214:
have the potential to adapt to changing conditions and learn new behaviors as they carry out their tasks. The online evolutionary process is implemented according to a physically distributed island model. Each robot optimizes an internal population of candidate solutions (intra-island variation), and two or more robots exchange candidate solutions when they meet (inter-island migration). In this way, each robot is potentially self-sufficient and the evolutionary process capitalizes on the exchange of controllers between multiple robots for faster synthesis of effective controllers.
114:
genes by the use of a global innovation number which increases as new genes are added. When adding a new gene the global innovation number is incremented and assigned to that gene. Thus the higher the number the more recently the gene was added. For a particular generation if an identical mutation occurs in more than one genome they are both given the same number, beyond that however the mutation number will remain unchanged indefinitely.
61:. It alters both the weighting parameters and structures of networks, attempting to find a balance between the fitness of evolved solutions and their diversity. It is based on applying three key techniques: tracking genes with history markers to allow crossover among topologies, applying speciation (the evolution of species) to preserve innovations, and developing topologies incrementally from simple initial structures ("complexifying").
151:
population. When a network's timer expires, its current fitness measure is examined to see whether it falls near the bottom of the population, and if so, it is discarded and replaced by a new network bred from two high-fitness parents. A timer is set for the new network and it is placed in the population to participate in the ongoing evaluations.
213:
odNEAT is an online and decentralized version of NEAT designed for multi-robot systems. odNEAT is executed onboard robots themselves during task execution to continuously optimize the parameters and the topology of the artificial neural network-based controllers. In this way, robots executing odNEAT
81:
Traditionally, a neural network topology is chosen by a human experimenter, and effective connection weight values are learned through a training procedure. This yields a situation whereby a trial and error process may be necessary in order to determine an appropriate topology. NEAT is an example of
154:
The first application of rtNEAT is a video game called Neuro-Evolving
Robotic Operatives, or NERO. In the first phase of the game, individual players deploy robots in a 'sandbox' and train them to some desired tactical doctrine. Once a collection of robots has been trained, a second phase of play
113:
and is represented by , if this genome is crossed with an identical genome (in terms of functionality) but ordered crossover will yield children that are missing information ( or ), in fact 1/3 of the information has been lost in this example. NEAT solves this problem by tracking the history of
93:-like feed-forward network of only input neurons and output neurons. As evolution progresses through discrete steps, the complexity of the network's topology may grow, either by inserting a new neuron into a connection path, or by creating a new connection between (formerly unconnected) neurons.
150:
In 2003, Stanley devised an extension to NEAT that allows evolution to occur in real time rather than through the iteration of generations as used by most genetic algorithms. The basic idea is to put the population under constant evaluation with a "lifetime" timer on each individual in the
85:
In order to encode the network into a phenotype for the GA, NEAT uses a direct encoding scheme which means every connection and neuron is explicitly represented. This is in contrast to indirect encoding schemes which define rules that allow the network to be constructed without explicitly
163:
An extension of Ken
Stanley's NEAT, developed by Colin Green, adds periodic pruning of the network topologies of candidate solutions during the evolution process. This addition addressed concern that unbounded automated growth would generate unnecessary structure.
264:
Erin J. Hastings, Ratan K. Guha, and
Kenneth O. Stanley (2009). "Automatic Content Generation in the Galactic Arms Race Video Game ". IEEE Transactions on Computational Intelligence and AI in Games, volume 4, number 1, pages 245-263, New York: IEEE Press,
254:
Matthew E. Taylor, Shimon
Whiteson, and Peter Stone (2006). "Comparing Evolutionary and Temporal Difference Methods in a Reinforcement Learning Domain". GECCO 2006: Proceedings of the Genetic and Evolutionary Computation
197:, a space-shooter game in which unique particle system weapons are evolved based on player usage statistics. Each particle system weapon in the game is controlled by an evolved
275:
Silva, Fernando; Urbano, Paulo; Correia, Luís; Christensen, Anders Lyhne (2015-09-15). "odNEAT: An
Algorithm for Decentralised Online Evolution of Robotic Controllers".
