Currently the optimization package expects the parameters to be Arrays, that
are dense in their implementation. Allowing Vector from the la package will
allow us to use sparse (or other) parameter implementations.
The most important use case is for stochastic gradient ascent.
Note that this should be implemented to be consistent with speed of the current
implementation.
Original issue reported on code.google.com by sameeersingh on 21 Feb 2012 at 6:55
Original issue reported on code.google.com by
sameeersingh
on 21 Feb 2012 at 6:55