Leveraging Class Hierarchies with Metric-Guided Prototype Learning

2020/07/06 by Garnot, Vivien Sainte Fare, Landrieu, Loic · 1 citation
#Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML)

paper · doi:10.48550/arxiv.2007.03047

Abstract

In many classification tasks, the set of target classes can be organized into a hierarchy. This structure induces a semantic distance between classes, and can be summarised under the form of a cost matrix, which defines a finite metric on the class set. In this paper, we propose to model the hierarchical class structure by integrating this metric in the supervision of a prototypical network. Our method relies on jointly learning a feature-extracting network and a set of class prototypes whose relative arrangement in the embedding space follows an hierarchical metric. We show that this approach allows for a consistent improvement of the error rate weighted by the cost matrix when compared to traditional methods and other prototype-based strategies. Furthermore, when the induced metric contains insight on the data structure, our method improves the overall precision as well. Experiments on four different public datasets - from agricultural time series classification to depth image semantic segmentation - validate our approach.

Cited by

• Feature Identification for Hierarchical Contrastive Learning

Related

• Inspecting class hierarchies in classification-based metric learning models
• Deep Taxonomic Networks for Unsupervised Hierarchical Prototype Discovery
• Fine-Grained Visual Categorization via Multi-stage Metric Learning
• Learning Hierarchical Feature Space Using CLAss-specific Subspace Multiple Kernel -- Metric Learning for Classification
• HIER: Metric Learning Beyond Class Labels via Hierarchical Regularization