Variational Autoencoders Pursue PCA Directions (by Accident).
Michal RolinekDominik ZietlowGeorg MartiusPublished in: CoRR (2018)
Keyphrases
- principal component analysis
- principal components analysis
- denoising
- face recognition
- dimensionality reduction
- image segmentation
- independent component analysis
- feature extraction
- principal components
- covariance matrix
- linear discriminant analysis
- principle component analysis
- feature space
- optical flow
- methods in computer vision
- traffic accidents
- high dimensional
- lower dimensional
- appearance based object recognition
- optical flow computation
- variational methods
- computer vision
- scale space
- negative matrix factorization
- k means
- feedforward neural networks
- singular value decomposition
- free energy
- high dimensional data
- low dimensional
- face images