Multi-Label Iterated Learning

Home

Discover

Bookmarks

Keep Up With Latest Trending Papers. Computer Science, AI and more.Subscribe

Artificial Intelligence

Multi-Label Iterated Learning

Multi-label Iterated Learning for Image Classification with Label Ambiguity

We propose multi-label iterated learning (mile) to incorporate the inductive biases of multi-label learning from single labels using the framework of iteratedlearning.

Mile is a simple yet effective procedure that builds a multi-labeldescription of the image by propagating binary predictions through successive generations of teacher and student networks with a learning bottleneck.

Experiments show that our approach exhibits systematic benefits on image accuracy as well as real f1 score, which indicates that mile deals better with label ambiguity than the standard training procedure, even when fine-tuningfrom self-supervised weights.

We also show that mile is effective reducing label noise, achieving state-of-the-art performance on real-world large-scale large-scale noisy data such as webvision.

Furthermore, mile improves performance in classincremental settings such as iirc and it is robust to distribution shifts.

Authors

Sai Rajeswar, Pau Rodriguez, Soumye Singhal, David Vazquez, Aaron Courville

◐ Latest Activity

◐ Recommended Follows