Roger Milroy has just returned from a year exchange at UC San Diego and is entering his final year of his BSc in Computer Science (AI). His main area of academic focus is in Artificial Intelligence generally and Neural Networks and Deep Learning specifically. He is currently working on making Artificial Intelligence, specifically in the area of Deep Learning, more explainable or transparent. There are already many approaches to this problem, including Sensitivity Analysis and Layerwise Relevance Propagation however these focus solely on interpreting individual decisions given a pretrained model. The project builds on these approaches to provide a more contextural approach taking more into account the training of the model. One of the advantages of this approach is the ability to verify a models training and thus also avoid data poisoning attacks.
