Online learning of Riemannian hidden Markov models in homogeneous Hadamard spaces

DOI10.1007/978-3-030-80209-7_5MaRDI QIDQ2117823zbMATH OpenOpenAlexFDO

Authors Quinten Tupker, Salem Said, Cyrus Mostajeran

Publication date 22 March 2022

Full work available at URL https://arxiv.org/abs/2102.07771

zbMATH Keywords

hidden Markov model\(k\)-means clusteringGaussian distributionexpectation-maximizationRiemannian manifoldstochastic gradient descent

Mathematics Subject Classification ID

Computational methods in Markov chains (60J22) Signal theory (characterization, reconstruction, filtering, etc.) (94A12) Image processing (compression, reconstruction, etc.) in information and communication theory (94A08)

Abstract: Hidden Markov models with observations in a Euclidean space play an important role in signal and image processing. Previous work extending to models where observations lie in Riemannian manifolds based on the Baum-Welch algorithm suffered from high memory usage and slow speed. Here we present an algorithm that is online, more accurate, and offers dramatic improvements in speed and efficiency.

Recommendations

• Riemannian online algorithms for estimating mixture model parameters
• Online Learning with Hidden Markov Models
• Bayesian online algorithms for learning in discrete hidden Markov models
• On-line learning for the infinite hidden Markov model
• Online expectation maximization based algorithms for inference in hidden Markov models

Cites work

• A Riemannian framework for tensor computing
• Baum's forward-backward algorithm revisited
• Computing the Karcher mean of symmetric positive definite matrices
• Exact equivalences and phase discrepancies between random matrix ensembles
• On-line estimation of hidden Markov model parameters based on the Kullback-Leibler information measure
• Riemannian Gaussian Distributions on the Space of Symmetric Positive Definite Matrices
• The Kähler mean of block-Toeplitz matrices with Toeplitz structured blocks

Cited in

(3)

• Riemannian online algorithms for estimating mixture model parameters
• Differential geometry with extreme eigenvalues in the positive semidefinite cone
• A HMM approach to the estimation of random trajectories on manifolds

Describes a project that uses

Uses Software

• Matrix Means Toolbox

This page was built for publication: Online learning of Riemannian hidden Markov models in homogeneous Hadamard spaces

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2117823)