Enhancing Supervised Learning with Unlabeled Data

By Sally A. Goldman, Yan Zhou. International Conference on Machine Learning, 2000

568 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/ea4feb953b86f6a099d61ffa70d21c59be99f76a

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

Example Selection for Bootstrapping Statistical Parsers (2003, background): “Goldman and Zhou (2000) show that, through careful selection of newly labeled examples, co-training can work even when the classiﬁers’ views do not satisfy the independence assumption.”
Semi-supervised learning by disagreement (2010, unknown): “Unlike [36], tri-training uses the same learning algorithm (e.g., C4.5 decision tree) to generate the three classifiers.”
Democratic co-learning (2004, background): “While democratic co-learning has similarities with statistical co-learning from our earlier work [7], there are major differences.”
Improve Computer-Aided Diagnosis With Machine Learning Techniques Using Undiagnosed Samples (2007, methodology): “Another important factor in co-training is how to produce the learned hypothesis with the refined classifiers, which is sometimes complicated and time-consuming [17].”
Latent-Class Hough Forests for 3D Object Detection and Pose Estimation (2014, methodology): “In the seminal work of Blum & Mitchell [1] it was stated that each classifier should be trained from different views/feature representations of the data, but it was later shown that using two classifiers trained originally on the same view will suffice [12], and this is the variant most similar to o”
Combining Committee-Based Semi-Supervised Learning and Active Learning (2010, methodology): “Li and Zhou[2] proposed an extension to TriTraining, called Co-Forest, in which an initial ensemble of random trees is trained on bootstrap subsamples generated from the given labeled dataset L.”
Analyzing Co-training Style Algorithms (2007, methodology): “For example, rather than using two views, Goldman and Zhou [8] used two diﬀerent supervised learning algorithms, Zhou and Li [16] used two diﬀerent parameter conﬁgurations of the same base learner, etc.”
Semi-Supervised Learning for Natural Language (2005, background): “Instead of assuming two independent views of the data, (Goldman and Zhou, 2000) uses two independent classifiers and one view of the data.”
Introduction to Semi-Supervised Learning (2009, methodology): “There are also many variants of co-training, including the Co-EM algorithm [134], single view [77, 38], single-view multiple-learner Democratic Colearning algorithm [201], Tri-Training [206], Canonical Correlation Analysis [204] and relaxation of the conditional independence assumption [92].”
Exploiting Unlabeled Data in Content-Based Image Retrieval (2004, methodology): “In fact, the semi-supervised learning mechanism embedded in Ssair is somewhat like the algorithm proposed by Goldman and Zhou [5], which uses two diﬀerent supervised learners instead of two attribute sets.”
A survey on semi-supervised learning (2019, methodology): “For example, one can use different hyperparameters for the supervised algorithms (Wang and Zhou 2007; Zhou and Li 2005a), or use different algorithms altogether (Goldman and Zhou 2000; Xu et al. 2012; Zhou and Goldman 2004). Wang and Zhou (2007) provided both theoretical and empirical analyses on wh”
Bootstrapping Coreference Classifiers with Multiple Machine Learning Algorithms (2003, methodology): “Goldman and Zhou (2000) and Steedman et al. (2003b) use two different learning algorithms in lieu of the multiple views required by standard co-training.1 The intuition is that the two learning algorithms can potentially substitute for the two views: different learners have different representation…”
Building text classifiers using positive and unlabeled examples (2003, background): “Finally, our work is related to learning using a small labeled set and a large unlabeled set [2][3][4][5][10] [11][25][26][28][35].”
Co-training by Committee: A New Semi-supervised Learning Framework (2008, methodology): “Zhou and Li [21] present a new Co-Training style SSL method called Tri-Training where the initial 3 classifiers are trained from data sets generated via bootstrap sampling from the original labeled training set.”
A Survey on Multi-view Learning (2013, methodology): “Zhou and Li (2005b) proposed another co-training style semi-supervised algorithm called tri-training, which does not require that the instance space be described with sufficient and redundant views, nor does it put any constraints on the supervised algorithm, as do Goldman and Zhou (2000).”