A zero-shot learning boosting framework via concept-constrained clustering

Qin Yue, Junbiao Cui, Liang Bai, Jianqing Liang, Jiye Liang

Abstract: Highlights•A general framework is proposed for boosting base zero-shot learning models.•Zero-shot learning is transformed into a co-training problem.•Concept-constrained clustering is proposed to estimate the data distribution.•An alternative optimization strategy is designed to achieve mutual guidance.