Read the original document by opening this link in a new tab.
Table of Contents
Abstract
Introduction
Permutation Invariance and Equivariance
Structure
Related Results
Deep Sets
Architecture
Other Related Works
Applications and Empirical Results
Summary
Deep Sets is a paper that discusses the problem of designing models for machine learning tasks defined on sets. The paper presents a generic framework to handle scenarios where input and output instances in a machine learning task are sets. It introduces the DeepSets architecture for both invariant and equivariant models. The paper also explores various applications of DeepSets including population statistic estimation, point cloud classification, set expansion, and outlier detection. Experimental results demonstrate the effectiveness of DeepSets in different tasks.