еще раз о нейроне имени Дженнифер Анистон Интересный свежачок из arxiv.org, как раз на тему
нейрона имени Дженнифер Анистон:
http://arxiv.org/abs/q-bio.NC/0603014Grandmother cells and the storage capacity of the human brain
Authors: John Collins, Dezhe Z. Jin
Comments: 8 pages, 3 figures
Subj-class: Neurons and Cognition
Quian Quiroga et al. have recently discovered neurons that appear to have the characteristics of grandmother (GM) cells. We show that, contrary to the general impression, a GM-cell representation can be optimally efficient, but that it is necessarily accompanied by a population of cells giving a conventional distributed representation of the input. From the data we calculate that only about 5% of the cells are in the active distributed-code population, and that the number of categories coded by the apparent GM-cells is less than 4200. This is rather low for them to be classic GM cells, unless there are certain plausible biases in cell selection. A natural alternative is that they are GM-like cells coding for explicit memories. We discuss the strong consequences both for the architecture of neural systems and for the statistics of neural firing and the patterns of synaptic connectivity.
То есть, утверждается, что кодирование небольшой инвариантной группой нейронов дублируется распределенным кодированием в большой популяции каких-нибудь других нейронов. Надо бы ознакомиться.
Current Mood: relaxedCurrent Music: Bohren und der Club of Gore: BLACK EARTH