EPQuant: A Graph Neural Network compression approach based on product quantization.
Linyong HuangZhe ZhangZhaoyang DuShuangchen LiHongzhong ZhengYuan XieNianxiong TanPublished in: Neurocomputing (2022)
Keyphrases
- neural network
- quantization noise
- data compression
- efficient compression
- random walk
- compression ratio
- uniform quantization
- lossy image compression
- graph representation
- graph model
- neural network is trained
- bipartite graph
- image compression
- pattern recognition
- quantization error
- graph structure
- graph matching
- artificial neural networks
- directed graph
- back propagation
- computational complexity
- prediction model
- life cycle
- graph databases
- fuzzy logic
- product design
- huffman coding
- bit rate
- bp neural network
- adaptive quantization
- transform coding
- compression scheme
- graph theory
- recurrent neural networks
- self organizing maps
- genetic algorithm