15/08/27 09:26:10.42 HOP0lL8T.net
>>587
それはこの論文のことだろ。
Towards real-time, mouse-scale cortical simulations
URLﾘﾝｸ(www.modha.org)
これを読む限りマウスの脳に【忠実な】シミュレーションとは程遠い。
まず個々のニューロンの機能をどの程度忠実にシミュレーションしてるか分からない。
シナプスに関してもそう。
それからモデル化された神経ネットワーク全体がマウスの脳のそれとどの程度似ているのかも分からない。
We deployed the simulator on a 4096-processor BlueGene/L supercomputer with 256 MB per CPU
. Using spiking neurons [2], we were able to represent 10^6 neurons (80% excitatory) and
6300 synapses per neuron in the 1 TB main memory of the system.
Using a synthetic pattern of neuronal interconnections, at a 1 ms resolution and an average
firing rate of 1 Hz, we were able to run 1s of model time in 10s of real time.
On other (smaller) models, we have observed biologically consistent dynamical properties such
as spontaneous formation of neuronal groups and synchronous/asynchronous firing patterns. Our goal is to continually refine the computational architecture and to add neurobiological details
such as macro and micro neuroanatomy, neuromodulators, and more detailed neuron and synapse
dynamical equations.
因みにニューロンの機能は完全には解明されていない。
URLﾘﾝｸ(redwood.berkeley.edu)
The overall input-output characteristic of a neuron is very complicated and
still not fully understood.