Various optimization methods and network architectures are used by convolutional\nneural networks (CNNs). Each optimization method and network architecture style have their own\nadvantages and representation abilities. To make the most of these advantages, evolutionary-fuzzyintegral-\nbased convolutional neural networks (EFI-CNNs) are proposed in this paper. The proposed\nEFI-CNNs were verified by way of face classification of age and gender. The trained CNNsâ?? outputs\nwere set as inputs of a fuzzy integral. The classification results were operated using either Sugeno\nor Choquet output rules. The conventional fuzzy density values of the fuzzy integral were decided\nby heuristic experiments. In this paper, particle swarm optimization (PSO) was used to adaptively\nfind optimal fuzzy density values. To combine the advantages of each CNN type, the evaluation of\neach CNN type in EFI-CNNs is necessary. Three CNN structures, AlexNet, very deep convolutional\nneural network (VGG16), and GoogLeNet, and three databases, computational intelligence\napplication laboratory (CIA), Morph, and cross-age celebrity dataset (CACD2000), were used in\nexperiments to classify age and gender. The experimental results show that the proposed method\nachieved 5.95% and 3.1% higher accuracy, respectively, in classifying age and gender.
Loading....