Variational Learning for Switching State-Space Models

By Zoubin Ghahramani, Geoffrey E. Hinton. 2000

497 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Signal Processing

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 Signal Processing

Citations in the corpus

Composing graphical models with neural networks for structured representations and fast inference (2016, methodology): “[1] [2] [3] [4] Gaussian mixture model Linear dynamical system Hidden Markov model Switching LDS [6] [2] [5] Mixture of Experts Driven LDS IO-HMM Factorial HMM [7]”
Variable Computation in Recurrent Neural Networks (2016, methodology): “Finally, the Switching State-Space Model of (Ghahramani & Hinton, 2000) combines HMMs and Linear Dynamical Systems: in this model, the HMM is used to switch between LDS parameters, and the experiments show that the HMM learns higher-level, slower dynamics than the LDS.”
Spatio-Temporal Data Mining (2017, methodology): “For example, variational approaches for switching state space models have been used to discover abnormal breathing episodes during sleep (also known as sleep apnea) based on chest volume measurements [Ghahramani and Hinton 2000].”
Variational Tempering (2014, methodology): “It was later applied to latent variable models in Ueda and Nakano [20], who suggest deterministic annealing for maximum likelihood estimation with incomplete data and specific Bayesian models such as latent factor models [7], hidden Markov models [10] and sparse factor models [22].”
A review of novelty detection (2014, background): “The Factorial SKF (FSKF) is a dynamic extension of the SKF, in which a cross-product of multiple factors is used rather than a single variable.”
A Deep Learning Approach to Link Prediction in Dynamic Networks (2014, background): “Linear dynamic systems (e.g. [6]) have more powerful hidden space but they cannot model the complex nonlinear dynamics created by nonlinear properties in data.”
Gene networks inference using dynamic Bayesian networks (2003, background): “Some well-known algorithms for learning Bayesian networks exist (Heckerman, 1995), and new algorithms for learning very complex models have recently been proposed (Ghahramani et al., 2000).”
Multicanonical Stochastic Variational Inference (2014, methodology): “In contrast to earlier work, we temper likelihoods instead of the posterior [7, 10, 21], which we will comment on later in this subsection.”
Pattern Recognition and Machine Learning (2006, unknown): “For example, the switching state space model (Ghahramani and Hinton, 1998) can be viewed as a combination of the hidden Markov model with a set of linear dynamical systems.”
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.”
Rectified Factor Networks (2015, methodology): “In the variational EM, the distribution Q is a variational approximation to the posterior (Ghahramani & Hinton, 1998).”
Bayesian Nonparametric Inference of Switching Dynamic Linear Models (2010, background): “For example, a ballistic missile makes an evasive maneuver; a country experiences a recession, a central bank intervention, or some national or global event; a honeyb e changes from awaggle to a turn right dance.”
An Unsupervised Ensemble Learning Method for Nonlinear Dynamic State-Space Models (2002, background): “See, for instance, Ghahramani and Hinton (2000), where variational methods resembling ensemble learning are applied to a state-space model where a hidden Markov model switches between different linear dynamical models.”