Ci2cv

Congratulations to Xueqian, George, Chaoyang and Simon for getting their papers accepted by CVPR! The CI2CV Lab, led by Dr Simon Lucey , is doing cutting edge research and developing technology in the fields of computer vision and machine learning. We are actively engaged in the following areas: Interpretable deep learning: Deep neural networks hav... e exhibited tremendous success in the past decade in various machine learning tasks, including computer vision, natural language processing, speech analysis, and reinforcement learning. Despite their success, the problem remains how well we understand deep models. Most importantly, as deep neural networks are getting increasingly popular, interpretability has become an exigent factor due to various concerns such as fairness, privacy, reliability, causality, transferability, transparency, plausibility, etc. Our work in this area aims to develop the necessary tools and theoretical frameworks required for a deeper interpretation of deep models and utilize this understanding towards enhanced models. Geometric deep learning: Geometric deep learning is an emerging field in machine learning that aims to learn from data with an underlying geometry, using deep neural networks. Traditional deep neural networks have demonstrated exceptional performance in Euclidean spaces. However, they may be sub-optimal when employed on complex data structures such as graphs and manifolds. Consequently, geometric deep learning focuses on injecting expert domain-specific geometric priors into deep models disguised as feature extraction techniques, optimization constraints, architectural choices, etc., in order to capture these underlying geometrical aspects of data more effectively and efficiently. Neural regularization: Generalization beyond training data is a critical aspect of any machine learning model. However, as the no-free-lunch theorem of machine learning states, there is no absolutely general-purpose model for a training set. In other words, with a particular inductive bias (certain ...Read More