/sci/ - Science & Math

sup /sci/

So I'm currently a grad student in CS working on my master's (since it's free, the company I work for is paying for it). I'm looking to get a jump start on my thesis project, any ideas? My field is AI with a focus on machine learning. I would say my "specialty" is the genetic algorithm.

Just want to probe your brains to see if you guys had any neat ideas for AI-oriented or machine-learning-oriented projects. Bonus points if it addresses a real world problem or issue. An example of what I'm looking for would be something like: utilizing the genetic algorithm to evolve for neural networks that become adaptive to knowledge, etc etc

The project is pretty much free game right now, so throw up any ideas you have.

>>2236814
oic

>>2236801
If I do pursue this, I would be using the Genetic Algorithm; using encoded genotype's as strings of character so that they would be easily mutable.

>>2236797
Natural selection isn't a theory.

It _was_ a theory when proposed by Darwin, but today, it has been solidified into a proven empirical fact.

I might go with some sort of evolution simulation. Artificial or natural selection

Gah! Too many ideas in my head!

>>2236779
For a master's thesis? No.

I'm almost entirely on my own. Of course they're vague and open-ended, I've only just started to conjure up some ideas.

>>2236765
Not a bad idea.

And yes, the natural selection sim would be really cool to see.

>>2236704
Ouch.

So /sci/ doesn't have simulation ideas?

>>2236673
Washington.

>>2236642
Yeah, you would be able to do things like that. It'd be pretty neat, but there is no way it would run on an end-user machine.

The final program will run in my university's super-computing facility (if I decide to go with that particular idea).

Bumping with random things.

>>2236628
/g/ is filled with lonely network admins and IT admins. I've been with /sci/ for a while now and I can say with confidence that there is some collective intelligence here.

/g/ is filled with uneducated, unintelligent people. They're useless.

Anyway, it's time for me to start thinking about my Computer Science master's thesis. Here are my ideas thus far (the first one adopted from /sci/):

- Accurate simulation of the solar system. User is given the ability to change parameters (masses, velocities, etc.) and even add or remove celestial bodies to see their effect.
- Dark matter distribution simulations, utilizing smoothed particle hydrodynamics (SPH)
- Natural selection simulation

All ideas are welcome; must be science related. Space, evolution, etc.