In the paper it mentions that During training, we reset the output layer when switching to a new object. In the brain, there are several ways the equivalent of a reset could occur..... So do we have any recent update of how specifically it is done? e.g, if I tried to touch on my coffee cup with my left hand, and then I switch to scroll my mouse. How do we tell that the scroll is a feature associated with the mouse not the coffee cup?
Thanks!!
This is still a bit of an open question. There are several hypotheses for how the reset could happen.
I’ve discussed my personal favorite hypothesis on the following thread:
I think the most obvious explanation is that one cannot always predict the location of the feature of the scroll wheel on the mouse from the position and orientation of the coffee cup. That lack of spatial correlation between the properties of the two objects is a powerful signal that the features on each likely do not belong to the same object.