Efficient Learning of Neural Networks with Evolutionary Algorithms

Abstract. In this article we present EANT2, a method that creates neural networks (NNs) by evolutionary reinforcement learning. The structure of NNs is developed using mutation operators, starting from a minimal structure. Their parameters are optimised using CMA-ES. EANT2 can create NNs that are very specialised; they achieve a very good performance while being relatively small. This can be seen in experiments where our method competes with a different one, called NEAT, to create networks that control a robot in a visual servoing scenario.

• Type: Conference
• Year: 2007
• Where: DAGM
• Authors: Nils T. Siebel, Jochen Krause, Gerald Sommer
• Keywords: DAGM 2007, Evolutionary Reinforcement, Image Processing, Neural Networks, Visual Servoing Scenario