ColdGANs: Taming Language GANs with Cautious Sampling Strategies

Training regimes based on Maximum Likelihood Estimation (MLE) suffer from known limitations, often leading to poorly generated text sequences. At the root of these limitations is the mismatch between training and inference, i.e. the so-called exposure bias, exacerbated by considering only the refere...

Ausführliche Beschreibung

Bibliographische Detailangaben

Veröffentlicht in:	arXiv.org 2020-06
Hauptverfasser:	Scialom, Thomas, Paul-Alexis Dray, Lamprier, Sylvain, Piwowarski, Benjamin, Staiano, Jacopo
Format:	Artikel
Sprache:	eng
Schlagworte:	Exploration Learning Maximum likelihood estimation Sampling Training
Online-Zugang:	Volltext
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!