Learning to Represent and Predict Sets with Deep Neural Networks

TL;DR

This paper introduces new deep learning techniques for representing and predicting sets, addressing challenges related to their unordered nature and internal relations, leading to improved modeling and prediction performance.

Contribution

It develops novel neural network methods for better set representations and predictions, overcoming previous bottlenecks and discontinuity issues.

Findings

Improved set representation models show consistent performance gains.

Set prediction models handle unordered data more effectively.

Techniques outperform existing approaches on various experiments.

Abstract

In this thesis, we develop various techniques for working with sets in machine learning. Each input or output is not an image or a sequence, but a set: an unordered collection of multiple objects, each object described by a feature vector. Their unordered nature makes them suitable for modeling a wide variety of data, ranging from objects in images to point clouds to graphs. Deep learning has recently shown great success on other types of structured data, so we aim to build the necessary structures for sets into deep neural networks. The first focus of this thesis is the learning of better set representations (sets as input). Existing approaches have bottlenecks that prevent them from properly modeling relations between objects within the set. To address this issue, we develop a variety of techniques for different scenarios and show that alleviating the bottleneck leads to consistent improvements across many experiments. The second focus of this thesis is the prediction of sets (sets as output). Current approaches do not take the unordered nature of sets into account properly. We determine that this results in a problem that causes discontinuity issues with many set prediction tasks and prevents them from learning some extremely simple datasets. To avoid this problem, we develop two models that properly take the structure of sets into account. Various experiments show that our set prediction techniques can significantly benefit over existing approaches.