Electronic memes for 2009-04-23 April 23, 2009 at 6:00 am
- Not true, but ubiquitous computing will change delivery http://tinyurl.com/covzx8 #
- Eight characters per minute, with your mind. It will only get faster. http://tinyurl.com/co3msa #
- But can it heat up pizza in 0.000001s? http://home.slac.stanford.edu/pressreleases/2009/20090421.htm #
Powered by Twitter Tools.
Robert – is there a tutorial or info anywhere about using the neuroevolutionary solver other than the general stuff on your download site? i got it running on Eclispe but where to start cuz I’d rather have to reverse engineer!
thx a ton
Wow! You are the first person to ever ask! Ok, so are you looking to develop your own experiment or work with the existing experiments? I have been planning to write up a tutorial for the system, but just haven’t finished it. I can update the Neuroevolutionary Solver page tonight with tutorial information so you can hopefully accomplish what you need.
Thx for the quick response! I would like to do both; develop my own experiments and work with the existing ones all for the purpose of learning & creating new algorithms/idea’s. I need to be able to simulate them and this is the perfect testbed. Then, after a simulation is working the way I want, I would like to test it further via programming the Roomba Create. FUN!
I would love to read your tutorial and will do so as your done!
Thx a lot!
Awesome! Ok, I will put up tutorial information about how to create experiments using the built in learning algorithms, (the easy part) and then how you would go about creating your own algorithm (the hard part). In fact, putting in your own algorithm is quite hard and dependent on the algorithm so I can really only offer guidelines for that, but I will do my best.
Why is it hard to add user algorithms? It looks easy if you were to use the Simbad api’s(override PerformBehavior()),etc.
Right? Hmm…
Ah, yeah that overrides the control for an agent. You can definitely place algorithms in there, and leverage NEAT or SDNEAT to build neural network controllers. What I was referring to was creating your own neural learning algorithm like NEAT or SDNEAT, embedding that in the system is quite complex.
I would like to be able to create my own neural learning algor. such as NEAT or SDNEAT if I get some inspiration after I play with the built in ones. I’ll follow your guidelines for this and hopefully it won’t be too complex!