Sufficient Representations for Categorical Variables

Jonathan Johannemann, Vitor Hadad, Susan Athey, Stefan Wager
8/26/2019
stat.MLcs.LG

Official code: grf-labs/sufrep (16 stars, GPL-3.0, last commit 2020-04-17).

Abstract

Many learning algorithms require categorical data to be transformed into real vectors before it can be used as input. Often, categorical variables are encoded as one-hot (or dummy) vectors. However, this mode of representation can be wasteful since it adds many low-signal regressors, especially when the number of unique categories is large. In this paper, we investigate simple alternative solutions for universally consistent estimators that rely on lower-dimensional real-valued representations of categorical variables that are "sufficient" in the sense that no predictive information is lost. We then compare preexisting and proposed methods on simulated and observational datasets.

View on arXivView PDF

Code Implementations(1)

grf-labs/sufrepOfficial100%
GPL-3.0

Sufficient Representation for Categorical Variables https://arxiv.org/abs/1908.09874v1

166R, TeXAug 28, 20196 years agoGPL-3.0

Similar Papers

Similarity encoding for learning with dirty categorical variables

83% match1 repo112018

Efficient Representations for High-Cardinality Categorical Variables in Machine Learning

82% match2025

Comparative Study on the Performance of Categorical Variable Encoders in Classification and Regression Tasks

79% match1 repo12024

ResBit: Residual Bit Vector for Categorical Values

78% match2023

Multi-Domain Causal Empirical Bayes Under Linear Mixing

78% match2026

Alternatives of Unsupervised Representations of Variables on the Latent Space

77% match2024

Cite this paper

@article{johannemann2019sufficient,
  title  = {Sufficient Representations for Categorical Variables},
  author = {Jonathan Johannemann and Vitor Hadad and Susan Athey and Stefan Wager},
  year   = {2019},
  eprint = {1908.09874},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1908.09874}
}

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/1908.09874.svg)](https://codeofpaper.com/paper/1908.09874)

Discussion