Evolution of Neural Network’s Architecture through Symbiotic Neuroevolution
نویسندگان
چکیده
In this paper an extension of SANE that simultaneously evolves the weights and architecture of an MLP neural network is presented. The symbiotic adaptive neuroevolution (SANE) system coevolves a population of neurons that cooperate to form a functioning neural network. Evolutionary Strategies (ES) is applied to evolve the network weights. In order to increase the evolving system performance and achieving global optimum convergence the concept of age is introduced in calculating the fitness. The current investigation focuses on evolving neurocontrollers for a nonlinear unstable system (cart-pole problem). The results indicate the suitability for using SANE to evolve weights and architecture of a neurocontroller. Index Terms – SANE, Coevolution, neural network, neurocontroller.
منابع مشابه
Incorporating Advice into Evolution of Neural Networks
Neuroevolution is a promising learning method in tasks with extremely large state and action spaces and hidden states. Recent advances allow neuroevolution to take place in real time, making it possible to e.g. construct video games with adaptive agents. Often some of the desired behaviors for such agents are known, and it would make sense to prescribe them, rather than requiring evolution to d...
متن کاملEvolving neural networks for strategic decision-making problems
Evolution of neural networks, or neuroevolution, has been a successful approach to many low-level control problems such as pole balancing, vehicle control, and collision warning. However, certain types of problems-such as those involving strategic decision-making-have remained difficult for neuroevolution to solve. This paper evaluates the hypothesis that such problems are difficult because the...
متن کاملForming Neural Networks Through Efficient and Adaptive Coevolution
This article demonstrates the advantages of a cooperative, coevolutionary search in difficult control problems. The symbiotic adaptive neuroevolution (SANE) system coevolves a population of neurons that cooperate to form a functioning neural network. In this process, neurons assume different but overlapping roles, resulting in a robust encoding of control behavior. SANE is shown to be more effi...
متن کاملSafe Mutations for Deep and Recurrent Neural Networks through Output Gradients
While neuroevolution (evolving neural networks) has a successful track record across a variety of domains from reinforcement learning to artificial life, it is rarely applied to large, deep neural networks. A central reason is that while random mutation generally works in low dimensions, a random perturbation of thousands or millions of weights is likely to break existing functionality, providi...
متن کاملEvolving Deep Neural Networks
e success of deep learning depends on nding an architecture to t the task. As deep learning has scaled up to more challenging tasks, the architectures have become dicult to design by hand. is paper proposes an automated method, CoDeepNEAT, for optimizing deep learning architectures through evolution. By extending existing neuroevolution methods to topology, components, and hyperparameters,...
متن کامل