Bayesian exponential family projections for coupled data sources
By Arto Klami, Seppo Virtanen, Samuel Kaski. 2010
33 citations recorded by the source corpus.
Exponential family extensions of principal component analysis (EPCA) have received a considerable amount of attention in recent years, demonstrating the growing need for basic modeling tools that do not assume the squared loss or Gaussian distribution. We extend the EPCA model toolbox by presenting the first exponential family multi-view learning methods of the partial least squares and canonical correlation analysis, based on a unified representation of EPCA as matrix factorization of the natural parameters of exponential family. The models are based on a new family of priors that are generally usable for all such factorizations. We also introduce new inference strategies, and demonstrate how the methods outperform earlier ones when the Gaussianity assumption does not hold.
Subfields: Probabilistic & Bayesian Methods
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Probabilistic & Bayesian Methods
- Samuel Kaski
- Arto Klami
- Seppo Virtanen
- Isabel Valera
- Zoubin Ghahramani
- Maria Lomeli
- Melanie F. Pradier
- Yarin Gal
Citations in the corpus
- Beyond CCA: Moment Matching for Multi-View Models (2016, neutral): “Some extensions of Gaussian CCA were proposed in the literature: exponential family CCA (Virtanen, 2010; Klami et al., 2010) and Bayesian CCA (see, e.”
- Bayesian Canonical correlation analysis (2013, methodology): “Klami et al. (2010) extended Bayesian CCA by generalizing from the Gaussian noise assumption to noise with any distribution in the exponential family.”
- A Survey of Multi-View Representation Learning (2016, unknown): “For example, this survey investigates a great number of deep multiview embedding models which exploit deep visual and linguistic features, while the other surveys focus on the traditional feature extraction methods for multi-view learning.”
- General Latent Feature Models for Heterogeneous Datasets (2017, background): “, 2009), exponential family partial least squares (Klami et al., 2012), and latent variable models with non-conjugate likelihoods (Wang and Blei, 2013; Khan et al.”