On 1 Jan 2017, at 19:25, Nir Krakauer <address@hidden> wrote:
Hi Francesco,
I agree with your approach. Matlab's Neural Network Toolbox is fairly widely
used. It would be a good thing to have a drop-in replacement in Octave. I
suggest that once you are happy with a set of your changes, you can post a
notice to the patch tracker and someone can review it for pushing to the Forge
repository.
I don't think this is a viable approach.
The nnet package is currently unmaintained, so probably nobody cares and nor
understands the existing code. It is very unlikely that anyone will ever take
the time to review Francesco's changes.
Francesco has already proven his skills and commitment while working on GSoC, so
I would suggest instead to mark Francesco as the new maintainer of nnet and let
him do the changes directly.