Colloquium - Domain-based Embeddings and Frameworks for Dynamics over Networks

March 23, 2020 - 4:00pm to 5:00pm
Zoom - See emails for details
Bijaya Adhikari
Virginia Tech

Which locations and staff should we monitor in order to detect pathogen outbreaks in hospitals? How do we predict the peak intensity of the influenza incidence in an interpretable fashion? How do we infer the states of all nodes in a critical infrastructure network where failures have occurred? Leveraging domain-based information should make it possible to answer these questions. However, several new challenges arise such as (a) presence of more complex dynamics (b) data sparsity and (c) mismatch between data and the process. Here, I propose to leverage domain-based frameworks, which include novel models and analysis techniques, and domain-based low dimensional representation learning to tackle these challenges for networks and time-series mining tasks. By developing novel frameworks, one can capture the complex dynamics accurately and analyze them more efficiently. At the same time, learning low-dimensional domain-aware embeddings capture domain specific properties more efficiently from sparse data, which is useful for subsequent tasks. Similarly, since domain-aware embeddings capture the model information directly from the data without any modeling assumptions they generalize better to new models. The domain-aware frameworks and embeddings I develop enable many applications in critical domains. For example, our domain-aware framework for C. Difficile outbreak detection has more than 95% accuracy. Similarly, our framework for product recommendation in e-commerce for queries with sparse data results in a 34% improvement over state-of-the-art e-commerce search engine. Additionally, by exploiting domain-aware embeddings, we outperform non-trivial competitors by up to 40% for influenza forecasting.


Bijaya AdhikariBijaya Adhikari is a PhD Candidate in the Department of Computer Science at Virginia Tech. He received his master's degree in computer science from Virginia Tech and a bachelor's degree from Vistula University in Warsaw, Poland. Currently, he works at the intersection of many important domains like epidemiology, critical infrastructure and AI/ML. He builds new algorithms and deep learning architectures incorporating domain knowledge to solve real challenging problems (e.g. his method had the best performance in the HHS1 region in the CDC flu forecasting challenge, his algorithms are in production at Walmart/Amazon etc). He has published at both top data mining and domain-specific venues (SIGKDD, ICDM, WWW, SDM, PLoS Computational Biology - his paper was selected by the editors to be featured on the PLOS Complexity channel as well). He has also won multiple awards from CS@VT (Pratt Fellowship), travel awards etc.