Improving the Performance of Fully Connected Neural Networks by Out-of-Place Matrix Transpose.
Shaohuai ShiPengfei XuXiaowen ChuPublished in: CoRR (2017)
Keyphrases
- fully connected
- neural network
- activation function
- perturbation theory
- conditional random fields
- scale free
- pattern recognition
- multilayer perceptron
- neural network model
- fuzzy logic
- feed forward
- fault diagnosis
- markov random field
- artificial neural networks
- recurrent neural networks
- neural nets
- expert systems
- genetic algorithm
- machine learning