Factorial Hidden Markov Models

By Michael I. Jordanzoubin. 1995

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

Variational Inference: A Review for Statisticians (2016, background): “This article, along with insights from statistical mechanics (Parisi 1988), led to a flurry of variational inference procedures for a wide class of models (Saul, Jaakkola, and Jordan 1996; Jaakkola and Jordan 1996, 1997; Ghahramani and Jordan 1997; Jordan et al. 1999).”
Learning Visual Predictive Models of Physics for Playing Billiards (2015, methodology): “Finally, methods for 3D tracking of Kyriazis et al. (2011); Salzmann & Urtasun (2011) use Physics simulators to constrain the search space during data association.”
An Explanation of In-context Learning as Implicit Bayesian Inference (2021, positive): “We model the entities and properties as independent Markov chains (i.e., a factorial HMM (Ghahramani & Jordan, 1997)), while the emissions depend on both.”
Multitalker Speech Separation With Utterance-Level Permutation Invariant Training of Deep Recurrent Neural Networks (2017, methodology): “…CASA and NMF have led to limited success [4], [5] and the most successful techniques, before the deep learning era, are based on probabilistic models [15]–[17], such as factorial GMM-HMM [18], that model the temporal dynamics and the complex interactions of the target and competing speech signals.”
Permutation invariant training of deep models for speaker-independent multi-talker speech separation (2016, methodology): “00 ©2017 IEEE 241 successful technique before the deep learning era is the model based approach [15,16,17] , such as factorial GMM-HMM [18] , that models the interaction between the target and competing speech signals and their temporal dynamics.”
Approximate inference for the loss-calibrated Bayesian (2011, methodology): “On the other hand, we also allow Q to be a restricted family of tractable distributions, in which case we are using the variational version of EM which only optimizes a lower bound but which has still been applied successfully in the past (Ghahramani and Jordan, 1997; Jordan et al., 1999).”
An Introduction to Variational Methods for Graphical Models (1999, methodology): “ Ghahramani and Jordan (1997) reported results on fitting an FHMM to the Bach chorale data set (Merz & Murphy, 1996).”
Deep Neural Networks for Single-Channel Multi-Talker Speech Recognition (2015, methodology): “…DNN to estimate senone posteriorprobabilities of the louder and softer speakers at each frame, aweighted .nite-state transducer (WFST)-based two-talker decoder to jointly estimate and correlate the speaker and speech,a speaker switching penalty estimated from the energy patternchange in…”
Bayesian Spiking Neurons I: Inference (2008, background): “In general, inference in coupled hidden Markov chain cannot be performed by simply passing messages in pairwise links between the hidden variables (Ghahramani & Jordan, 1997).”
Dynamic Bayesian Networks for Audio-Visual Speech Recognition (2002, methodology): “…improvements to the existing methods for visual feature selection and we focus on two models for isolated word audio-visual speech recognition: the coupled hidden Markov model (CHMM) [3] and the factorial hidden Markov model (FHMM) [4], which are special cases of the dynamic Bayesian networks [5].”
Graphical Models, Exponential Families, and Variational Inference (2008, unknown): “5(b) shows a factorial HMM, in which multiple chains are coupled by their links to a common set of observed variables [89].”
A Pronoun Anaphora Resolution System based on Factorial Hidden Markov Models (2011, background): “These designs allow similar beneﬁts as longer dependencies of higher-order HMMs but avoid the problem of intractability.”
REDD : A Public Data Set for Energy Disaggregation Research (2011, methodology): “Exact posterior inference in the FHMM model is not tractable (we use 4 states per device, and typically around 20 devices per home for a total of 4 20 ≈ 1 × 10 12 different combinations of hidden states), so we use a blocked Gibbs sampling scheme: we fix the hidden states of all but one of the chain”
Expectation propagation for approximate inference in dynamic Bayesian networks (2002, background): “Examples of variational approaches for dynamic Bayesian algorithms are (Ghahramani and Hinton, 1998) for switching linear dynamical systems and (Ghahramani and Jordan, 1997) for factorial hidden Markov models.”
Understanding Video Events: A Survey of Methods for Automatic Interpretation of Semantic Occurrences in Video (2009, background): “For further details regarding HMMs, see [67] and [68].”
Multimodal Machine Learning: A Survey and Taxonomy (2017, methodology): “Some of the earliest approaches to use graphical models for multimodal fusion include generative models such as coupled [155] and factorial hidden Markov models [70] alongside dynamic Bayesian networks [67].”