Distribution Re-weighting and Voting Paradoxes
We explore a specific type of distribution shift called domain expertise, in which training is
limited to a subset of all possible labels. This setting is common among specialized human …
limited to a subset of all possible labels. This setting is common among specialized human …
[PDF][PDF] Combining Binary Classifiers Leads to Nontransitive Paradoxes
We study collections of binary classifiers trained on different pairs of target classes and
observe cases of nontransitive disagreement. Noting a similar phenomenon in voting theory …
observe cases of nontransitive disagreement. Noting a similar phenomenon in voting theory …