Neural Architecture Search and Analysis: Robustness and Sparsity

This project aims to study deep neural networks behaviours against clean and adversarial datasets. In addition, projects will investigate different types of attacks on deep neural networks and study the potentiality of making neural robust against these attacks. We will study the mutational robustness of the networks and study links and connections that influence the generalization ability of the networks. In general, we study the classification and pattern recognition tasks. However, we will not limit the study to mere classification tasks but extend it to multi-task learning and other machine learning tasks.

This project is open for a self-funded, strongly motivated candidate with a strong degree in computer science, mathematics, electronic engineering, and neuroscience. With strong candidates, it is possible to prepare scholarship applications such as Commonwealth and others.