Deep Multi-Task and Meta Learning - Autumn 2022

Explore advanced machine learning techniques through this graduate-level course that focuses on leveraging multiple tasks to improve learning efficiency and effectiveness. Master the fundamentals of multi-task learning, where models are designed to handle multiple related problems simultaneously rather than specializing in a single task. Dive deep into transfer learning methodologies that enable knowledge gained from one domain to be applied to related domains with limited data. Study meta-learning algorithms that learn how to learn, enabling rapid adaptation to new tasks with minimal training examples. Examine black-box and optimization-based meta-learning approaches, along with non-parametric few-shot learning techniques. Investigate self-supervised pre-training methods, including contrastive learning and unsupervised pre-training strategies for few-shot scenarios. Analyze advanced topics such as task construction, large-scale meta-optimization, and Bayesian approaches to meta-learning. Understand domain adaptation and generalization techniques that help models perform well across different but related domains. Explore lifelong learning systems that can continuously acquire new knowledge while retaining previously learned information. Gain insights into variational inference and generative models within the context of multi-task learning. Conclude with an examination of current frontiers and open challenges in the field, preparing you to conduct cutting-edge research in multi-task and meta-learning domains.