[Home]History of Fe4r

Robo Home | Changes | Preferences | AllPages

Revision 21 . . May 5, 2006 0:11 EST by GrubbmGait [reverted from vandalism]
Revision 20 . . May 5, 2006 0:00 EST by

Difference (from prior major revision) (no other diffs)

Changed: 1,2c1,48
rebuilt Lockian gagging spook first Cinderella apostate farmers Orphically:adjustments trampled [online poker ] locator decreases reclaiming deport [poker ] renounce nesting?[party poker ] - Tons of interesdting stuff!!!


Great, I want to try it. Where can I download it?


How does it move?

Like Raiko does :D

How does it fire?

It tries to aim with a NeuralNet!!

How does the melee strategy differ from one-on-one strategy?

No meele stategy

It imports the NeuraLib.

| Version | rank | raiting| comment|
| 0.1(RRGC)| 211 | 1584.3 | the basic structure |
| 0.2(RRGC) | --- | --- | its sloooooooow |
| 0.3| 212| 1589.17 | new Movement, improved NeuralNet|

Will be a 1on1 Bot with a GF-Gun and Wavesurfing.
The first version will have a PM-Gun.
The auther will be Krabb :)

It's a misuse of time (and nerves) to implement PMG if your bot will not use it. PM is a nest for bugs, I finally left it buggy in QuickHack 0.6... --lRem

But i think it is easier to programm a PMG than a GFG. --Krabb

Maybe for some people it is. As for me, GFG is both shorter and easier. Unless you're talking of that nano type PMG (which in my version is of codesize around 550)... --lRem

I'd say PM is shorter and GF is easier to debug. -- Kawigi

Itīll be a NeuralNet bot now :D Just because itīs such an intresting thing!--Krabb

NeuralNet you say... Arfticial Neural Network is THE disappointment bringer. I'm not talking only about Robocode. Simply this whole subject seems so interesting, and even exciting, _before_ you get to know it. Afterwards it becomes just another algorithm that you use realy rarely. I've been toying with theese for a few months just to get to know the models, learning algorithms and their application. But I've never stumbled upon real life problem solved by them. Also NeuralTargeting is a bit of missed effort. --lRem

Yes maybe you are right :) I heard so much about Neural Networks and i decided to test it. And Roboking is a good possibility to do it. I donīt expect a top 100 place itīs just a toy. But i think not every kind of Neural Networks has been tested. A Neural Network could be not only used for direct targeting, maybe you can determinate the best gun against a special enemy quicker with a NN or you analyse your GF statistic with a NN. Just some thoughts which came to my mind, when i was reading Articles about NN... But what exactly did you do with Neural Networks? --Krabb

I was teaching the nets to recognize things (usualy simple images, but tried also with texts) and to play simple games. As for robocode - analysing GF stats with NN maybe could help something, but most probably it's not worth the effort/slowdown. Determining the best gun with NN is a good idea, but first you have to have 2 good guns ;] --lRem

I have 2 kinds of guns one against wavesurfing bots and the other against random or static movements. Each of them is not realy good but a combination of this two guns would be better then a single one :) --Krabb

Ouch, version 0.2 is slooow and often generates no score because it "does not perform any actions in a reasonable time". --GrubbmGait

Sorry, it should be faster now! --Krabb

Robo Home | Changes | Preferences | AllPages