Batch Mode Active Learning for Multi-Label Classification with Informative Label Correlation Mining
Bang Zhang (NICTA)
NICTA SML SEMINARDATE: 2012-02-09
TIME: 11:00:00 - 12:00:00
LOCATION: NICTA - 7 London Circuit
CONTACT: JavaScript must be enabled to display this email address.
ABSTRACT:
The performances of supervised learning techniques on classification problems heavily rely on the quality of their training examples. But the acquisition of high quality training examples requires significant efforts from human annotators. In this talk, I will present a novel multi-label batch model active learning (MLBAL) approach that allows the learning algorithm to actively select a batch of informative example-label pairs from which it learns at each learning iteration, so as to learn accurate classifiers with less annotation efforts. There are a few important issues need to be considered for MLBAL: (1) Unlike binary active learning, the active selection granularity for MLBAL is fined from example to example-label pair. (2) Different labels are usually not independent, and label correlations are helpful to boost the performance of MLBAL. (3) Since the amount of possible label combinations increases exponentially with respect of the number of labels, an efficient mining method is required to discover the informative label correlations for MLBAL. The presented approach takes into account all these issues, and the empirical studies demonstrate its effectiveness.
BIO:
Bang Zhang (Matt) is a fourth year PhD student at University of New South Wales studying computer vision and machine learning. He is also sponsored by National ICT Australia. He is a student member of IEEE and has published 14 papers in the top computer vision and pattern recognition conferences, e.g., ICCV, ICPR, ICIP, ACMMM, WACV, MIR., etc. He won the CiSRA best research paper award at University of New South Wales in 2009. He spent 3 months at Microsoft Research Asia as a research intern working on large scale internet image search in 2010. He received his BSc degree from Sun Yat-Sen University, P.R.China in 2004, and MSc degree in University of New South Wales in 2007.
