Access to Unlabeled Data can Speed up Prediction Time

By Ruth Urner, Shai Shalev-Shwartz, Shai Ben-David. 2011

55 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Semi-supervised learning, causality, and the conditional cluster assumption (2019, background): “ve a lower risk. In a sense, knowing P(X C)helps decide which parts of the support are most important to get right. This, in turn, can be particularly helpful when dealing with model misspeciﬁcation (Urner et al., 2011). For further discussion of this point, we refer to Peters et al. (2017, Section ”
Knowledge Distillation: A Survey (2020, background): “The knowledge transfer between a fully-supervised teacher model and a student model using the unlabeled data is also introduced for semi-supervised learning (Urner et al., 2011).”
Towards Understanding Knowledge Distillation (2019, background): “The only other theoretical analysis we are aware of is by Urner et al. (2011), who study distillation as a form of semi-supervised learning.”
Domain adaptation–can quantity compensate for quality? (2013, methodology): “The basic idea of our construction is to apply a simple two step procedure, similar to the one suggested in [14] in the context of semi-supervised learning.”
A New PAC-Bayesian Perspective on Domain Adaptation (2015, unknown): “As we shall see in Section 7, once specialized to linear classiﬁers, d T ( ρ ) behaves nicely in this context.”
Joint distribution optimal transportation for domain adaptation (2017, methodology): “To provide some guarantees on our method, we consider an adaptation of the notion probabilistic Lipschitzness introduced in [21, 22] which assumes that two close instances must have the same labels with high probability.”
A survey on semi-supervised learning (2019, background): “This approach can also be applied to inductive learners that have a computationally expensive prediction phase: we can train an inductive semi-supervised learning method on all available data, and pass its predictions for the unlabelled data along with the labelled data to a computationally more eff”