Exploiting Kernel Sparsity and Entropy for Interpretable CNN Compression.
Yuchao LiShaohui LinBaochang ZhangJianzhuang LiuDavid S. DoermannYongjian WuFeiyue HuangRongrong JiPublished in: CoRR (2018)
Keyphrases
- kernel function
- compression ratio
- data compression
- high dimensional
- cellular neural networks
- image compression
- kernel methods
- compression algorithm
- support vector
- compression scheme
- information theory
- convolutional neural network
- information entropy
- information theoretic
- sparse representation
- mutual information
- feature space
- lossless compression
- compression rate
- positive definite
- sparsity constraints
- reproducing kernel hilbert space