Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Yang, Scott Cheng-Hsin, Eaves, Baxter, Schmidt, Michael, Swanson, Ken, Shafto, Patrick
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2403.10424
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866909154825732096
author Yang, Scott Cheng-Hsin
Eaves, Baxter
Schmidt, Michael
Swanson, Ken
Shafto, Patrick
author_facet Yang, Scott Cheng-Hsin
Eaves, Baxter
Schmidt, Michael
Swanson, Ken
Shafto, Patrick
contents Tabular data is common yet typically incomplete, small in volume, and access-restricted due to privacy concerns. Synthetic data generation offers potential solutions. Many metrics exist for evaluating the quality of synthetic tabular data; however, we lack an objective, coherent interpretation of the many metrics. To address this issue, we propose an evaluation framework with a single, mathematical objective that posits that the synthetic data should be drawn from the same distribution as the observed data. Through various structural decomposition of the objective, this framework allows us to reason for the first time the completeness of any set of metrics, as well as unifies existing metrics, including those that stem from fidelity considerations, downstream application, and model-based approaches. Moreover, the framework motivates model-free baselines and a new spectrum of metrics. We evaluate structurally informed synthesizers and synthesizers powered by deep learning. Using our structured framework, we show that synthetic data generators that explicitly represent tabular structure outperform other methods, especially on smaller datasets.
format Preprint
id arxiv_https___arxiv_org_abs_2403_10424
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Structured Evaluation of Synthetic Tabular Data
Yang, Scott Cheng-Hsin
Eaves, Baxter
Schmidt, Michael
Swanson, Ken
Shafto, Patrick
Machine Learning
Tabular data is common yet typically incomplete, small in volume, and access-restricted due to privacy concerns. Synthetic data generation offers potential solutions. Many metrics exist for evaluating the quality of synthetic tabular data; however, we lack an objective, coherent interpretation of the many metrics. To address this issue, we propose an evaluation framework with a single, mathematical objective that posits that the synthetic data should be drawn from the same distribution as the observed data. Through various structural decomposition of the objective, this framework allows us to reason for the first time the completeness of any set of metrics, as well as unifies existing metrics, including those that stem from fidelity considerations, downstream application, and model-based approaches. Moreover, the framework motivates model-free baselines and a new spectrum of metrics. We evaluate structurally informed synthesizers and synthesizers powered by deep learning. Using our structured framework, we show that synthetic data generators that explicitly represent tabular structure outperform other methods, especially on smaller datasets.
title Structured Evaluation of Synthetic Tabular Data
topic Machine Learning
url https://arxiv.org/abs/2403.10424