BRUNO: A Deep Recurrent Model for Exchangeable Data

Iryna Korshunova, Jonas Degrave, Ferenc Huszár, Yarin Gal, Arthur Gretton +1 more
2/21/2018
stat.ML

Official code: IraKorshunova/bruno (35 stars, MIT, last commit 2020-07-01).

Abstract

We present a novel model architecture which leverages deep learning tools to perform exact Bayesian inference on sets of high dimensional, complex observations. Our model is provably exchangeable, meaning that the joint distribution over observations is invariant under permutation: this property lies at the heart of Bayesian inference. The model does not require variational approximations to train, and new samples can be generated conditional on previous samples, with cost linear in the size of the conditioning set. The advantages of our architecture are demonstrated on learning tasks that require generalisation from short observed sequences while modelling sequence variability, such as conditional image generation, few-shot learning, and anomaly detection.

View on arXivView PDF

Code Implementations(1)

IraKorshunova/brunoOfficial100%
MIT

a deep recurrent model for exchangeable data

359PythonJul 13, 20186 years agoMIT
deep-neural-networksexchangeable-structuresfew-shot-learninggaussian-processesomniglot+2 more

Similar Papers

Variational Feature Compression for Model-Specific Representations

83% match2026

Self-Attention Between Datapoints: Going Beyond Individual Input-Output Pairs in Deep Learning

82% match1 repo4172021

Stochastic Layer-Wise Shuffle for Improving Vision Mamba Training

82% match1 repo212024

Detecting Model Misspecification in Amortized Bayesian Inference with Neural Networks: An Extended Investigation

82% match2024

Ladder Variational Autoencoders

81% match1 repo1042016

Deep Unsupervised Learning using Nonequilibrium Thermodynamics

81% match1 repo5462015

Cite this paper

@article{korshunova2018bruno,
  title  = {BRUNO: A Deep Recurrent Model for Exchangeable Data},
  author = {Iryna Korshunova and Jonas Degrave and Ferenc Huszár and Yarin Gal and Arthur Gretton and Joni Dambre},
  year   = {2018},
  eprint = {1802.07535},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1802.07535}
}

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/1802.07535.svg)](https://codeofpaper.com/paper/1802.07535)

Discussion