Funding for New Open-set Domain Adaptation Method
ECE/Khoury Professor Yun Raymond Fu received $100K from the Cisco Research University Funding Committee for his project on “Self-debiased Open-set Domain Adaptation.”
Most domain adaptation methods in machine learning focus on addressing the biased data distribution between the source and the target domains. While in the real world, as we usually know little about the target domain, it is unrealistic to assume that all the target samples share the same class space with the source domain samples. We propose a new adversarial-training-based Open-set Domain Adaptation method via self-supervised learning. Such an algorithm will overcome the challenges of data distribution shift between the features of different domains, and the category openness of cross-domain labels.