The list-length effect means simply that adding new items to a list harms memory for
other list items. For the pseudo-recurrent model, this easily falls out of the manner in which
serial learning is done. The further back in a list an item occurs, the more poorly it is
remembered (see Figures 5 and 6). Thus, when more items are added to a serial list, the less
chance the earlier ones have of being remembered. While this is true, of course, of standard
backpropagation models, as can be seen in Figures 5 and 6, beyond one or two items back, all
items have basically been forgotten. The pseudo-recurrent network’s ability to forget
gradually allows it to show a plausible list-length effect
The list-length effect means simply that adding new items to a list harms memory forother list items. For the pseudo-recurrent model, this easily falls out of the manner in whichserial learning is done. The further back in a list an item occurs, the more poorly it isremembered (see Figures 5 and 6). Thus, when more items are added to a serial list, the lesschance the earlier ones have of being remembered. While this is true, of course, of standardbackpropagation models, as can be seen in Figures 5 and 6, beyond one or two items back, allitems have basically been forgotten. The pseudo-recurrent network’s ability to forgetgradually allows it to show a plausible list-length effect
การแปล กรุณารอสักครู่..
