Student Seminar Series
DATE: Tuesday, February 6th, 2018, 2:00pm
LOCATION: MSB 1147 (Colloquium Room).
SPEAKERS: Franco Liang, Graduate Student, Statistics, UC Davis
TITLE: "Block-based Partitioning for Extreme Multi-label Classification"
ABSTRACT: The objective in extreme multi-label classification is to learn a classifier that tags a data point with the most relevant subset of labels from a large label set. In recent years, various approaches are popular, but it remains challenging to deal with a huge number of features and labels, or to deliver high prediction accuracy when the traditional low-rank assumption on training sets is invalid. In this talk, I introduce a block-based partitioning method as a pretreatment to extreme multi-label classification. Real data experiments on large-scale datasets demonstrate that our method improves prediction time while preserves almost the same level of prediction accuracy as existing methods.
This seminar series is organized by PhD Students Andrew Blandino, Dmitriy Izyumin, and Benjamin Roycraft.