82:
a topology and weight evolving artificial neural network (TWEANN) which attempts to simultaneously learn weight values and an appropriate topology for a neural network.
155:
allows players to pit their robots in a battle against robots trained by some other player, to see how well their training regimens prepared their robots for battle.
223:
101:
The competing conventions problem arises when there is more than one way of representing information in a phenotype. For example, if a genome contains neurons
69:
On simple control tasks, the NEAT algorithm often arrives at effective networks more quickly than other contemporary neuro-evolutionary techniques and
245:
Kenneth O. Stanley and Risto
Miikkulainen (2002). "Evolving Neural Networks Through Augmenting Topologies". Evolutionary Computation 10 (2): 99-127
193:
Content-Generating NEAT (cgNEAT) evolves custom video game content based on user preferences. The first video game to implement cgNEAT is
802:
137:
interpreter. This implementation of NEAT is considered the conventional basic starting point for implementations of the NEAT algorithm.
780:
812:
678:
669:
58:
807:
492:
642:
582:
759:- A 3D version of Picbreeder, where you interactively evolve 3D objects that are encoded with CPPNs and evolved with NEAT.
688:
651:
629:
616:
607:
134:
722:
817:
660:
638:
573:
464:
Proceedings of the
Artificial Intelligence and Interactive Digital Entertainment Conference (AIIDE 2005) Demo Papers
126:
706:
42:
591:
500:
IAAI 2007: Proceedings of the
Nineteenth Annual Innovative Applications of Artificial Intelligence Conference
456:
786:
349:
70:
762:
783:- A visual tutorial series on NEAT, including solving the classic pole balancing problem using NEAT in R
434:
17:
513:
354:
503:
390:
375:
310:
194:
54:
455:
Kenneth O. Stanley; Ryan
Cornelius; Risto Miikkulainen; Thomas D’Silva & Aliza Gold (2005).
775:
367:
302:
38:
408:
359:
334:
292:
284:
117:
These innovation numbers allow NEAT to match up genes which can be crossed with each other.
475:"Comparing Evolutionary and Temporal Difference Methods in a Reinforcement Learning Domain"
726:
595:
447:
50:
474:
517:
521:
202:
46:
796:
427:"Phased Searching with NEAT: Alternating Between Complexification And Simplification"
426:
314:
86:
representing every connection and neuron, allowing for more compact representation.
548:
379:
561:
540:
612:
181:
is specialized to evolve large scale structures. It was originally based on the
398:
Proceedings of the
Genetic and Evolutionary Computation Conference (GECCO-2002)
363:
767:
693:
482:
GECCO 2006: Proceedings of the Genetic and Evolutionary Computation Conference
90:
391:"Efficient Reinforcement Learning Through Evolving Neural Network Topologies"
718:
625:
416:
Proceedings of the 2003 IEEE Congress on Evolutionary Computation (CEC-2003)
178:
173:
130:
674:
371:
306:
557:
288:
665:
493:"Stochastic Optimization for Collision Selection in High Energy Physics"
771:
508:
297:
599:
544:
409:"Evolving Adaptive Neural Networks with and without Adaptive Synapses"
684:
587:
407:
Kenneth O. Stanley; Bobby D. Bryant & Risto Miikkulainen (2003).
702:
578:
697:
603:
552:
634:
125:
The original implementation by Ken Stanley is published under the
753:- Online, collaborative art generated by CPPNs evolved with NEAT.
565:
198:
182:
787:"Artificial intelligence learns Mario level in just 34 attempts
774:
video demonstrating an implementation of NEAT learning to play
647:
569:
473:
Matthew E. Taylor; Shimon Whiteson & Peter Stone (2006).
756:
744:
732:
750:
621:
335:"Evolving Neural Networks Through Augmenting Topologies"
738:
27:
Genetic algorithm for making artificial neural networks
656:
389:
Kenneth O. Stanley & Risto Miikkulainen (2002).
333:
Kenneth O. Stanley & Risto Miikkulainen (2002).
