We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: Sampling Techniques in Bayesian Target Encoding

Abstract: Target encoding is an effective encoding technique of categorical variables and is often used in machine learning systems for processing tabular data sets with mixed numeric and categorical variables. Recently en enhanced version of this encoding technique was proposed by using conjugate Bayesian modeling. This paper presents a further development of Bayesian encoding method by using sampling techniques, which helps in extracting information from intra-category distribution of the target variable, improves generalization and reduces target leakage.
Subjects: Machine Learning (cs.LG); Discrete Mathematics (cs.DM); Machine Learning (stat.ML)
Cite as: arXiv:2006.01317 [cs.LG]
  (or arXiv:2006.01317v2 [cs.LG] for this version)

Submission history

From: Michael Larionov [view email]
[v1] Mon, 1 Jun 2020 23:40:00 GMT (105kb,D)
[v2] Mon, 23 Nov 2020 13:27:07 GMT (87kb,D)

Link back to: arXiv, form interface, contact.