Weakly Supervised Machine Learning

In modern machine learning methods, predictors are trained using a large amount of fully annotated data. However, due to high labeling costs or privacy concerns, it is often difficult to obtain such full supervision. To solve this problem, we are developing a general framework for training a classifier under weak supervision. Our framework includes various weakly supervised learning problems, and we are exploring a systematic solution based on empirical risk minimization. This framework can also be combined with deep learning, enabling high generalization capability.

RIKEN Center for Advanced Intelligence Project

• du Plessis, M. C., Niu, G., & Sugiyama, M. Analysis of learning from positive and unlabeled data. NeurIPS2014, pp. 703-711, 2014. https://papers.nips.cc/paper/2014/hash/35051070e572e47d2c26c241ab88307f-Abstract.html
• Lu, N., Zhang, T., Niu, G., & Sugiyama, M. Mitigating overfitting in supervised classification from two unlabeled datasets: A consistent risk correction approach. AISTATS2020, pp. 1115-1125, 2020. https://proceedings.mlr.press/v108/lu20c.html
• Sugiyama, M., Bao, H., Ishida, T., Lu, N., Sakai, T., & Niu, G. Machine Learning from Weak Supervision: An Empirical Risk Minimization Approach, The MIT Press, to appear.