FSPool: Learning Set Representations with Featurewise Sort Pooling

Yan Zhang, Jonathon Hare, Adam Prügel-Bennett
6/6/2019
cs.LGcs.AIstat.ML

Official code: Cyanogenoid/fspool (41 stars, MIT, last commit 2023-10-04).

Abstract

Traditional set prediction models can struggle with simple datasets due to an issue we call the responsibility problem. We introduce a pooling method for sets of feature vectors based on sorting features across elements of the set. This can be used to construct a permutation-equivariant auto-encoder that avoids this responsibility problem. On a toy dataset of polygons and a set version of MNIST, we show that such an auto-encoder produces considerably better reconstructions and representations. Replacing the pooling function in existing set encoders with FSPool improves accuracy and convergence speed on a variety of datasets.

View on arXivView PDF

Code Implementations(1)

Cyanogenoid/fspoolOfficial100%
MIT

[ICLR 2020] FSPool: Learning Set Representations with Featurewise Sort Pooling

417CSS, TeXJun 6, 20193 years agoMIT

Similar Papers

Permutation-Invariant Set Autoencoders with Fixed-Size Embeddings for Multi-Agent Learning

79% match2023

Learning Representations of Sets through Optimized Permutations

78% match1 repo362018

End-to-End Object Detection with Transformers

78% match1 repo15.3k2020

Symmetric Shape-Preserving Autoencoder for Unsupervised Real Scene Point Cloud Completion

78% match1 repo232023

Equivariant vs. Invariant Layers: A Comparison of Backbone and Pooling for Point Cloud Classification

78% match1 repo42023

Towards Human-AI Complementarity with Prediction Sets

78% match1 repo32024

Cite this paper

@article{zhang2019fspool,
  title  = {FSPool: Learning Set Representations with Featurewise Sort Pooling},
  author = {Yan Zhang and Jonathon Hare and Adam Prügel-Bennett},
  year   = {2019},
  eprint = {1906.02795},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1906.02795}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/1906.02795.svg)](https://codeofpaper.com/paper/1906.02795)

Discussion