Domain Adaptation with Invariant Representation Learning - What Transformations to Learn?

This course focuses on Domain Adaptation with Invariant Representation Learning, exploring the challenges and solutions in unsupervised domain adaptation using neural networks. The learning outcomes include understanding the limitations of fixed mappings across domains and developing techniques to incorporate domain-specific information for optimal representation. The course teaches skills such as minimizing Jensen-Shannon Divergence, adversarial training, translation, optimization, and contrastive training. The teaching method involves a lecture by the speaker followed by real-world data experiments. The intended audience includes individuals interested in transfer learning, domain adaptation, and machine learning applications in computational biology.