计算神经科学1） 笔记 Rajesh P Rao 讲

- Hubel Wisel response of a neuro in an Intact Cat Brain 

phoneomen they find: when the bar of light goes 45 degree angle, the spike has a very robust response.Frequency of spikes = f(Light bar's orientation) "receptive field" of a neuro as part of  

orientaion of a bar of light that produces the best repsonse

concepts:

on center, off surround receptive field.Each cell tends to respond to light input in only a small area of the retina and visual field. An on-center, off-surround cell becomes more active when only the center of this area is illuminated and less active when the only edges of this area (the surround) are illuminated. Generally a cell is not affected much by input in far away areas of the retina (although recent studies have begun to show that some subtle long-range communication may exist). Also we have off center, on surround receptive field.

Descriptive Model:

Oriented receptive field of a neuron in primary visual cortex (V1）How do quantify this ？using reverse correlation 

Mechinstic model :How do we go from LGN RF converges to V1 RF? - Arrange the inputs gives feedforward connection to V1 cell, but it does not take into account other recurrent response 

Interpretive model

Question: whya re receptive field in V1 shapes like in this way, and what are the computational advanatge of such receptive fields

-efficent coding hypothesis, is this the most efficent way to reconstruct the images 

what are the Receptive fields that minmize the total square pixelwise errors between Image 1 and image 2 and are as Indepdent as possible?

- sparse coding

-predictive coding