Limits...
Synaptic plasticity enables adaptive self-tuning critical networks.

Stepp N, Plenz D, Srinivasa N - PLoS Comput. Biol. (2015)

Bottom Line: 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.

View Article: PubMed Central - PubMed

Affiliation: Center for Neural and Emergent Systems, Information and System Sciences Lab, HRL Laboratories LLC, Malibu, California, United States of America.

ABSTRACT
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.

Show MeSH
Evolution of synaptic conductances in the network.A) The evolution of E synapses in the E–I network shows that conductances develop a bimodal distribution, heavily favoring weights close to zero. The initial E synaptic conductances were set to 0.5 nS for all synapses. B) The mean square error (MSE) between E synaptic conductances with and without perturbations during network evolution. Each perturbation causes a jump in weight differences, with weights continuing to diverge between perturbations. C) The evolution of I synapses in the E–I network shows that the synaptic weights develop a unimodal, and qualitatively exponential, distribution of synaptic conductances. The initial I synaptic conductances were set at 0.5 nS for all synapses. D) The MSE between I synaptic conductance histograms with and without perturbations. Inhibitory weights tend to recover from perturbations more than excitatory weights.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4295840&req=5

pcbi.1004043.g006: Evolution of synaptic conductances in the network.A) The evolution of E synapses in the E–I network shows that conductances develop a bimodal distribution, heavily favoring weights close to zero. The initial E synaptic conductances were set to 0.5 nS for all synapses. B) The mean square error (MSE) between E synaptic conductances with and without perturbations during network evolution. Each perturbation causes a jump in weight differences, with weights continuing to diverge between perturbations. C) The evolution of I synapses in the E–I network shows that the synaptic weights develop a unimodal, and qualitatively exponential, distribution of synaptic conductances. The initial I synaptic conductances were set at 0.5 nS for all synapses. D) The MSE between I synaptic conductance histograms with and without perturbations. Inhibitory weights tend to recover from perturbations more than excitatory weights.

Mentions: As argued above, self-tuning to criticality requires change within the network, which is most readily effected by altering synaptic conductances. Fig. 6 shows the changing distribution of these synaptic “weights” over the course of the simulation. While the effect of perturbations on E and I synaptic weights is evident visually, by the three ridges in the weight-time plot respectively (Fig. 6a,c) it can be further quantified by comparing the perturbed and unperturbed weights using a simple mean square error measure (see Methods). This measure shows that each perturbation caused strong change to the synaptic conductances in both E and I weights, which outlasted the perturbation. Thus, these perturbations also significantly changed the effective network topology as well.


Synaptic plasticity enables adaptive self-tuning critical networks.

Stepp N, Plenz D, Srinivasa N - PLoS Comput. Biol. (2015)

Evolution of synaptic conductances in the network.A) The evolution of E synapses in the E–I network shows that conductances develop a bimodal distribution, heavily favoring weights close to zero. The initial E synaptic conductances were set to 0.5 nS for all synapses. B) The mean square error (MSE) between E synaptic conductances with and without perturbations during network evolution. Each perturbation causes a jump in weight differences, with weights continuing to diverge between perturbations. C) The evolution of I synapses in the E–I network shows that the synaptic weights develop a unimodal, and qualitatively exponential, distribution of synaptic conductances. The initial I synaptic conductances were set at 0.5 nS for all synapses. D) The MSE between I synaptic conductance histograms with and without perturbations. Inhibitory weights tend to recover from perturbations more than excitatory weights.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4295840&req=5

pcbi.1004043.g006: Evolution of synaptic conductances in the network.A) The evolution of E synapses in the E–I network shows that conductances develop a bimodal distribution, heavily favoring weights close to zero. The initial E synaptic conductances were set to 0.5 nS for all synapses. B) The mean square error (MSE) between E synaptic conductances with and without perturbations during network evolution. Each perturbation causes a jump in weight differences, with weights continuing to diverge between perturbations. C) The evolution of I synapses in the E–I network shows that the synaptic weights develop a unimodal, and qualitatively exponential, distribution of synaptic conductances. The initial I synaptic conductances were set at 0.5 nS for all synapses. D) The MSE between I synaptic conductance histograms with and without perturbations. Inhibitory weights tend to recover from perturbations more than excitatory weights.
Mentions: As argued above, self-tuning to criticality requires change within the network, which is most readily effected by altering synaptic conductances. Fig. 6 shows the changing distribution of these synaptic “weights” over the course of the simulation. While the effect of perturbations on E and I synaptic weights is evident visually, by the three ridges in the weight-time plot respectively (Fig. 6a,c) it can be further quantified by comparing the perturbed and unperturbed weights using a simple mean square error measure (see Methods). This measure shows that each perturbation caused strong change to the synaptic conductances in both E and I weights, which outlasted the perturbation. Thus, these perturbations also significantly changed the effective network topology as well.

Bottom Line: 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.

View Article: PubMed Central - PubMed

Affiliation: Center for Neural and Emergent Systems, Information and System Sciences Lab, HRL Laboratories LLC, Malibu, California, United States of America.

ABSTRACT
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.

Show MeSH