Latent Dimensionality Estimation for Probabilistic Canonical Correlation Analysis Using Normalized Maximum Likelihood Code-Length.
Tomohiko NakmauraTomoharu IwataKenji YamanishiPublished in: DSAA (2017)
Keyphrases
- canonical correlation analysis
- normalized maximum likelihood
- code length
- change point detection
- gray code
- partial least squares
- least squares
- spatially adaptive
- high dimensional
- non stationary
- generative model
- latent variables
- dimensionality reduction
- mdl principle
- minimum description length
- probabilistic model
- information theoretic
- channel coding
- channel capacity
- image compression
- image sequences
- multiscale