733:"Evolutionary Complexity Research Group at UCF"
491:Shimon Whiteson & Daniel Whiteson (2007).
201:, similarly to the evolution technique in the
241:
239:
224:Evolutionary acquisition of neural topologies
8:
457:"Real-Time Learning in the NERO Video Game"
185:theory and is an active field of research.
768:MarI/O - Machine Learning for Video Games
507:
353:
296:
739:NERO: Neuro-Evolving Robotic Operatives
235:
31:NeuroEvolution of Augmenting Topologies
735:- Ken Stanley's current research group
443:
432:
18:NeuroEvolution of Augmented Topologies
7:
763:BEACON Blog: What is neuroevolution?
41:(GA) for the generation of evolving
747:- an example application of cgNEAT
741:- an example application of rtNEAT
624:(not an exact implementation) and
25:
789:NEAT explained via MarI/O program
59:The University of Texas at Austin
598: (archived 2021-05-15)) and
89:The NEAT approach begins with a
1:
681:(not an exact implementation)
729: (archived 2023-12-05))
834:
803:Artificial neural networks
364:10.1162/106365602320169811
171:
43:artificial neural networks
707:Go (programming language)
205:interactive art program.
813:Evolutionary computation
342:Evolutionary Computation
277:Evolutionary Computation
49:technique) developed by
808:Evolutionary algorithms
745:GAR: Galactic Arms Race
442:Cite journal requires
71:reinforcement learning
129:. It integrates with
97:Competing conventions
425:Colin Green (2004).
289:10.1162/evco_a_00141
518:2006hep.ex....7012W
818:Genetic algorithms
757:"EndlessForms.com"
195:Galactic Arms Race
55:Risto Miikkulainen
776:Super Mario World
57:in 2002 while at
39:genetic algorithm
16:(Redirected from
825:
781:"GekkoQuant.com"
751:"PicBreeder.org"
528:
526:
520:. Archived from
511:
497:
485:
479:
467:
461:
451:
445:
440:
438:
430:
419:
413:
401:
395:
383:
357:
339:
319:
318:
300:
272:
266:
262:
256:
252:
246:
243:
21:
833:
832:
828:
827:
826:
824:
823:
822:
793:
792:
727:Wayback Machine
715:
596:Wayback Machine
536:
534:Implementations
531:
524:
495:
490:
477:
472:
459:
454:
441:
431:
424:
411:
406:
393:
388:
355:10.1.1.638.3910
337:
332:
328:
323:
322:
274:
273:
269:
263:
259:
253:
249:
244:
237:
232:
220:
211:
191:
176:
170:
161:
148:
143:
123:
99:
79:
67:
51:Kenneth Stanley
28:
23:
22:
15:
12:
11:
5:
831:
829:
821:
820:
815:
810:
805:
795:
794:
791:
790:
784:
778:
765:
760:
754:
748:
742:
736:
730:
714:
713:External links
711:
710:
709:
700:
691:
682:
672:
663:
654:
645:
632:
619:
610:
585:
576:
555:
535:
532:
530:
529:
527:on 2015-10-28.
509:hep-ex/0607012
487:
486:
469:
468:
452:
444:|journal=
421:
420:
403:
402:
385:
384:
329:
327:
324:
321:
320:
283:(3): 421–449.
