ZCA Whitening
Description
ZCA Whitening is an image preprocessing method that leads to a transformation of data such that the covariance matrix is the identity matrix, leading to decorrelated features.
Image Source: Alex Krizhevsky
Papers Using This Method
Isotropy Matters: Soft-ZCA Whitening of Embeddings for Semantic Code Search2024-11-26Covariance-corrected Whitening Alleviates Network Degeneration on Imbalanced Classification2024-08-30Whitening Consistently Improves Self-Supervised Learning2024-08-14Implicit ZCA Whitening Effects of Linear Autoencoders for Recommendation2023-08-15Orthogonal SVD Covariance Conditioning and Latent Disentanglement2022-12-11Improving Covariance Conditioning of the SVD Meta-layer by Orthogonality2022-07-05Finite Versus Infinite Neural Networks: an Empirical Study2020-07-31Iterative Normalization: Beyond Standardization towards Efficient Whitening2019-04-06Decorrelated Batch Normalization2018-04-23Active Convolution: Learning the Shape of Convolution for Image Classification2017-03-27