Dynamic Batch Mode Active Learning via L1 Regularization.

Shayok Chakraborty Vineeth N Balasubramanian Sethuraman Panchanathan
Introduction The generation of humongous amounts of data in today's digital world together with the scarcity of human labor asso- ciated with labeling the individual data points, has spurred research in the field of active learning. Active learning algo- rithms automatically select the exemplar data instances from an unlabeled set and thereby reduce human annotation ef- fort in training a classifier. Conventional methods of active learning have focused on the pool-based strategy where the classifier selects one point at a time from an unlabeled pool and is updated ...