An Average Classification Algorithm

Rooyen, Brendan van; Menon, Aditya Krishna; Williamson, Robert C.

doi:10.48550/arxiv.1506.01520

Cited by 1 publication

(2 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…That is, training on the contaminated data without modifying the loss still minimizes the clean BER. This result has been previously observed for the 0/1 loss [25] and general symmetric losses [42,6]. The above argument gives a simple derivation from Prop.…”

Section: Symmetric Lossessupporting

confidence: 84%

“…1 enables a simple proof of a known result concerning symmetric losses, i.e., losses for which ℓ(t, 1) + ℓ(t, −1) is constant, such as the sigmoid loss. In particular, symmetric losses are immune to label noise under MCMs, meaning the original loss ℓ can be minimized on data drawn from the MCM and still optimize the clean BER [25,42,6].…”

Section: Mutual Contamination Models and Unbiased Lossesmentioning

confidence: 99%

See 1 more Smart Citation

Learning from Label Proportions: A Mutual Contamination Framework

Scott,

Zhang

2020

Preprint

View full text Add to dashboard Cite

Learning from label proportions (LLP) is a weakly supervised setting for classification in which unlabeled training instances are grouped into bags, and each bag is annotated with the proportion of each class occurring in that bag. Prior work on LLP has yet to establish a consistent learning procedure, nor does there exist a theoretically justified, general purpose training criterion. In this work we address these two issues by posing LLP in terms of mutual contamination models (MCMs), which have recently been applied successfully to study various other weak supervision settings. In the process, we establish several novel technical results for MCMs, including unbiased losses and generalization error bounds under non-iid sampling plans. We also point out the limitations of a common experimental setting for LLP, and propose a new one based on our MCM framework.

show abstract

Section: Symmetric Lossessupporting

confidence: 84%

Section: Mutual Contamination Models and Unbiased Lossesmentioning

confidence: 99%