DSD: Regularizing Deep Neural Networks with Dense-Sparse-Dense Training Flow.
Song HanJeff PoolSharan NarangHuizi MaoShijian TangErich ElsenBryan CatanzaroJohn TranWilliam J. DallyPublished in: CoRR (2016)
Keyphrases
- neural network
- motion field estimation
- training process
- pattern recognition
- dense stereo
- dense optical flow
- flow estimation
- sparse data
- stereo correspondence
- feedforward neural networks
- back propagation
- optical flow
- multi layer perceptron
- recurrent neural networks
- sparse coding
- deep learning
- self organizing maps
- dense reconstruction
- high dimensional
- dense motion estimation