We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

stat.ML

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo

Statistics > Machine Learning

Title: Multivariate Gaussian Variational Inference by Natural Gradient Descent

Abstract: This short note reviews so-called Natural Gradient Descent (NGD) for multivariate Gaussians. The Fisher Information Matrix (FIM) is derived for several different parameterizations of Gaussians. Careful attention is paid to the symmetric nature of the covariance matrix when calculating derivatives. We show that there are some advantages to choosing a parameterization comprising the mean and inverse covariance matrix and provide a simple NGD update that accounts for the symmetric (and sparse) nature of the inverse covariance matrix.
Comments: 11 pages, 0 figures; second version fixed a single typo
Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG); Robotics (cs.RO); Statistics Theory (math.ST)
Cite as: arXiv:2001.10025 [stat.ML]
  (or arXiv:2001.10025v2 [stat.ML] for this version)

Submission history

From: Tim Barfoot [view email]
[v1] Mon, 27 Jan 2020 19:20:03 GMT (20kb,D)
[v2] Mon, 19 Oct 2020 13:54:16 GMT (20kb)

Link back to: arXiv, form interface, contact.