Hey Arasu, it makes a lot of sense to start using the optimization stuff Wu has implemented in the neural nets package. Currently, we have some kind of duplication there.
So feel free to propose such changes :) Sergey On Sun, Jan 10, 2016 at 6:18 PM, Arasu Arun <[email protected]> wrote: > Dear Developers, > > I wanna implement rmsProp - it was listed as an task for deep learning: > https://github.com/shogun-toolbox/shogun/wiki/GSoC_2015_project_deep_learning > . > I learnt it through Dr. Hinton's course and am familiar with the theory > behind it. > > The core rmsProp functions were written by Wu Lin but the training > algorithms haven't been written in the neural nets package, yet (as far as > I've seen). > > Do I open an issue on Github if I wanna implement this? I'm not sure what > the protocol is. Also, if the same developer (Wu Lin) who wrote the > previous rmsProp functions want to implement other parts, too, I'll of > course back off. :P > > Thank you! > -- > Arasu Arun >
