Tree-Based Indirect Encodings for Evolutionary Development of Neural Networks
نویسندگان
چکیده
This paper focuses on TWEANN (Topology and Weight Evolving Artificial Neural Network) methods based on indirect developmental encodings. TWEANNs are Evolutionary Algorithms (EAs) which evolve both topology and parameters (weights) of neural networks. Indirect developmental encoding is an approach inspired by multi-cellular organisms’ development from a single cell (zygote) known from Nature. The possible benefits of such encoding can be seen in Nature: for example, human genome consists of roughly 30 000 genes, which describe more than 20 billion neurons, each linked to as many as 10 000 others. In this work we examine properties of known tree-based indirect developmental encodings: Cellular Encoding and Edge Encoding. Well known Genetic Programming is usualy used to evolve tree structures. We have employed its successors: Gene Expression Programming (GEP) and Grammatical Evolution (GE) to optimize the trees. The combination of well designed developmental encoding and proper optimization method should bring compact genomes able to describe large-scale, modular neural networks. We have compared GE and GEP using a benchmark and found that GE was able to find solution about 7 times faster then GEP. On the other hand GEP solutions were more compact.
منابع مشابه
A Taxonomy for Artificial Embryogeny
A major challenge for evolutionary computation is to evolve phenotypes such as neural networks, sensory systems, or motor controllers at the same level of complexity as found in biological organisms. In order to meet this challenge, many researchers are proposing indirect encodings, that is, evolutionary mechanisms where the same genes are used multiple times in the process of building a phenot...
متن کاملNovel Radial Basis Function Neural Networks based on Probabilistic Evolutionary and Gaussian Mixture Model for Satellites Optimum Selection
In this study, two novel learning algorithms have been applied on Radial Basis Function Neural Network (RBFNN) to approximate the functions with high non-linear order. The Probabilistic Evolutionary (PE) and Gaussian Mixture Model (GMM) techniques are proposed to significantly minimize the error functions. The main idea is concerning the various strategies to optimize the procedure of Gradient ...
متن کاملR-HybrID: Evolution of Agent Controllers with a Hybrisation of Indirect and Direct Encodings
Neuroevolution, the optimisation of artificial neural networks (ANNs) through evolutionary computation, is a promising approach to the synthesis of controllers for autonomous agents. Traditional neuroevolution approaches employ direct encodings, which are limited in their ability to evolve complex or large-scale controllers because each ANN parameter is independently optimised. Indirect encodin...
متن کاملDIFFERENT NEURAL NETWORKS AND MODAL TREE METHOD FOR PREDICTING ULTIMATE BEARING CAPACITY OF PILES
The prediction of the ultimate bearing capacity of the pile under axial load is one of the important issues for many researches in the field of geotechnical engineering. In recent years, the use of computational intelligence techniques such as different methods of artificial neural network has been developed in terms of physical and numerical modeling aspects. In this study, a database of 100 p...
متن کاملShortcomings with Tree-Structured Edge Encodings for Neural Networks
In evolutionary algorithms a common method for encoding neural networks is to use a tree-structured assembly procedure for constructing them. Since node operators have difficulties in specifying edge weights and these operators are execution-order dependent, an alternative is to use edge operators. Here we identify three problems with edge operators: in the initialization phase most randomly cr...
متن کامل