Proceedings of the 12th International Conference on Cognitive Modeling (ICCM 2013) (2013) pp. 1-6
At first glance, dendritic+ processing appears to be an extraneous level of detail. Nevertheless, it facilitates richer neural encoding schemes that ultimately lead to simpler networks while improving their neurobiological plausibility. Dendritic+ processing is an example of several modeling tradeoffs: how local complexifications can improve global simplicity, and how functional network circuitry can be traded against representational circuitry. In this paper we introduce dendritic+ processing and its use within a model of narrow slit viewing based on an emergic network architecture (Leibovitz, 2013a). Here, emergic units are intended to abstract functional interactions beyond the level of individual neurons, while emergic ports support structured neural representations that implicate dendritic+ processes. More specifically, visual information such as the amount of blue colour is effectively encoded along with its spatial extent. The visual information can flow over a field of neurons, and because each neuron is also updated with its own spatial extent, this enables an individual neuron to dynamically pick out the portion of the visual flow that is within its remapped receptive field for further processing. The narrow slit views are automatically integrated into the wider anorthoscopic view maintained “in the flow”. This phenomenon emerges at a simple sensory level (supporting perception). This paper suggests one high-level functional role for dendritic+ processing, of maintaining information coherence under continuous motion. Perhaps dendritic+ processing can simplify your network too!
Mendeley saves you time finding and organizing research
Choose a citation style from the tabs below