Evolving the placement and density of neurons in the hyperneat substrate
作者： Sebastian Risi , Joel Lehman , Kenneth O. Stanley
关键词: Neuroevolution 、 Artificial neural network 、 Computer science 、 HyperNEAT 、 Node (circuits) 、 Hypercube 、 Artificial intelligence 、 Neuroevolution of augmenting topologies 、 Representation (mathematics)
摘要: The Hypercube-based NeuroEvolution of Augmenting Topologies (HyperNEAT) approach demonstrated that the pattern weights across connectivity an artificial neural network (ANN) can be generated as a function its geometry, thereby allowing large ANNs to evolved for high-dimensional problems. Yet it left user question where hidden nodes should placed in geometry is potentially infinitely dense. To relieve from this decision, paper introduces extension called evolvable-substrate HyperNEAT (ES-HyperNEAT) determines placement and density based on quadtree-like decomposition hypercube novel insight about relationship between node placement. idea representation encodes ANN contains implicit information therefore exploited avoid need evolve explicit In paper, proof concept, ES-HyperNEAT discovers working placements simple navigation domain own, eliminating configure substrate by hand suggesting potential power new approach.
acm.org
core.ac.uk
ucf.edu 
sci-hub.se 参考文章(20)
Jason Gauci, Kenneth O. Stanley, A case study on the critical role of geometric regularity in machine learning national conference on artificial intelligence. pp. 628- 633 ,(2008)
Peter Bentley, Sanjeev Kumar, Three ways to grow designs: a comparison of embryogenies for an evolutionary design problem genetic and evolutionary computation conference. pp. 35- 43 ,(1999)
J. Bongard, Evolving modular genetic regulatory networks congress on evolutionary computation. ,vol. 2, pp. 1872- 1877 ,(2002) , 10.1109/CEC.2002.1004528
Jan Drchal, Jan Koutnik, Miroslav Snorek, HyperNEAT controlled robots learn how to drive on roads in simulated environment congress on evolutionary computation. pp. 1087- 1092 ,(2009) , 10.1109/CEC.2009.4983067
R. A. Finkel, J. L. Bentley, Quad trees a data structure for retrieval on composite keys Acta Informatica. ,vol. 4, pp. 1- 9 ,(1974) , 10.1007/BF00288933
Kenneth O. Stanley, Compositional pattern producing networks: A novel abstraction of development Genetic Programming and Evolvable Machines. ,vol. 8, pp. 131- 162 ,(2007) , 10.1007/S10710-007-9028-8
Jason Gauci, Kenneth O. Stanley, Autonomous evolution of topographic regularities in artificial neural networks Neural Computation. ,vol. 22, pp. 1860- 1898 ,(2010) , 10.1162/NECO.2010.06-09-1042
Jimmy Secretan, Nicholas Beato, David BD Ambrosio, Adelein Rodriguez, Adam Campbell, Kenneth O Stanley, Picbreeder Proceeding of the twenty-sixth annual CHI conference on Human factors in computing systems - CHI '08. pp. 1759- 1768 ,(2008) , 10.1145/1357054.1357328
Kenneth O. Stanley, Risto Miikkulainen, Evolving neural networks through augmenting topologies Evolutionary Computation. ,vol. 10, pp. 99- 127 ,(2002) , 10.1162/106365602320169811
Jeff Clune, Benjamin E. Beckmann, Charles Ofria, Robert T. Pennock, Evolving coordinated quadruped gaits with the HyperNEAT generative encoding congress on evolutionary computation. pp. 2764- 2771 ,(2009) , 10.1109/CEC.2009.4983289