note to self - next steps

[LudoTheHUN]

I'm just re-reading some of the material in http://www-ia.hiof.no/~rolando/, specifically http://www-ia.hiof.no/~rolando/01-report.pdf and I'm finally starting to get it (intuitively)... (or at least I think I am).

I'm thinking of revisiting https://github.com/LudoTheHUN/adatx ... as my next project, making it much less naive... Basically, think gradient decent in the space of all functions that are not 'far away' from the current one... (ala 'grains' on page 15 here http://www-ia.hiof.no/~rolando/01-report.pdf) , and a much bigger reliance of the 'function evaluation function's as a way to specify desired behaviour, which is inline with generative testing.

A few sub-projects fall out as prerequisites, it will be very useful to have a notion of a distance metric between values (eg: a number is closer to another number then a map)... same for functions, eg: + and - are close to each other then to identity. If you could place values and function in their respective spaces, you could look in the neighbourhood for likely alternatives / substitutions in the global program search...