No abstract

No abstract

No abstract

Contemporary neural networks simulate the operation of brain subsystems with substantially higher delity than was possible a few years ago. These improvements offer new opportunities to use the cognitive competence of neural networks in uid mechanics. One such opportunity is developed by taking concepts from their origin in neuroscience through a neural network to the uid mechanics. The discussion focuses on one brain subsystem (the "What" pathway of the visual system), adapts an existing neural network model of the visual system (the boundary contour system/feature contour system), and shows how a ow eld scalar can excite a pattern of simulated neural activity. The neural network uses three types of shunting neurons found in the retina, lateral geniculate nucleus, and cortical area V1. When shown the density eld of a supersonic channel ow, certain neurons in the neural network become active when a shock passes nearby. These active neurons provide a pattern of neural activity that could feed into higher-level What pathway neural networks to make more sophisticated inferences about structures present in a ow. The challenges posed by embedding the current neural network in these higher-level brain structures suggest a promising research agenda.