DKM: Differentiable k-Means Clustering Layer for Neural Network Compression.
Minsik ChoKeivan Alizadeh-VahidSaurabh AdyaMohammad RastegariPublished in: ICLR (2022)
Keyphrases
- neural network
- multi layer
- data compression
- single layer
- image compression
- neural network model
- compression algorithm
- artificial neural networks
- compression scheme
- back propagation
- objective function
- network architecture
- compression ratio
- recurrent neural networks
- neural network is trained
- self organizing maps
- loss function
- knn
- k means
- pattern recognition
- application layer
- hopfield neural network
- feed forward neural networks
- feed forward
- fuzzy neural network
- fuzzy logic
- backpropagation neural network
- cluster analysis
- genetic algorithm
- multi layer perceptron
- fuzzy artmap
- multiple layers
- compression rate
- activation function
- neural nets
- radial basis function