We can revert back to Ben's initial implementation in the way parameters
are masked during optimisation (i.e. extract the parameters to be optimised
rather than setting the gradient/error of the fixed parameters to 0).
Original issue reported on code.google.com by remi.bar...@gmail.com on 27 Jul 2009 at 12:00
Original issue reported on code.google.com by
remi.bar...@gmail.comon 27 Jul 2009 at 12:00