Active Learning

By Burr Settles. 2012

695 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Science Applications, Developmental and Educational Psychology, Statistical Learning Theory

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

In Computer Science Applications

In Developmental and Educational Psychology

In Statistical Learning Theory

Citations in the corpus

Hypersim: A Photorealistic Synthetic Dataset for Holistic Indoor Scene Understanding (2020, background): “compute node). We will include the cost of rendering each image in our online code repository, so the marginal value and marginal cost of each image can be analyzed jointly in downstream applications [59]. Annotation Costs In total, annotating our scenes took 369 hours, leading to an average annotat”
Making deep neural networks right for the right scientific reasons by interacting with their explanations (2020, neutral): “Here we cover the basics of pool-based active learning, and refer the reader to two excellent surveys [38, 39] for more details.”
A survey on semi-supervised learning (2019, background): “Since labelling data is generally costly, the challenge lies in the selection of unlabelled data points whose labels would bemost informative (Settles 2012).”
Perceptual Annotation: Measuring Human Vision to Improve Computer Vision (2014, background): “Standard approaches for incorporating humans into the machine learning process [1] have focused on individual human annotators for labeling difficult or ambiguous training data for continual improvement of a class model.”
Evaluation in artificial intelligence: from task-oriented to ability-oriented measurement (2017, background): “Other related areas are domain adaptation (Jiang, 2008), multi-task learning (Caruana, 1997; Thrun, 1996), lifelong learning (Thrun and Pratt, 2012), incremental learning (Khreich et al, 2012) and active learning (Settles, 2012).”
Active Fairness in Algorithmic Decision Making (2018, background): “ from the active learning setting in which the system actively queries training examples that optimize learning e.g. by balancing exploration and exploitation or maximizing the expected model change (Settles 2012). We foresee future work studying synergies in systems that attain fairness by actively”
Active Learning of Uniformly Accurate Inter-atomic Potentials for Materials Simulation (2018, background): “[21] Burr Settles.”
Toward systematic review automation: a practical guide to using machine learning tools in research synthesis (2019, background): “This is a variant of active learning (AL) [29].”
Snorkel DryBell: A Case Study in Deploying Weak Supervision at Industrial Scale (2018, background): “Active learning [36] methods select data points for human annotators to label.”
Pitfalls of In-Domain Uncertainty Estimation and Ensembling in Deep Learning (2020, methodology): “ensembling techniques are used to improve it. These settings include, but are not limited to, regression, language modeling (Gal, 2016), image segmentation (Gustafsson et al., 2019), active learning (Settles, 2012) and reinforcement learning (Buckman et al., 2018; Chua et al., 2018). We focus on in-”
The Importance of Calibration for Estimating Proportions from Annotations (2018, background): “…this work also relates to the problem of active learning , where the goal is to interactively choose instances to be labeled, in a way that maximizes accuracy while minimizing the total cost of annotation (Beygelzimer et al., 2009; Baldridge and Osborne, 2004; Rai et al., 2010; Settles, 2012).”