SPGLAD: A Self-paced Learning-Based Crowdsourcing Classification Model

Abstract

Crowdsourcing platforms like Amazon’s Mechanical Turk provide fast and effective solutions of collecting massive datasets for performing tasks in domains such as image classification, information retrieval, etc. Crowdsourcing quality control plays an essential role in such systems. However, existing algorithms are prone to get stuck in a bad local optimum because of ill-defined datasets. To overcome the above drawbacks, we propose a novel self-paced quality control model integrating a priority-based sample-picking strategy. The proposed model ensures the evident samples do better efforts during iterations. We also empirically demonstrate that the proposed self-paced learning strategy promotes common quality control methods.