[ABE-L] Abel Rodriguez - Data Science Academic Seminar - INSPER

[Hedibert Lopes]

Speaker:

Abel Rodriguez (https://stat.uw.edu/about-us/people/abel-rodriguez)


University:

University of Washington


Title:

Bayesian Record Linkage Models:  New Priors and Data Sources


Abstract:

Record linkage is the task of identifying records in databases that refer
to the same entity in the absence of a unique identifier.  Record linkage
arises in a variety of applications, ranging from public health to defense
and national security. While work on discriminative approaches to record
linkage goes back to the late sixties and early seventies, approaches based
on generative models that allow for full uncertainty quantification have
become popular over the last 10 years.  This presentation will describe two
of our recent contributions to this literature. First, we describe a novel
allelic partition prior for micro clustering that allows for the encoding
of relevant prior information and shows superior empirical performance in
deduplication tasks. Next, we describe how to incorporate relational
information (such as the social network of the individuals to which the
records correspond) to improve the quality of the linkages.  This is joint
work with my former students Juan Sosa and Brenda Betancourt.




October 07, 2021


12pm de São Paulo, Brasil (UTC/GMT -03:00)


Zoom link: https://zoom.us/j/95660781314
-------------- Próxima Parte ----------
Um anexo em HTML foi limpo...
URL: <http://lists.ime.usp.br/pipermail/abe/attachments/20210930/623d6bb5/attachment.htm>