December 2017
Intermediate to advanced
536 pages
14h 23m
English
The experiments on the stacked autoencoder suggest that pretraining should be done with low values of learning rate. This ensures better convergence and performance during the fine tuning.
Read now
Unlock full access