| Abstract: |
Similarity learning has gained much attention recently in a variety of real-world applications, where the goal is to learn a function that maps input patterns to a target space while preserving the semantic distance in the input space. In the domain of brain data analysis, similarity learning is an important problem for multi-subject group-contrasting studies, such as the classification or clustering on brain scans of multiple subjects for neurological disorder diagnosis. In this talk, I will focus on the graph similarity learning problem for brain network data analysis. I will first overview the background and existing research in this field. Then I will introduce an end-to-end deep learning framework called “Higher-order Siamese GCN” for similarity learning on fMRI brain networks. The proposed framework learns the brain network representations via a supervised metric-based approach, which uses Siamese neural networks with two graph convolutional networks as the twin networks. It performs higher-order convolutions by incorporating higher-order proximity in graph convolutional networks to characterize and learn the community structure in brain connectivity networks. Experimental results on four real fMRI datasets will be presented, which indicates the potential use cases of the proposed framework. Finally, I will talk about other related problems and possible future work in this direction. |
| Bio: |
Dr. Guixiang Ma is a Machine Learning Research Scientist at Intel Labs. She received her Ph.D. from the Department of Computer Science at University of Illinois at Chicago (UIC) in May, 2019. During her doctoral study, she worked as a research assistant in the Big Data and Social Computing Lab at UIC, advised by Prof. Philip S. Yu. Her research interest lies in the fields of machine learning, data mining, graph theory, and their applications in brain, social and information networks. In particular, she has authored a number of papers on graph mining for brain data analysis in top-tier conferences. Recently, she has been focusing on building deep graph models, such as graph neural network (GNN) based approaches, for brain network analysis. |