Simple, robust, scalable semi-supervised learning via expectation regularization

By Gideon S. Mann, Andrew McCallum. 2007

184 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

Optimization with Non-Differentiable Constraints with Applications to Fairness, Recall, Churn, and Other Goals (2018, methodology): “Mann and McCallum (2007) and follow-on work (Bellare et al., 2009; Mann and McCallum, 2010) optimized probabilistic models with constraints in order to incorporate side information about the prior priors on class labels, which in the context of a binary classifier is a special case of rate…”
Posterior Regularization for Structured Latent Variable Models (2010, methodology): “By approximating this problem, we recover either the generalized expectation constraints framework of Mann and McCallum (2007), or with a further approximation we recover a special case of the posterior regularization framework presented in Section 2.”
Semi-Supervised Learning via Generalized Maximum Entropy (2010, methodology): “Recently constraint driven semi-supervised approaches have attracted attention, (Bellare et al., 2009; Chang et al., ; Liang et al., 2009; Mann & McCallum, 2007).”
Generalized Expectation Criteria for Semi-Supervised Learning with Weakly Labeled Data (2010, methodology): “Since the initial proposal of GE criteria (under the nameexpectation regularization) in Mann and McCallum (2007), there has been a flurry of recent work on generalizd expectation criteria and related methods which apply expectation constraints for weak learning.”
Generalized Expectation Criteria for Semi-Supervised Learning of Conditional Random Fields (2008, background): “Mann and McCallum (2007) introduce a special case of GE, label regularization , and demonstrate its effectiveness for training maximum entropy classiﬁers.”
Feature Noising for Log-Linear Structured Prediction (2013, background): “Semisupervised dropout has the advantage of only using the predicted label probabilities on the unlabeled data to modulate an L2 regularizer, rather than requiring more heavy-handed modeling of the unlabeled data as in entropy regularization or expectation regularization (Mann and McCallum, 2007).”
Introduction to Semi-Supervised Learning (2009, background): “• expectation regularization [124], which may have deep connections with the class proportion constraints in [36, 33, 89, 210];”
$\propto$SVM for learning with label proportions (2013, methodology): “Related Learning Settings: In semi-supervised learning, Mann & McCallum (2007) and Bellare et al. (2009) used an expectation regularization term to encourage model predictions on the unlabeled data to match the given proportions.”