Learning from measurements in exponential families

By Percy Liang, Michael I. Jordan, Daniel Klein. 2009

127 citations recorded by the source corpus.

Subfields: Artificial Intelligence, 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 Statistical Learning Theory

Citations in the corpus

Snorkel: Rapid Training Data Creation with Weak Supervision (2017, background): “Other forms include crowdsourced labels [37, 50], rules and heuristics for labeling data [39, 52], and others [29, 30, 30, 46, 51].”
Training Classifiers with Natural Language Explanations (2018, background): “Previous works have relied on identifying relevant parts of the input such as labeling features (Druck et al., 2009; Raghavan et al., 2005; Liang et al., 2009), highlighting rationale phrases in Both cohorts showed signs of optic nerve toxicity due to ethambutol.”
Harnessing Deep Neural Networks with Logic Rules (2016, background): “Though there do exist general frameworks that allow encoding various structured constraints on latent variable models (Ganchev et al., 2010; Zhu et al., 2014; Liang et al., 2009), they either are not directly applicable to the NN case, or could yield inferior performance as in our empirical study.”
Label-Free Supervision of Neural Networks with Physics and Domain Knowledge (2016, background): “Natural language processing has seen many successful applications of constraint learning [10, 2, 5], and the recent work of [12] has provided a fresh perspective on the idea of learning with labeling functions, rather than labels, in the form of Data Programming.”
(Almost) No Label No Cry (2014, unknown): “The latter, including AMM, (conv/alter)-∝SVM, iteratively optimize a cost over labelings (always consistent with label proportions for AMM, not always for (conv/alter)-∝SVM).”
Using Universal Linguistic Knowledge to Guide Grammar Induction (2010, methodology): “As with prior work (Liang et al., 2009b), we assume a degenerate q ( β ) ≡ δ β ∗ ( β ) for tractability reasons, i.e., all mass is concentrated on some single β ∗ .”
Training Complex Models with Multi-Task Weak Supervision (2018, methodology): “…2017a; Hoffmann et al. 2011; Takamatsu, Sato, and Nakagawa 2012), co-training methods (Blum and Mitchell 1998), pattern-based supervision (Gupta and Manning 2014; Zhang et al. 2017a), and feature-annotation techniques (Mann and McCallum 2010; Zaidan and Eisner 2008; Liang, Jordan, and Klein 2009).”
Data-dependent Gaussian Prior Objective for Language Generation (2020, background): “In contrast, for structured probabilistic models, posterior regularization (PR) and related frameworks (Ganchev et al., 2010; Liang et al., 2009; Bellare et al., 2009) provide a general means to impose knowledge constraints during model estimation.”
Active Learning (2012, background): “, 2009), or constraints on the objective function (Druck et al., 2008; Liang et al., 2009; Small et al., 2011).”
Posterior Regularization for Structured Latent Variable Models (2010, methodology): “Liang et al. (2009) use a uniform distribution over f ( X , Y ) ± e , which they call “box noise”.”
Toward interactive training and evaluation (2011, methodology): “Our interactive paradigm and active learning with expectation constraints [7, 15] could also be combined, allowing a mixture of user-initiated and system-initiated interactions.”
Zero-shot Learning of Classifiers from Natural Language Quantification (2018, background): “Prominent among these are the Constraint-driven learning (Chang et al., 2007a), Generalized Expectation (Mann and McCallum, 2010) and Posterior Regularization (Ganchev et al., 2010) and Bayesian Measurements (Liang et al., 2009) frameworks.”
Active Learning by Labeling Features (2009, methodology): “Note that Liang et al. (2009) only use this method in synthetic experiments, and instead use a method similar to total uncertainty for experiments in part-of-speech tagging.”
An Introduction to Conditional Random Fields (2010, background): “In order to deal with this, several different types of regularization terms have been proposed that take the unlabelled data into account, including entropy regularization [39, 45], generalized expectation criteria [69], posterior regularization [32, 38], and measurement-based learning [62].”
Weakly supervised learning via statistical sufficiency (2016, background): “These are for instance self-training and co-training [Blum and Mitchell, 1998; Nigam and Ghani, 2000], multi-view learning [de Sa, 2005] and graph cuts [Blum and Chawla, 2001] for semi-supervised learning, the framework of collective graphical models [Sheldon and Dietterich, 2011; Bernstein and Shel”
Offline Policy Selection under Uncertainty (2020, background): “, generalized expectation criteria [Mann and McCallum, 2010], learning from measurements [Liang et al., 2009], and regularized Bayes [Zhu et al.”
On-the-Job Learning with Bayesian Decision Theory (2015, unknown): “Making active partial observations on structures and has been explored in the measurements framework of [27] and in the distant supervision setting [28].”
Snorkel: rapid training data creation with weak supervision (2019, background): “Other forms include crowdsourced labels [37, 50], rules and heuristics for labeling data [39, 52], and others [29,30,30,46,51].”