ALWANN: Automatic Layer-Wise Approximation of Deep Neural Network Accelerators without Retraining.
Vojtech MrazekZdenek VasícekLukás SekaninaMuhammad Abdullah HanifMuhammad ShafiquePublished in: CoRR (2019)
Keyphrases
- neural network
- multi layer
- single layer
- artificial neural networks
- multiple layers
- fully automatic
- closed form
- training algorithm
- real time
- pattern recognition
- multilayer perceptron
- recurrent neural networks
- approximation algorithms
- fuzzy artmap
- feed forward neural networks
- computing systems
- bp neural network
- associative memory
- feed forward
- error bounds
- radial basis function
- pairwise
- genetic algorithm
- data sets