Correlated Topic Models

By Yezhou Huang. 2014

926 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.

Citations in the corpus

Assumed Density Filtering Methods for Learning Bayesian Neural Networks (2016, background): “We cope by upper bounding the offending expectation (Blei and Lafferty 2006), E q 2 4ln X”
TopicRNN: A Recurrent Neural Network with Long-Range Semantic Dependency (2016, background): “For example, the correlated topic model (Blei and Lafferty, 2006) uses a log-normal distribution.”
Steering Output Style and Topic in Neural Response Generation (2017, background): “Topic models (Blei et al., 2003; Lafferty and Blei, 2006), on the other hand, are admixtures that capture word co-occurrence statistics by using a much smaller number of topics that can be more freely combined to explain a single document (and this makes it harder to visualize the topics and…”
Structured Stochastic Variational Inference (2014, methodology): “Unlike SVI, it can also address models without tractable local conditionals, such as multilevel logistic regressions (Gelman and Hill, 2007) or the correlated topic model (Blei and Lafferty, 2006).”
DART: a Dataset of Arguments and their Relations on Twitter (2016, unknown): “We tested both Latent Dirichlet Allocation (Blei et al., 2003), and more powerful models such as Correlated Topic Models (Blei and Lafferty, 2006), but the interpretability of the clusters did not improve (Chang et al., 2009).”
Advances in Variational Inference (2017, background): “tion methods which involve inner optimization routines [16], [206], [222] often become prohibitively slow for practical inference tasks.”
WarpLDA: a Cache Efficient O(1) Algorithm for Latent Dirichlet Allocation (2015, background): “, correlation of topics [5, 12], temporal dynamics of topics [6, 4], or relationship to labels [37, 36].”
Topic modelling for qualitative studies (2017, background): “Correlated topic models (CTM) avoid this unrealistic assumption, admitting that some topics are closer to each other and share words with each other; CTMs use logistic normal distribution instead of Dirichlet to model correlations between topics [14].”
Style in the long tail: discovering unique interests with latent variable models in large scale social E-commerce (2014, background): “This can be seen as a simpler way of achieving what the Correlated Topic Model [3] does, but with an easier implementation that is trivial to do in a mapreduce framework.”
Provable Alternating Gradient Descent for Non-negative Matrix Factorization with Strong Correlations (2017, unknown): “For example, when a document contains the topic “machine learning”, it is more likely to contain the topic “computer science” than “geography” (Blei & Lafferty, 2006).”
Variational learning for latent Gaussian model of discrete data (2012, methodology): “In CTM, this required the use of the multinomial logit likelihood which first transforms the realvalued vector z to a probability vector and then use multinomial distribution to model the topics.”
Autoencoding Variational Inference For Topic Models (2017, methodology): “Although we approximate the Dirichlet prior in LDA with a logistic normal, this is not the same idea as a correlated topic model (Blei & Lafferty, 2006), because we use a diagonal covariance matrix.”
Dependent Multinomial Models Made Easy: Stick-Breaking with the Polya-gamma Augmentation (2015, methodology): “In this section we show that MCMC sampling in a correlated topic model based on the stick breaking construction (SB-CTM) can be signiﬁcantly more eﬃcient than sampling in the LN-CTM while maintaining the same integration advantage over EM.”
A Survey on Bayesian Deep Learning (2016, background): “References [7, 113] are proposed to incorporate temporal information, and Reference [6] extends LDA by assuming correlations among topics.”
An analysis of the coherence of descriptors in topic modeling (2015, methodology): “…other prominent methods featuring in the topic coherence experiments that have not been discussed so far include the Correlated Topic Model (CTM) of Blei and Lafferty (2006), which attempts to directly model correlation between the latent topics themselves, and the Pólya Urn method proposed by…”
MedLDA: maximum margin supervised topic models (2012, methodology): “In fact, the likelihood component of MedLDA can be any other form of generative topic model, such as correlated topic models (Blei and Lafferty , 2005), or latent space Markov random fields, such as exponential family harmoniums (Welling et al., 2004; Xing et al., 2005; Chen et al., 2010).”
TensorFlow Distributions (2017, background): “For example, theMultinomial-LogisticNormal distribution [7] fails to meet this contract.”