Abbas, F., Triplett, M. A., Goodhill, G. J., & Meyer, M. P. (2017). A Three-Layer Network Model of Direction Selective Circuits in the Optic Tectum. Frontiers in neural circuits, 11, 88.
[Abbas 2017] Define 3 layers as: axons of RGCs, supervisial inhibitory interneurons (SIN), and principal or paraventricular neurons (PVN). ZF larvae. Claim that at first 2 levels there are only 3 principal directions (a trifoil, equally split), but they get recalculated into 4 at last layer (rostro-caudal is absent in the RGCs input, but is present in OT). In intro: there seems to be a controversy in ZF, about how exactly inhibition is tuned in direction-selective circuits in the OT, and whether it is sharpening, or defining the responses. Model: just 10 neurons, as each one represents an entire population; 3 layers, ReLU units, diff equations (not too discrete). Custom set input sensitivity curves and tuned connectivity scheme, reproduces tuning from Ca img experiments
[Abbas 2017] Define 3 layers as: axons of RGCs, supervisial inhibitory interneurons (SIN), and principal or paraventricular neurons (PVN). ZF larvae. Claim that at first 2 levels there are only 3 principal directions (a trifoil, equally split), but they get recalculated into 4 at last layer (rostro-caudal is absent in the RGCs input, but is present in OT). In intro: there seems to be a controversy in ZF, about how exactly inhibition is tuned in direction-selective circuits in the OT, and whether it is sharpening, or defining the responses. Model: just 10 neurons, as each one represents an entire population; 3 layers, ReLU units, diff equations (not too discrete). Custom set input sensitivity curves and tuned connectivity scheme, reproduces tuning from Ca img experiments
