1/2 of 1/2 done (Sabbatical Part 6)
Some time ago I promised I'd keep this space up to date on how my return to grad school was doing. Turns out I'm pretty lazy with doing that.
I've offically now completed 2 of 4 credits for my course work. Last term I took 'Genetic Programming' and had the opportunity to learn plenty about how to 'evolve' program code to solve search/classification problems. Essentially you draw a tree where the leafs are values that can be used as arguments for the internal nodes that function as operators. The canonical example is something like the following:
You then perform evolutionary based operators to change the layout/contents of the tree. You select the 'best' new offspring and repeat until you hit your threshold or run out of time. Fun stuff. I did most of the coding using DEAP, a GP package for Python. Result have even been posted to github. Just the code though, no write ups. My final project was a regression to recreate 'Shadow Economy' data. This summer I'm taking a reading course that should carry me until September where I have one course left, and thesis writing and sabbatical proper begins.
Lessons learned:
- 'Just a few more minutes' of code writing in DEAP usually means a couple of hours
- I have workable experience with evolutionary computing now, good news too because my thesis is going to be based on this methodology
- This dogfooding process has really exposed me to the actual research life-cycle. Get papers. Get rich yo.
