Adhd_classification_dbn_extraction

Guide : Prof Sundaram Suresh (NTU- Singapore) Area: Deep learning neural networks for feature extraction in high dimensional neuro imaging data. Tools used: Standard neuro imaging software for preprocessing, a MATLAB deep learning toolbox DeeBNet. I used deep learning algorithms including RBM’s and CNN’s to train on an open source MRI data set and classify unseen fMRI scans as having ADHD or not. I was able to achieve accuracy scores of 64% which is incrementally better than the current start of art(as of 2016). The project was challenging due to the high dimensionality of the input data and the meager number of test samples.