machine learning - How many hidden layers are there in a Deep Neural Network -
can tell me "usually" how many layers deep neural network have? how deep deep enough?
to knowledge, still difficult specific number of hidden layers. can tell me, examples, how many hidden layers researchers, developers use in deep learning projects?
many thanks.
it totally depends on problem try model. more layers have, harder it's train network (more computation power needed). deeper layer however, more complex problems can solve.
geoffrey hinton wrote in tutorial:
how many lines of code should ai program use , how long should each line be? – silly question.
• deep belief nets give creator lot of freedom.
– how best make use of freedom depends on task.
– enough narrow layers can model distribution on binary vectors (sutskever & hinton, 2007)
• if freedom scares you, stick convex optimization of shallow models inadequate doing artificial intelligence.
from know number of layers not big. here ( imagenet large-scale visual recognition challenge 2014) e.g. google team used net 22 layers.
Comments
Post a Comment