Timezone: »

 
Poster
Cost-Sensitive Self-Training for Optimizing Non-Decomposable Metrics
Harsh Rangwani · shrinivas ramasubramanian · Sho Takemori · Kato Takashi · Yuhei Umeda · Venkatesh Babu R

Wed Nov 30 09:00 AM -- 11:00 AM (PST) @ Hall J #101
in Poster Session 3 »

Self-training based semi-supervised learning algorithms have enabled the learning of highly accurate deep neural networks, using only a fraction of labeled data. However, the majority of work on self-training has focused on the objective of improving accuracy whereas practical machine learning systems can have complex goals (e.g. maximizing the minimum of recall across classes, etc.) that are non-decomposable in nature. In this work, we introduce the Cost-Sensitive Self-Training (CSST) framework which generalizes the self-training-based methods for optimizing non-decomposable metrics. We prove that our framework can better optimize the desired non-decomposable metric utilizing unlabeled data, under similar data distribution assumptions made for the analysis of self-training. Using the proposed CSST framework, we obtain practical self-training methods (for both vision and NLP tasks) for optimizing different non-decomposable metrics using deep neural networks. Our results demonstrate that CSST achieves an improvement over the state-of-the-art in majority of the cases across datasets and objectives.

Keywords: Semi-Supervised Learning · Long-tailed learning · Non-Decomposable Objectives

Author Information

Harsh Rangwani (Indian Institute of Science)
shrinivas ramasubramanian (Indian Institute of Science)
Sho Takemori (Fujitsu Limited)
Kato Takashi (Fujitsu Limited)
Yuhei Umeda (Fujitsu)
Venkatesh Babu R (Indian Institute of Science)

More from the Same Authors