Theory of Disagreement-Based Active Learning

By Steve Hanneke. 2014

306 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.

In Artificial Intelligence

Citations in the corpus

Scalable Private Learning with PATE (2018, background): “In Machine Learning Theory, active learning (Cohn et al., 1994) has been shown to allow learning from fewer labeled examples than the passive case (see e.g. Hanneke (2014)).”
Deep Batch Active Learning by Diverse, Uncertain Gradient Lower Bounds (2019, background): “Some uncertainty sampling approaches have theoretical guarantees on statistical consistency (Hanneke, 2014; Balcan et al., 2006).”
Practical Evaluation and Optimization of Contextual Bandit Algorithms (2018, methodology): “To achieve this, we use techniques from disagreement-based active learning (Hanneke, 2014; Hsu, 2010).”
The power of localization for efficiently learning linear separators with noise (2013, background): “anneke 2007; Dasgupta et al. 2007; Castro and Nowak 2007; Balcan et al. 2008; Koltchinskii 2010; Beygelzimer et al. 2010; Wang 2011; Dasgupta 2011; Raginsky and Rakhlin 2011; Balcan and Hanneke 2012; Hanneke 2014]. However, despite many efforts, except for very simple noise models (random classiﬁcat”
Efficient and Parsimonious Agnostic Active Learning (2015, methodology): “In order to quantify the extent of gains over passive learning, we measure the hardness of our problem using the disagreement coefficient [Hanneke, 2014], which is defined as 5ω in our statement of the low-noise condition (10) corresponds to 1/κ in the results of Castro and Nowak [2008]. θ = θ(h∗)…”
Making Contextual Decisions with Low Technical Debt (2016, background): “In active learning [11, 26, 27, 44], the algorithm helps select examples to label for a supervised learner.”
Teaching a black-box learner (2019, background): “In many cases, θ has a bound that is independent of the cardinality of X ; see Hanneke (2014) for several examples.”
A Contextual Bandit Bake-off (2018, unknown): “We also experimented with a variant we call active (cid:15) -greedy, that uses notions from disagreement-based active learning (Hanneke, 2014; Hsu, 2010) in order to reduce uniform exploration to only actions that could plausibly be taken by the optimal policy.”