Friday, 13th October - 11am Vlad Niculae : Seminar

Title: Marginalizing over Latent Permutations Efficiently

Abstract:

We present an algorithm for learning latent variable models by marginalizing over the space of discrete structures, in particular for latent permutations in a compact O(n) parametrization based on the permutahedron. We will cover algorithmic contributions in finding k-best permutations, and applications in learning directed acyclic graphs. This presentation is based on "DAG Learning on the Permutahedron" (https://openreview.net/forum?id=m9LCdYgN8-6)

Bio:

Vlad Niculae is an assistant professor in the Language Technology Lab, at the Informatics Institute (IvI), University of Amsterdam. His research is on machine learning for natural language processing, focusing on sparsity, structure, and optimization. He obtained his PhD in 2018 from Cornell University, advised by Claire Cardie, and was previously a postdoc at the Instituto de Telecomunicações in Lisbon. Vlad is an ELLIS Scholar and a member of the Amsterdam ELLIS unit.

 vCal  iCal