Title: Sparse Independent Component Analysis and Any-way Independent Component Analysis for Multimodal Fusion in Imaging Genomics
Dr. Calhoun, Advisor
Dr. Davenport, Chair
Abstract: The objective of the proposed research is to develop new algorithms that leverage sparsity and mutual information across data modalities built upon the independent component analysis framework to improve the performance of current multimodal fusion approaches. To alleviate the signal-background separation difficulties in sources of genetic data, I propose a sparse independent component analysis by enhancing a robust sparsity measure, Hoyer index. Hoyer index is scale-invariant and well suited for independent component analysis frameworks since the scale of decomposed sources is arbitrary. The proposed sparse independent component analysis is further extended into two data modalities as a sparse parallel independent component analysis for applications on imaging genomics, in order to investigate the association between brain imaging and genomics. Moreover, to increase the flexibility and robustness in mining multimodal data, I propose an aNy-way independent component analysis algorithm, which optimizes the entire correlation structure of linked components across any number of modalities via the Gaussian independent vector analysis and simultaneously optimizes independence via separate (parallel) independent component analyses. Lastly, I apply the proposed aNy-way independent component analysis to multimodal brain imaging data to extract multi-aspect brain functional and anatomical relation.