267:
257:
247:
234:
233:
231:
228:
227:
226:
219:
216:
210:
207:
203:NEAT Particles
190:
187:
172:Main article:
169:
166:
160:
159:Phased pruning
157:
147:
144:
142:
139:
122:
121:Implementation
119:
98:
95:
78:
75:
66:
63:
47:neuroevolution
26:
24:
14:
13:
10:
9:
6:
4:
3:
2:
830:
819:
816:
814:
811:
809:
806:
804:
801:
800:
798:
788:
785:
782:
779:
777:
773:
769:
766:
764:
761:
758:
755:
752:
749:
746:
743:
740:
737:
734:
731:
728:
724:
723:NEAT Homepage
720:
719:NEAT Homepage
717:
716:
712:
708:
704:
701:
699:
695:
692:
690:
686:
683:
680:
676:
673:
671:
667:
664:
662:
658:
655:
653:
649:
646:
644:
640:
636:
633:
631:
627:
623:
620:
618:
614:
611:
609:
605:
601:
597:
593:
589:
586:
584:
580:
577:
575:
571:
567:
563:
559:
556:
554:
550:
546:
542:
538:
537:
533:
523:
519:
515:
510:
505:
501:
494:
489:
488:
483:
476:
471:
470:
465:
458:
453:
449:
436:
428:
423:
422:
417:
410:
405:
404:
399:
392:
387:
386:
381:
377:
373:
369:
365:
361:
356:
351:
348:(2): 99–127.
347:
343:
336:
331:
330:
325:
316:
312:
308:
304:
299:
294:
290:
286:
282:
278:
271:
268:
261:
258:
251:
248:
242:
240:
236:
229:
225:
222:
221:
217:
215:
208:
206:
204:
200:
196:
188:
186:
184:
180:
175:
167:
165:
158:
156:
152:
145:
140:
138:
136:
132:
128:
120:
118:
115:
112:
108:
104:
96:
94:
92:
87:
83:
76:
74:
72:
64:
62:
60:
56:
52:
48:
44:
40:
36:
32:
19:
694:EvolutionNet
522:the original
499:
481:
463:
435:cite journal
415:
397:
345:
341:
326:Bibliography
280:
276:
270:
260:
250:
212:
192:
177:
162:
153:
149:
124:
116:
110:
106:
102:
100:
88:
84:
80:
68:
34:
30:
29:
626:neat-python
613:neat-python
298:10071/10504
255:Conference.
65:Performance
797:Categories
679:Javascript
670:Javascript
539:Stanley's
230:References
141:Extensions
91:perceptron
675:Neataptic
622:NeuralFit
592:MultiNEAT
588:MultiNEAT
579:SharpNEAT
350:CiteSeerX
179:HyperNEAT
174:HyperNEAT
168:HyperNEAT
77:Algorithm
73:methods.
657:RubyNEAT
541:original
372:12180173
315:20815070
307:25478664
218:See also
133:, a GNU
772:YouTube
725:at the
685:Neat-Ex
594:at the
566:NEAT 4J
514:Bibcode
37:) is a
703:goNEAT
689:Elixir
666:neatjs
652:Python
630:Python
617:Python
608:Python
600:mtNEAT
549:rtNEAT
545:mtNEAT
380:498161
378:
370:
352:
313:
305:
209:odNEAT
189:cgNEAT
146:rtNEAT
135:scheme
635:Encog
562:JNEAT
525:(PDF)
504:arXiv
496:(PDF)
478:(PDF)
460:(PDF)
412:(PDF)
394:(PDF)
376:S2CID
338:(PDF)
311:S2CID
265:2009.
131:Guile
770:, a
705:for
696:for
687:for
677:for
668:for
661:Ruby
659:for
650:for
648:peas
641:and
639:Java
637:for
628:for
615:for
606:and
602:for
581:for
574:Java
572:for
570:ANJI
551:for
547:and
448:help
368:PMID
303:PMID
199:CPPN
183:CPPN
109:and
53:and
35:NEAT
698:C++
604:C++
558:ECJ
553:C++
360:doi
293:hdl
285:doi
127:GPL
45:(a
799::
643:C#
583:C#
568:,
564:,
560:,
543:,
512:.
502:.
498:.
480:.
462:.
439::
437:}}
433:{{
414:.
396:.
374:.
366:.
358:.
346:10
344:.
340:.
309:.
301:.
291:.
281:23
279:.
238:^
105:,
721:(
590:(
516::
506::
484:.
466:.
450:)
446:(
429:.
418:.
400:.
382:.
362::
317:.
295::
287::
111:C
107:B
103:A
33:(
20:)
Text is available under the Creative Commons Attribution-ShareAlike License. Additional terms may apply.