Convolutional Neural Network Feature Extraction Using Covariance Tensor Decomposition (Open Access)
Research and Programming Projects | | Links: Paper | Source | Medium Blog
This work introduces a new algorithm based on multilinear algebra for feature extraction, which later is plugged into a CNN to perform classification. During a single feed-forward step, we generate the kernels for a CNN architecture by computing the covariance tensor of the data and factorizing it by Tucker decomposition.
Tensor Decomposition, Feature extraction, Kernel, Convolutional Neural Networks, Matrix decomposition
PUBLISHED
