DiscLDA: Discriminative Learning for Dimensionality Reduction and Classification

By Simon Lacoste-Julien, Fei Sha, Michael I. Jordan. 2008

442 citations recorded by the source corpus.

Subfields: Representation & Self-Supervised Learning, Probabilistic & Bayesian Methods

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Representation & Self-Supervised Learning

In Probabilistic & Bayesian Methods

Citations in the corpus

Spectral Methods for Supervised Topic Models (2014, unknown): “…regression response for each document, multiclass sLDA [21] that learns with discrete classification responses, discriminative LDA (DiscLDA) [14] that incorporates classification response via discriminative linear transformations on topic mixing vectors, and MedLDA [22] [23] that employs a…”
Visualization of genetic disease-phenotype similarities by multiple maps t-SNE with Laplacian regularization (2014, methodology): “It has been successfully applied to visualize documents [19], breast cancer CADx imaging data [20], and many other domains. t-SNE models the similarity among data points by probability distance rather than Euclidean distance.”
Sparse Additive Generative Models of Text (2011, methodology): “They ﬁnd a comparable result of 69% using support vector machines; alternative latent variable models Discriminative LDA (Lacoste-Julien et al., 2008) and Supervised LDA (Blei & McAuliﬀe, 2007) do worse.”
A Model of Text for Experimentation in the Social Sciences (2016, methodology): “To demonstrate that the covariates provide useful predictive information, we compare the proposed structural topic model (STM) to latent Dirichlet allocation (LDA), the Dirichlet multinomial regression topic model (DMR), and the sparse additive generative text model (SAGE).”
An overview of topic modeling and its current applications in bioinformatics (2016, methodology): “The typical supervised topic models include supervised LDA (sLDA) (Mcauliffe and Blei 2008), the discriminative variation on LDA (discLDA) (Lacoste-Julien et al. 2009), and maximum entropy discrimination LDA (medLDA) (Zhu et al.”
Deep Poisson Factor Modeling (2015, methodology): “Although other DP-based discriminative topic models have been proposed [16, 21], they rely on approximations in order to combine the topic model, usually LDA, with softmax-based classification approaches.”
Structured sparsity through convex optimization (2011, background): “, 2004) • Learning hyperparameters from data • Identifiability and interpretation/evaluation of results • Discriminative tasks (Blei and McAuliffe, 2008; Lacoste-Julien et al., 2008; Mairal et al., 2009c) • Optimization and local minima”
Incorporating Lexical Priors into Topic Models (2012, methodology): “…extension thereof that have focused on improving the semantic coherence of the words in each topic (Grifﬁths et al., 2005; Wallach, 2005; Grifﬁths et al., 2007), the document topic distributions (Blei and Mcauliffe, 2008; Lacoste-Julien et al., 2008) or other aspects (Blei. and Lafferty., 2009).”
Learning Tags from Unsegmented Videos of Multiple Human Actions (2011, background): “Recent LDA variants have included supervised response variables [3], [14], [31].”
Supervised topic models for clinical interpretability (2016, background): “The most natural extension is supervised LDA (McAuliffe and Blei, 2007), though other attempts exist (Zhu et al., 2012; Lacoste-Julien et al., 2009).”
Baselines and Bigrams: Simple, Good Sentiment and Topic Classification (2012, methodology): “Table 4: On 3 20-newsgroup subtasks, we compare to DiscLDA (Lacoste-Julien et al., 2008) and ActiveSVM (Schohn and Cohn, 2000).”
Statistical topic models for multi-label document classification (2011, methodology): “We chose to compare with results from both Lewis et al. (2004) and Eyheramendy et al.”
Latent Dirichlet Allocation Models for Image Classification (2013, unknown): “A second approach is to incorporate a class label variable in the generative model [10], [3], [25], [13], [28], [16].”
Weakly Supervised Joint Sentiment-Topic Detection from Text (2012, methodology): “DiscLDA uses a class-dependent linear transformation to project a K-dimensional (K latent topics) document-topic distribution into a L-dimensional space (L document labels), while Labeled LDA simply defines a one-to-one correspondence between LDA’s latent topics and document labels.”