Less is More: Active Learning with Support Vector Machines

By Greg Schohn, David A. Cohn. 2000

989 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

Query Strategies for Evading Convex-Inducing Classifiers (2010, background): “The ﬁeld of active learning also studies a form of query-based optimization (Schohn and Cohn, 2000).”
Active Bias: Training More Accurate Neural Networks by Emphasizing High Variance Samples (2017, positive): “As reported in the previous studies [42, 6], models can sometimes achieve lower generalization error after being trained with only a subset of actively selected training data.”
Near-Optimal Evasion of Convex-Inducing Classifiers (2010, background): “The field of active learning also studies a form of query based optimization (Schohn and Cohn, 2000).”
Easy Questions First? A Case Study on Curriculum Learning for Question Answering (2016, methodology): “Here, we choose the question q i ∈ Q \ Q 0 whose predicted answer ˆ a i is farthest from the decision boundary: max z ∗ w T φ ( q i , a ∗ , z ∗ , K ) = max ˆ z w T φ ( q, ˆ a, ˆ z , K ) + ∆(ˆ a, a ∗ ) .”
Active Learning as Non-Convex Optimization (2009, methodology): “Heuristic offline active learning algorithms—for example many algorithms for Support Vector Machines (SVMs) that greedily label the point with the smallest absolute margin— are practical and useful in a variety of settings (Schohn and Cohn, 2000; Tong and Koller, 2001).”
Interpretable LSTMs For Whole-Brain Neuroimaging Analyses (2018, background): “(3) Their performance in decoding the cognitive state of an individual from the underlying brain activity does often not improve well with large datasets [8].”
Classifying large data sets using SVMs with hierarchical clusters (2003, methodology): “However, ASVM takes much longer than CB-SVM for very large data sets that do not .t in the memory because it needs to scan the entire data set at each round to select the closest data point, thereby generating too much I/O cost to undergo as many rounds as it needs to get enough training h data.”
Active learning of visual descriptors for grasping using non-parametric smoothed beta distributions (2012, background): “Finally, in Section 7we draw our conclusions and comment on future developments.”
A Survey of Active Learning Algorithms for Supervised Remote Sensing Image Classification (2011, background): “The heuristic taking advantage of this property is called margin sampling (MS) [32], [33].”
Active Learning and Intrinsically Motivated Exploration in Robots : Advances and Challenges (2010, background): “There have been active-extensions to most of the existing learning methods, e.g., logistic regression [12], support vector machines [13], and gaussian processes [14]–[16].”
Active Learning (2012, methodology): “Settles et al. (2008b) introduced a scenario called multiple-instance active learning, illustrated in Figure 7.2 for (a) computer vision and (b) natural language processing tasks. In multiple-instance (MI) learning, instances are grouped into bags (i.e., multi-sets), and it is the bags, rather than ”
Combining active and semi-supervised learning for spoken language understanding (2005, methodology): “Active learning has also been applied to support-vector machines (Schohn and Cohn, 2000; Tong and Koller, 2001), as well as Boosting and Bagging (Abe and Mamitsuka, 1998).”
Interactive deduplication using active learning (2002, methodology): “An intuitive method for measuring uncertainty for separator based classiﬁers like SVMs and regression is to make it inversely proportional to the distance of the instance from the separator [30, 27].”
Deep Batch Active Learning by Diverse, Uncertain Gradient Lower Bounds (2019, background): “In the context of linear classiﬁ-cation, Tong and Koller (2001); Schohn and Cohn (2000); Tur et al. (2005) propose uncertainty sampling methods that query examples that lie closest to the current decision boundary.”
Online Choice of Active Learning Algorithms (2003, methodology): “The first algorithm relies on kernel machines and was recently independently proposed by three research groups (Tong & Koller, 2001; Schohn & Cohn, 2000; Campbell et al., 2000).”
Teachable robots: Understanding human teaching behavior to build more effective robot learners (2008, background): “Input Output”
Comparison of Classification Algorithms and Training Sample Sizes in Urban Land Classification with Landsat Thematic Mapper Imagery (2014, methodology): “In this research, we used a margin-sampling algorithm [41,42], which takes advantage of SVM.”