A CNN model framework and simulator for biological sensory systems
D. Bálya, Botond Roska, T. Roska, F. S. Werblin · SZTAKI Publication Repository (Hungarian Academy of Sciences) · 2001
An abstract neuron model framework and simulator is proposed for qualitative spatio-temporal studies for biologic sensory systems. The proposed framework defines a huge number of layer-ordered simple base cells with neuromorphic relevant parameters (1). The created models can be translated to hardware-level general multi-layer Cellular Neural/non-linear Neural Network (CNN) templates (2). The framework is used to create a complete mammalian retina model. The definition of the model elements is based on retinal anatomy and electro-physiology measurements (1). The developed model complexity is moderate, compared to a fully neuromorphic one (3) therefore chances of its implementation on a complex- cell CNN Universal Machine (4) chips (5) using multi- layer technology (6) and algorithmic programmability are good. .