January 2018
Beginner to intermediate
284 pages
8h 35m
English
Initialize the network according to a certain distribution, such as a normal distribution or a uniform distribution, with very small weights that are close to zero (called symmetry breaking). Different parts of the network will get distinct updates due to this randomness and thus grow diversely. For example,
, where
is a zero mean, unit standard deviation is Gaussian, and
is a small number, for example, 0.01 or 0.001. Or ...
Read now
Unlock full access