Unsupervised Embedding of Hierarchical Structure in Euclidean Space

We investigated the effectiveness of embedding data in Euclidean space using VAE-based methods to improve unsupervised hierarchical clustering.

[ Under Review | Paper | Code]

COVID-CT-Dataset:a CT scan dataset about COVID-19

We construct one of the first open-source COVID-19 CT scan dataset and we provide experimental studies to demonstrate this dataset is useful for developing AI-based diagnosis models of COVID-19

[ Under Review | Paper | Code]

Melodic Phrase Segmentation By Deep Neural Networks

We explore and adapt various neural network architectures to see if they can be generalized to work with the symbolic representation of music and produce satisfactory melodic phrase segmentation.

[ ArXiv | Paper | Code]

Sample-Efficient Deep Learning for COVID-19 Diagnosis Based on CT Scans

We develop sample-efficient deep learning methods for COVID-19 diagnosis based on CT scans, addressing the challenge of limited training data in medical imaging during the early stages of the pandemic.

[ medRxiv | Paper]

Graphite: A Graph-Based Extreme Multi-Label Short Text Classifier for Keyphrase Recommendation

We propose Graphite, a graph-based approach for extreme multi-label short text classification applied to keyphrase recommendation, leveraging graph neural networks to capture relationships between keyphrases and text.

[ ArXiv | Paper]

GraphEx: A Graph-Based Extraction Method for Advertiser Keyphrase Recommendation

We present GraphEx, a graph-based extraction method for advertiser keyphrase recommendation that leverages graph structures to improve the quality and relevance of recommended keyphrases for advertising campaigns.

[ IEEE ICDE 2025 | Paper]

BroadGen: A Framework for Generating Effective and Efficient Advertiser Broad Match Keyphrase Recommendations

We introduce BroadGen, a framework designed to generate effective and efficient broad match keyphrase recommendations for advertisers, optimizing both relevance and coverage in advertising campaigns.

[ ArXiv | Paper]