A Winner-Take-All Method for Training Sparse Convolutional Autoencoders.
Alireza MakhzaniBrendan J. FreyPublished in: CoRR (2014)
Keyphrases
- high accuracy
- significant improvement
- cost function
- dynamic programming
- detection method
- sparse matrix
- preprocessing
- computational cost
- training algorithm
- training process
- input data
- deep belief networks
- reduced set
- training phase
- rbf neural network
- clustering method
- training samples
- neural network
- training set
- feature extraction
- image processing