Deep Multi-Task and Meta Learning
Deep Multi-Task and Meta-Learning
While deep learning has had great success in supervised and reinforcement learning issues like image classification, speech recognition, and game-playing, these models are specialized to a large extent for the specific task they are trained for. This course will examine how the structure resulting from numerous tasks can be used to learn more efficiently or effectively in a situation where there are multiple tasks to solve. This includes the following:
goal-conditioned reinforcement learning approaches that take advantage of the structure of the specified goal space to learn a large number of tasks much more quickly
meta-learning approaches aimed towards learning efficient learning algorithms capable of fast-learning new tasks
Where a problem necessitates learning a series of activities, curriculum, and lifelong learning, utilizing their shared structure to enable knowledge trajectories
This Deep Multi-Task Course is For
This is a post-graduate course. Students will be able to comprehend and use state-of-the-art multi-task learning and meta-learning algorithms by the end of the course and will be prepared to pursue research on these issues.