Learning Representations of Sets through Optimized Permutations

Yan Zhang, Jonathon Hare, Adam Prügel-Bennett
12/10/2018
cs.LGcs.CVstat.ML

Official code: Cyanogenoid/perm-optim (36 stars, last commit 2019-05-01).

Abstract

Representations of sets are challenging to learn because operations on sets should be permutation-invariant. To this end, we propose a Permutation-Optimisation module that learns how to permute a set end-to-end. The permuted set can be further processed to learn a permutation-invariant representation of that set, avoiding a bottleneck in traditional set models. We demonstrate our model's ability to learn permutations and set representations with either explicit or implicit supervision on four datasets, on which we achieve state-of-the-art results: number sorting, image mosaics, classification from image mosaics, and visual question answering.

View on arXivView PDF

Code Implementations(1)

Cyanogenoid/perm-optimOfficial100%

[ICLR 2019] Learning Representations of Sets through Optimized Permutations

362Shell, PythonSep 23, 20187 years ago

Similar Papers

Set2Seq Transformer: Temporal and Positional-Aware Set Representations for Sequential Multiple-Instance Learning

79% match2024

On permutation-invariant neural networks

79% match2024

FSPool: Learning Set Representations with Featurewise Sort Pooling

78% match1 repo412019

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

78% match2023

Scalable Permutation-Aware Modeling for Temporal Set Prediction

76% match2025

Permutation Equivariance of Transformers and its Applications

76% match1 repo192023

Cite this paper

@article{zhang2018learning,
  title  = {Learning Representations of Sets through Optimized Permutations},
  author = {Yan Zhang and Jonathon Hare and Adam Prügel-Bennett},
  year   = {2018},
  eprint = {1812.03928},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1812.03928}
}

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/1812.03928.svg)](https://codeofpaper.com/paper/1812.03928)

Discussion