Heterogeneous Uncertainty Sampling for Supervised Learning

By David D. Lewis, Jason Catlett. 1994

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

Understanding Dataset Difficulty with V-Usable Information (2021, neutral): “Estimating instance difficulty is also evocative of instance selection for active learning (Lewis & Catlett, 1994; Fu et al., 2013; Liu & Motoda, 2002); however these estimates could change as the dataset picks up new instances.”
The Unreasonable Effectiveness of Noisy Data for Fine-Grained Recognition (2015, neutral): “This contrasts with common approaches that focus sampling in uncertain regions [28, 1, 35, 14].”
Survey of the State of the Art in Natural Language Generation: Core tasks, applications and evaluation (2017, methodology): “For example Li et al. (2016) proposed a persona-based model in which the decoder lstm is conditioned on embeddings obtained from tweets pertaining to individual speakers/authors.”
Mining Text Data (2012, methodology): “The work in [80] studies the different kinds of distributional assumptions of words in documents.”
A Survey of Active Learning Algorithms for Supervised Remote Sensing Image Classification (2011, methodology): “The active learning framework has demonstrated its effectiveness when applied to large datasets needing accurate selection of examples [ 18 ].”
Semi-Supervised Active Learning for Sequence Labeling (2009, methodology): “, 1992) or based on uncertainty sampling (Lewis and Catlett, 1994).”
Closing the Loop: Fast, Interactive Semi-Supervised Annotation With Queries on Features and Instances (2011, background): “So far, attempts to “reuse” active learning data have yielded mixed results (Lewis and Catlett, 1994; Baldridge and Osborne, 2004).”
Selective Sampling In Natural Language Learning (1995, methodology): “Figure 4 compares our random selection scheme with a batch selection scheme (similar to that of Lewis and Catlett (1994)).”
Active Learning with Feedback on Features and Instances (2006, background): “pling, the classifier selects the most uncertain instance (Lewis and Catlett, 1994), for a given model ( X ) and a pool of unlabeled instances ( ).”
Information-Theoretic Measures of Dataset Difficulty (2021, background): “Uncertainty sampling, for example, picks the instances that the partially trained model is least certain about (Lewis and Catlett, 1994; Lewis and Gale, 1994; Nigam et al., 2000), which could be interpreted as a measure of difﬁculty.”
ViewAL: Active Learning With Viewpoint Entropy for Semantic Segmentation (2019, unknown): “5 [24, 23].”
SMOTEBoost: Improving Prediction of the Minority Class in Boosting (2003, background): “Researchers have dealt with class imbalance by oversampling the minority class samples with replacement, and/or under-sampling the majority class [19, 20, 21, 22].”
Building Natural Language Interfaces to Web APIs (2017, background): “However, several key distinctions make classic active learning techniques like uncertainty sampling [14] not directly applicable.”
Data Mining for Imbalanced Datasets: An Overview (2005, background): “These applications range from telecommunications management (Ezawa et al., 1996), bioinformatics (Radivojac et al., 2004), text classification (Lewis and Catlett, 1994; Dumais et al., 1998; Mladeni6 and Grobelnik, 1999; Cohen, 1995b), speech recognition (Liu et al., 2004), to detection of oil…”
A Survey on Active Learning and Human-in-the-Loop Deep Learning for Medical Image Analysis (2019, methodology): “Pool based sampling has been applied to several real world tasks, prior to the advent of deep learning (Lewis and Catlett (1994); McCallum and Nigam (1998); Set-tles and Craven (2008); Zhang and Chen (2002); Hauptmann et al. (2006).”
A Survey of Text Classification Algorithms (2012, background): “Some of the earliest implementation of classifiers may be found in [80, 81, 87, 127].”