IRC Log Viewer » #firebreath » 2010-10-02

IRC Nick Time (GMT-7) Message
taxilian 12:10 wow, lotta people here for a Saturday
WarGloom 12:10 :)
taxilian 12:10 I find that I can't quite resist the urge to put occasional things in my papers that my professor couldn't help but comment on, just to see if they actually read the papers
someday I'll get a professor without a sense of humor and probably get docked for it
WarGloom 12:10 lol
nitrogenycs 12:10 how often do they miss these comments?
taxilian 12:10 it's surprisingly rare, actually
but I keep them relevant enough (just funny) that I haven't been docked yet
nitrogenycs 12:10 :)
taxilian 12:10 this one is a genetic algorithm, and I ran into a combination of bugs that resulted a population that never changed because over time they all became the same; not quite good enough for the solution, but closeish
so after explaining the issue in my writeup, I added: "Ideas on how the results of this early experiment could be used to better educate socioeconomically disadvantaged students in rural areas of southern Kentucky will be saved for another paper as it is outside of the scope of this document. "
nitrogenycs 12:10 lol, I bet you get docked if you ever get a professor from the rural areas of southern Kentucky
taxilian 12:10 the trick is to know your audience =]
nitrogenycs 12:10 :) ya
taxilian 12:10 I had one professor tell me "Be careful writing this way; some professors would dock you for this. However, after reading a stack of papers, I found the levity refreshing"
nitrogenycs 12:10 yeah, always depends on the person
taxilian 12:10 I consider that an accurate representation of what you'll find in the industry, though; you always have to know your audience when writing something, or you'll not be well received
in some cases not having anything amusing in the paper will result in nobody wanting to read it
nitrogenycs 12:10 It's been ages since I toyed with GAs. I started looking into them for AI, but it's so hard to get useful stuff out of them once the tasks get a bit complexer
taxilian 12:10 other cases you'll not be taken seriously
yeah; kinda a fun idea, though
nitrogenycs 12:10 that's true
yes, the idea is interesting. Maybe it just needs more hardware power.
taxilian 12:10 well, some complex problems it probably works well for
seems like it would be one interesting method of finding routes on a map, for example
or circuit board layouts
nitrogenycs 13:10 yes, that always depends on the definiton of "complex"
taxilian 13:10 true
nitrogenycs 13:10 yes, I know it's used for finding routes on the map, travelling salesman problem
taxilian 13:10 it does well with simple but abstract problems, I guess
nitrogenycs 13:10 it gives very close to optimal results in a tiny fraction of the time needed for the optimal solution
yes, problems you can solve locally and don't depend a lot on context
taxilian 13:10 when there is reasonably good heuristic function
nitrogenycs 13:10 yes
taxilian 13:10 this is a "code breaking" algorithm; of course, it's very fake, since it requires that you know what you're looking for :-P
but it's a sound test of the principles
nitrogenycs 13:10 I've seen an example a long time ago where they used a GA to make a "something" which moves as fast as possible. It started with a simple brick and then the "creature" would randomly grow legs and stuff. The fitness was calculated based on the time it took it to go some distance.
the outcomes looked quite interesting
taxilian 13:10 lol
that does sound interesting
nitrogenycs 13:10 this is a similar video, but it lacks the cool randomness of the character :
taxilian 13:10 HUH
bleh, dump caps lock
nitrogenycs 13:10 this one here is pretty cool:
taxilian 13:10 I might have to show these to my professor
I think he'd enjoy them
ever seen this?
nitrogenycs 13:10 no, just finished the first level :)
it's cool with different characters
i've seen the real mario somewhere
but it plays awkward on german keyboards, because Z and Y is switched compared to the US keyboards