support Graph Theory Based Sparsification for Graph Neural Networks

arXiv:2609.31466

An unsupervised sparsification framework that controls the length of rerouted communication paths and the load they place on retained edges. Across 19 benchmarks, sparse supports retain 10–50% of edges while recovering or closely approaching full-graph GNN performance.

arXiv ↗ · PDF ↗ · Code ↗

Figure 2 · Support-graph structure — excerpt from the linked paper.

EDiS: Edge Disjoint Subgraph Sparsification Framework for Graph Neural Networks

arXiv:2610.09059

Decompose a graph once, then reuse cached edge-disjoint subgraphs to compose a different sparse training graph each epoch. EDiS separates structural extraction from epoch-level sampling and evaluates this approach across 19 node classification benchmarks.

arXiv ↗ · PDF ↗

Figure 2 · EDiS pipeline — excerpt from the linked paper.

V2W-BERT: A Framework for Effective Hierarchical Multiclass Classification of Software Vulnerabilities

DSAA 2021 · Best Application Paper · DSAA 2021

A BERT-based framework maps vulnerability descriptions to a hierarchy of software weaknesses. Link prediction and a reconstruction objective support classification even when labeled examples for particular weaknesses are scarce.

Paper ↗ · arXiv ↗ · PDF ↗ · Code ↗

Figure 2 · V2W-BERT architecture — excerpt from the linked paper.

Evolutionary algorithm using adaptive fuzzy dominance and reference point for many-objective optimization

Swarm and Evolutionary Computation 2019

An evolutionary optimization method combines adaptive fuzzy dominance and reference points to balance convergence and diversity when optimizing many competing objectives.

Paper ↗ · Code ↗ · Author manuscript ↗

Figure 2 · Preferred reference points (author manuscript) — excerpt from the linked paper.