From: Deep neural network-based bottleneck feature and denoising autoencoder-based dereverberation for distant-talking speaker identification
Value
Batch size
128
Learning rate of Bernoulli-Bernoulli RBM
0.02
Learning rate of Gaussian-Bernoulli RBM
0.002
Weight decay
0.0002
Number of iterations in pre-training
50
number of iterations in fine-tuning
100