Applying Piecewise Linear Approximation for DNN Non-Linear Activation Functions to Bfloat16 MACs.
Seok Young KimChang Hyun KimSeon Wook KimPublished in: ICEIC (2021)
Keyphrases
- support vector machine
- activation function
- radial basis function
- multi layer perceptron
- neural network
- neural architecture
- training process
- hidden layer
- feed forward
- neural nets
- complex valued
- multilayer perceptron
- feed forward neural networks
- artificial neural networks
- back propagation
- hidden nodes
- learning rate
- knn
- artificial intelligence
- network architecture
- training set
- feature space
- particle swarm optimization
- training phase
- pattern recognition