Lamarckian training of feedforward neural networks.
Paulo Cortez, Miguel Rocha, José Neves · 2001
Abstract. Living creatures improve their adaptation capabilities to a changing world by means of two orthogonal processes: evolution and lifetime learning. Within Articial Intelligence, both mechanisms in-spired the development of non-orthodox problem solving tools, namely Genetic and Evolutionary Algorithms (GEAs) and Articial Neural Net-works (ANNs). Several local search gradient-based methods have been developed for ANN training, with considerable success; however, in some situations, such procedures may lead to local minima. Under this sce-nario, the combination of evolution and learning techniques, may lead to better results (e.g., global optima). Comparative tests on several Ma-chine Learning tasks attest this claim.