Flexpoint: An Adaptive Numerical Format for Efficient Training of Deep Neural Networks.
Urs KösterTristan WebbXin WangMarcel NassarArjun K. BansalWilliam ConstableOguz ElibolStewart HallLuke HornofAmir KhosrowshahiCarey KlossRuby J. PaiNaveen RaoPublished in: CoRR (2017)
Keyphrases
- neural network
- training algorithm
- training process
- multimedia
- artificial neural networks
- genetic algorithm
- training examples
- databases
- pattern recognition
- cost effective
- recurrent networks
- feedforward neural networks
- multi layer perceptron
- fuzzy systems
- data structure
- back propagation
- object detection
- deep architectures
- backpropagation algorithm
- decision trees
- feed forward neural networks
- query processing
- training phase
- hidden markov models
- lightweight
- learning rate
- neural nets
- multilayer perceptron
- neural network model
- computationally expensive
- real time