Fast differentiable sorting and ranking

Jul 12, 2020



Sorting is an elementary building block of modern software. In machine learning and statistics, it is commonly used in robust statistics, order statistics and ranking metrics. However, sorting is a piecewise linear function and as a result includes many kinks at which it is non-differentiable. More problematic, the ranking operator is a piecewise constant function, meaning that its derivatives are null or undefined. While numerous works have proposed differentiable proxies to sorting and ranking, they do not achieve the O(n log n) time complexity one could expect from a sorting or ranking operation. In this paper, we propose the first differentiable sorting and ranking operators with O(n log n) time and O(n) space complexity. Our proposal in addition enjoys exact computation and differentiation. We achieve this feat by casting differentiable sorting and ranking as projections onto a permutahedron, the convex hull of permutations, and using a reduction to isotonic optimization. Empirically, we confirm that our approach is an order of magnitude faster than existing approaches. We also showcase two novel applications: differentiable Spearman's rank coefficient and differentiable least trimmed squares.



About ICML 2020

The International Conference on Machine Learning (ICML) is the premier gathering of professionals dedicated to the advancement of the branch of artificial intelligence known as machine learning. ICML is globally renowned for presenting and publishing cutting-edge research on all aspects of machine learning used in closely related areas like artificial intelligence, statistics and data science, as well as important application areas such as machine vision, computational biology, speech recognition, and robotics. ICML is one of the fastest growing artificial intelligence conferences in the world. Participants at ICML span a wide range of backgrounds, from academic and industrial researchers, to entrepreneurs and engineers, to graduate students and postdocs.

Store presentation

Should this presentation be stored for 1000 years?

How do we store presentations

Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%


Recommended Videos

Presentations on similar topic, category or speaker