Deep learning through generative and developmental system

Deep learning through supervised and unsupervised learning has demonstrated human competitive performance on some visual tasks; however, evolution played an important role in the development of biological visual systems. Thus evolutionary approaches, specifically the Hypercube-based NeuroEvolution of Augmenting Topologies, are applied to deep learning tasks in this paper. Results indicate HyperNEAT alone struggles in image classification, but trains effective feature extractors for other machine learning approaches.