Affect Estimation in 3D Space Using Multi-Task Active Learning for Regression

Acquisition of labeled training samples for affective computing is usually costly and time-consuming, as affects are intrinsically subjective, subtle uncertain, hence multiple human assessors needed to evaluate each sample. Particularly, affect estimation in the 3D space valence, arousal dominance, assessor has perform evaluations three dimensions, which makes labeling problem even more challenging. Many sophisticated machine learning approaches have been proposed reduce data requirement various other domains, but so far few considered computing. This paper proposes two multi-task active regression approaches, select most beneficial label, by considering primitives simultaneously. Experimental results on VAM corpus demonstrated that our optimal sample selection can result better performance than random several traditional single-task approaches. Thus, they help alleviate computing, i.e., be obtained from fewer queries.