Random Initialization of Theta
33 views
Skip to first unread message
Antztein
Nov 13, 2011, 7:34:37 PM11/13/11
to NYC Machine Learning Review
Why do we multiply rand(m,n) with 2*init_epsilon and then subtract
init_epsilon?
Isn't there a better way in octave to just get a matrix of random
numbers from the range [-epsilon, epsilon]
What I don't understand is multiplying by twice epsilon and then
subtracting it once.
init_epsilon?
Isn't there a better way in octave to just get a matrix of random
numbers from the range [-epsilon, epsilon]
What I don't understand is multiplying by twice epsilon and then
subtracting it once.
Daniel Lamblin
Nov 14, 2011, 12:54:49 AM11/14/11
to nyc-machine-l...@googlegroups.com
rand give you a positive value in the range of 0 to 1. Scaling it as we do gives the range [-epsilon, epsilon]
Reply all
Reply to author
Forward
0 new messages