Published in Preprint, 2024
TL;DR: In Extreme Multi-Label Text Classification (XMTC), we train a standalone attention model outside the pipeline, integrate it, and fine-tune, enabling faster learning with less data. Specifically, we transfer transfer L2R models to fine-tune attention in XMTC for ICD coding
Recommended citation: Your Name, You. (2024). "Paper Title Number 1." Journal 1. 1(1). https://arxiv.org/pdf/2410.23066
Published in ISAIM, 2018
TL;DR: We propose a new algorithm for learning buyer behavior parameters efficiently, adaptable to any price constraints set by the seller. This approach offers practical insights for profit-maximizing pricing strategies, inventory management, and promotion planning.
Recommended citation: Saharoy, Debjyoti, and Theja Tulabandhula. "An Online Algorithm for Learning Buyer Behavior under Realistic Pricing Restrictions." arXiv preprint arXiv:1803.01968 (2018). https://isaim2018.cs.ou.edu/papers/ISAIM2018_ML_Saharoy_Tulabandhula.pdf
Published in arXiv, 2016
TL;DR: Polynomial time $(n, d)$-universal set over ${0,1}$, size improvement over Bshouty’s.
Recommended citation: Saharoy, Debjyoti, and Shailesh Vaya. "Universal Sets and Cover-Free Families." arXiv preprint arXiv:1606.01376 (2016). https://arxiv.org/abs/1606.01376
Published in arXiv, 2014
TL;DR: We explore budget-constrained network upgradeable problems on undirected graphs, presenting algorithms for: 1. Finding a maximum weight constrained spanning tree within a budget, improving upon prior results. 2. Solving the longest path problem in a directed acyclic graph (DAG) with limited improvements, achieving efficient algorithms and approximations.
Recommended citation: Saharoy, Debjyoti, and Sandeep Sen. "Approximation algorithms for budget constrained network upgradeable problems." arXiv preprint arXiv:1412.3721 (2014). https://arxiv.org/abs/1412.3721