2022/06/24 by Cheng Chen, Cheng, Chen, Hilal Asi +3 · 1 voice · 2 citations
Computer Science · #FOS: Computer and information sciences #FOS: Mathematics #Human-Computer Interaction (cs.HC) #Machine Learning (cs.LG) #Machine Learning and Algorithms #Machine Learning and Data Classification #Music and Audio Processing #Statistics Theory (math.ST)
paper · pdf · doi:10.48550/arxiv.2206.12041
openalex publication_date 2022/06/24 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
The construction of most supervised learning datasets revolves around collecting multiple labels for each instance, then aggregating the labels to form a type of "gold-standard". We question the wisdom of this pipeline by developing a (stylized) theoretical model of this process and analyzing its statistical consequences, showing how access to non-aggregated label information can make training well-calibrated models more feasible than it is with gold-standard labels. The entire story, however, is subtle, and the contrasts between aggregated and fuller label information depend on the particulars of the problem, where estimators that use aggregated information exhibit robust but slower rates of convergence, while estimators that can effectively leverage all labels converge more quickly if they have fidelity to (or can learn) the true labeling process. The theory makes several predictions for real-world datasets, including when non-aggregate labels should improve learning performance, which we test to corroborate the validity of our predictions.