Yesterday I wrote about beginning work on a “Selectionist Neural Network Experiment Editor & Runner“.
I am getting started. But that is a long term effort. Much too long a delay of reinforcement.
How about a shorter term project that will still move the ball forward at the same time? Something a little more fun, and something that will allow more people to play with the neural net sooner?
I have not really looked into how it would all work, but I could get a rudimentary version of the selectionist neural network model (which José Burgos refers to as SelNet) out pretty quickly to work on Swift Playgrounds. That is, NO graphical user interface. Just some example code showing how I run the Burgos & García-Leal 2015 procedure.
I could author a series of “lessons” that show how to use the network library. You could modify the code to your heart’s desire on the iPad. Such lessons might be usable in courses or lab work. You could modify them to fit your situation. You could use them to learn enough to create your own lessons, or Swift Playgrounds apps.
I’m liking the idea as a way of developing the longer term, larger project, incrementally. As I get a bit of new functionality working, I could release it. Very simple stuff at first, and build complexity and capabilities. I think I’d set a goal of a release/month with new stuff.
And maybe someone could play around with integrating SelNet into the game that comes with Swift Playgrounds. The game character should learn, right?
Interested? Leave a reply below.