Synaptic plasticity enables adaptive self-tuning critical networks.
Synaptic plasticity enables adaptive self-tuning critical networks.
复制标题
DOI：
10.1371/journal.pcbi.1004043
复制
发表时间：
2015-01
影响因子：
4.3
通讯作者：
Srinivasa N
中科院分区：
文献类型：
--
作者：
Stepp N;Plenz D;Srinivasa N
在休息时，哺乳动物的大脑皮层会显示自发的神经活动。单个神经元在休息期间的尖峰被描述为不规则和异步的。相比之下，最近在体内和体外人群自发活动的措施，使用LFP，EEG，MEG或功能磁共振成像表明，默认状态的皮质是关键的，表现为自发的，规模不变的，级联的活动被称为神经元雪崩。关键性使网络保持最佳信息处理的状态，但这种观点似乎很难与明显不规则的单神经元尖峰相协调。在这里，我们模拟了一个10，000个神经元，确定性的，可塑性的尖峰神经元网络。我们表明，短期和长期突触可塑性的结合使这些网络能够在面对内在的（即自我维持的）异步尖峰时表现出临界性。短暂的外部扰动导致自适应的，长期的修改内在的网络连接通过长期的兴奋性可塑性，而长期的抑制性可塑性，使网络的快速自我调整回到一个临界状态。临界状态的特征在于分支参数在1附近振荡，临界指数接近-3/2，并且自相似参数在0.5和1之间的长尾分布。神经网络，无论是人工的还是生物的，都是由连接在一起的单个单元组成的，这些单元相互发送和接收称为尖峰的能量包。虽然简单描述，但存在大量可能的实现，实例化和各种神经网络。其中一些网络在随机性和有序性之间，以及在衰变和爆炸之间保持着临界平衡。为特定网络选择正确的属性和参数以达到这种临界状态可能是困难和耗时的。单元之间的连接强度可能会随着时间的推移通过突触可塑性而改变，我们利用这种机制来创建一个自我调整到临界状态的网络。更具体地说，来自兴奋性和抑制性可塑性的相反力量的相互作用创造了一种平衡，允许自我调节发生。这种自我调整采用相对简单的尖峰单位，并以创建复杂行为的方式将它们连接起来。我们的研究结果对在硬件中实现的人工神经网络的设计具有影响，其中参数调整可能是昂贵的，但也可能提供对生物网络关键性质的洞察。
During rest, the mammalian cortex displays spontaneous neural activity. Spiking of single neurons during rest has been described as irregular and asynchronous. In contrast, recent in vivo and in vitro population measures of spontaneous activity, using the LFP, EEG, MEG or fMRI suggest that the default state of the cortex is critical, manifested by spontaneous, scale-invariant, cascades of activity known as neuronal avalanches. Criticality keeps a network poised for optimal information processing, but this view seems to be difficult to reconcile with apparently irregular single neuron spiking. Here, we simulate a 10,000 neuron, deterministic, plastic network of spiking neurons. We show that a combination of short- and long-term synaptic plasticity enables these networks to exhibit criticality in the face of intrinsic, i.e. self-sustained, asynchronous spiking. Brief external perturbations lead to adaptive, long-term modification of intrinsic network connectivity through long-term excitatory plasticity, whereas long-term inhibitory plasticity enables rapid self-tuning of the network back to a critical state. The critical state is characterized by a branching parameter oscillating around unity, a critical exponent close to -3/2 and a long tail distribution of a self-similarity parameter between 0.5 and 1. Neural networks, whether artificial or biological, consist of individual units connected together that mutually send and receive parcels of energy called spikes. While simply described, there is a vast space of possible implementations, instantiations, and varieties of neural networks. Some of these networks are critically balanced between randomness and order, and between death by decay and death by explosion. Selecting just the right properties and parameters for a particular network to reach this critical state can be difficult and time-consuming. The strength of connections between units may change over time via synaptic plasticity, and we exploit this mechanism to create a network that self-tunes to criticality. More specifically, the interplay of opposing forces from excitatory and inhibitory plasticity create a balance that allows self-tuning to take place. This self-tuning takes relatively simple spiking units and connects them in a way that creates complex behavior. Our results have implications for the design of artificial neural networks implemented in hardware, where parameter tuning can be costly, but may provide insight into the critical nature of biological networks as well.
登录
查看更多内容
影响因子：
3.5
作者：
Cruz-Albrecht, Jose M.;Derosier, Timothy;Srinivasa, Narayan
通讯作者：
Srinivasa, Narayan
影响因子：
64.8
作者：
Markram, H;Tsodyks, M
通讯作者：
Tsodyks, M
影响因子：
4.3
作者：
Boerlin M;Machens CK;Denève S
通讯作者：
Denève S
影响因子：
2.9
作者：
Bertschinger, N;Natschl채ger, T
通讯作者：
Natschl채ger, T
影响因子：
19.6
作者：
Levina, A.;Herrmann, J. M.;Geisel, T.
通讯作者：
Geisel, T.