Training Deep Networks without Learning Rates Through Coin Betting.
Francesco OrabonaTatiana TommasiPublished in: NIPS (2017)
Keyphrases
- learning rate
- adaptive learning rate
- backpropagation algorithm
- learning algorithm
- gaussian kernels
- convergence rate
- covering numbers
- feed forward neural networks
- hidden layer
- training examples
- back propagation
- convergence speed
- convergence theorem
- feedforward neural networks
- neural network
- training samples
- training set
- uniform convergence
- delta bar delta
- weight vector
- learning machines
- vc dimension
- multi class