mryellow

> priority sweeps, how could this be simply implemented with the current code? I assume it involves marking the experience memory with some value that represents good experience vs bad...

Not sure I have it saved here, think it may have been an incomplete draft, and not that interesting otherwise. They were using ReinforceJS, had modified this bit https://github.com/karpathy/reinforcejs/blob/0b9315a69c55f7d66a9d3839a0a90dd067be45db/lib/rl.js#L1091 to...

`google: "this.learnFromTuple(e[0], e[1], e[2], e[3], e[4], e[5])"` On Learning Coordination Among Soccer Agents http://robocup.csu.edu.cn/web/wp-content/uploads/2012/12/data/pdfs/robio12-116.pdf

Hangon, only result, but not it, although I've seen this paper before.... and don't think it passed in the score, but checked it before firing `learnFromTuple`... So that's a wild...

> how it should be done properly Looks like that check pattern you included on the ConvnetJS which dumps it into `window.foo` or `module.exports.foo` is best practice. Not something I've...

I've wrapped it using this pattern, once `Math.tanh` is polyfilled for NodeJS `v0.10.40` everything works as expected. execfile.js ``` var vm = require("vm"); var fs = require("fs"); module.exports = function(path,...

Yeah similar to this: https://github.com/karpathy/convnetjs/blob/master/src/convnet_export.js Think in the end it's probably not really needed, depends on if this lib is about being a block of code in one file, simple...

> is it normally better to let the system "work out" the rules? Personally I've found the simpler the rewards the better. Great example is: http://cs.stanford.edu/people/karpathy/convnetjs/demo/rldemo.html That wall proximity reward...

Could filter it after the fact, but with the right setup should be able to learn the rules. If the filter is captured in the experiences then it's something the...

Morning shower brain kicked in and I'm off on a little bit of a tangent there. The filter and learning in past epsilon stuff is more about when you _must_...