Built a supervised learning framework that reliably identifies clinical conditions for epilepsy patients based on background EEG patterns. Extract features from EEG by computing band power with Fourier transforms.
Achieved above 85% AUC in classifying clinical outcomes. Visualize the findings and presented the poster at ACNS Annual Meeting.
Designed a deep learning architecture to model the link between the galaxies distribution and its underlying dark matter distribution.
Formulated the task to be a semantic segmentation problem and explored the use of convolutional neural networks to perform the mapping.
Outperformed the standard benchmark method of the field while having much more scaling and generalization abilities. Submitted the paper to KDD 2019 as the first author.
