Timezone: »

 
Poster
Learning Deep Parsimonious Representations
Renjie Liao · Alex Schwing · Richard Zemel · Raquel Urtasun

Tue Dec 06 09:00 AM -- 12:30 PM (PST) @ Area 5+6+7+8 #169 #None

In this paper we aim at facilitating generalization for deep networks while supporting interpretability of the learned representations. Towards this goal, we propose a clustering based regularization that encourages parsimonious representations. Our k-means style objective is easy to optimize and flexible supporting various forms of clustering, including sample and spatial clustering as well as co-clustering. We demonstrate the effectiveness of our approach on the tasks of unsupervised learning, classification, fine grained categorization and zero-shot learning.

Author Information

Renjie Liao (UofT)
Alex Schwing (University of Illinois at Urbana-Champaign)
Richard Zemel (Vector Institute/University of Toronto)
Raquel Urtasun (University of Toronto)

More from the Same Authors