# Tag Info

1

What they mean is that as long as you rotate $C_i$ and $C_j$ by the same amount, the dot product will be the same. This is rotational invariance because the dot product is invariant to coherent rotation of the relevant vectors. In neural terms, the correlation between the activity of two neurons (in a population representing a one dimensional circular ...

0

This is a very difficult question that we don't know the answer to yet. Here are some references. Impermanence of dendritic spines in live adult CA1 hippocampus Alessio Attardo, James E. Fitzgerald & Mark J. Schnitzer http://www.nature.com/nature/journal/v523/n7562/full/nature14467.html Strikingly, CA1 spine turnover dynamics differed sharply from ...

1

It is generally thought that thalamic input comes in layer 4, feed back from higher areas come through layer 1 to layer 2/3 and feed forward is sent from the deeper layers. see Canonical Microcircuits for Predictive Coding Andre Bastos, W. Martin Usrey, Rick Adams, George Mangun, Pascal Fries, and Karl Friston ...

0

Try: The development of topography in the visual cortex: a review of models NV Swindale - Network: Computation in neural systems, 1996 - Taylor & Francis. https://www.cs.cmu.edu/afs/cs/academic/class/15883-f13/readings/swindale-1996.pdf or: Congenital visual pathway abnormalities: a window onto cortical stability and plasticity MB Hoffmann, SO Dumoulin ...

4

It is something of an oversimplification to say that there are separate visual pathways for both color and shape. There are many cells, even in V1, which are selective for both colour and shape (or at least orientation). While there are regions more sensitive to some features than others, there are plenty of neurons which combine features. It is also ...

-2

It is correct that only a small percentage of neurons increase their activities relative to their baseline level in response to new stimuli or to more abstract thoughts. There are many computational advantages to this. For instance, in deep learning, sparse coding is a class of unsupervised methods for learning sets of over-complete bases to represent ...

0

They are bound unconsciously, and this is known as the binding problem. You can make experiments (using very short durations and/or "masks") however, to interfere with this process. The result is that sometimes you consciously bind the wrong color and shape.

Top 50 recent answers are included