Jeevan Thapa

Jeevan Thapa

Machine Learning Researcher · Ph.D. candidate at RIT

Publications

My research develops adaptive machine learning methods for models, representations, and user behavior that change over time. Citation metrics and the complete record are available on Google Scholar.

Bayesian structure inference framework for continual network depth and width adaptation

International Conference on Machine Learning · 2024

Bayesian Adaptation of Network Depth and Width for Continual Learning

A Bayesian continual-learning framework that dynamically adapts network depth and width to evolving tasks through beta–Bernoulli process priors.

Code
BibTeX
@inproceedings{thapa2024bayesian,
  title     = {Bayesian Adaptation of Network Depth and Width for Continual Learning},
  author    = {Thapa, Jeevan and Li, Rui},
  booktitle = {Proceedings of the 41st International Conference on Machine Learning},
  year      = {2024}
}
Architecture of continual SVD-LoRA and results on continual user modeling

European Conference on Information Retrieval · 2026

Evolving Mixture of Low-Rank Experts for Continual User Modeling

A mixture-of-rank-1-experts architecture that enables parameter-efficient adaptation to sequential recommendation data.

Code
BibTeX
@inproceedings{thapa2026evolving,
  title     = {Evolving Mixture of Low-Rank Experts for Continual User Modeling},
  author    = {Thapa, Jeevan and Zhao, S. and Shindo, K.},
  booktitle = {Proceedings of the European Conference on Information Retrieval},
  year      = {2026}
}

Under review

Cross-Task Representation Alignment for Exemplar-Free Class-Incremental Learning

Jeevan Thapa and Rui Li · Manuscript under review