Limits...
Effects of Spike Anticipation on the Spiking Dynamics of Neural Networks.

de Santos-Sierra D, Sanchez-Jimenez A, Garcia-Vellisca MA, Navas A, Villacorta-Atienza JA - Front Comput Neurosci (2015)

Bottom Line: We show that the introduction of intermediary neurons in the network enhances spike anticipation and analyse how these variations in spike anticipation can significantly change the firing regime of the neural network according to its functional and structural properties.In addition we show that the interspike interval (ISI), one of the main features of the neural response associated with the information coding, can be closely related to spike anticipation by each spike, and how synaptic plasticity can be modulated through that relationship.This study has been performed through numerical simulation of a coupled system of Hindmarsh-Rose neurons.

View Article: PubMed Central - PubMed

Affiliation: Group of Biometrics, Biosignals and Security, Research Centre for Smart Buildings and Energy Efficiency (CeDInt), Technical University of Madrid Madrid, Spain ; Laboratory of Computational System Biology, Center for Biomedical Technology, Technical University of Madrid Madrid, Spain.

ABSTRACT
Synchronization is one of the central phenomena involved in information processing in living systems. It is known that the nervous system requires the coordinated activity of both local and distant neural populations. Such an interplay allows to merge different information modalities in a whole processing supporting high-level mental skills as understanding, memory, abstraction, etc. Though, the biological processes underlying synchronization in the brain are not fully understood there have been reported a variety of mechanisms supporting different types of synchronization both at theoretical and experimental level. One of the more intriguing of these phenomena is the anticipating synchronization, which has been recently reported in a pair of unidirectionally coupled artificial neurons under simple conditions (Pyragiene and Pyragas, 2013), where the slave neuron is able to anticipate in time the behavior of the master one. In this paper, we explore the effect of spike anticipation over the information processing performed by a neural network at functional and structural level. We show that the introduction of intermediary neurons in the network enhances spike anticipation and analyse how these variations in spike anticipation can significantly change the firing regime of the neural network according to its functional and structural properties. In addition we show that the interspike interval (ISI), one of the main features of the neural response associated with the information coding, can be closely related to spike anticipation by each spike, and how synaptic plasticity can be modulated through that relationship. This study has been performed through numerical simulation of a coupled system of Hindmarsh-Rose neurons.

No MeSH data available.


Related in: MedlinePlus

Synaptic plasticity under spike anticipation. (A) Neural network with a loop composed of three intermediate neurons between master (M) and slave (S), and where information flux converges into the neuron X. Red arrow points the synapse whose plasticity is analyzed by monitoring the evolution of its coupling weight kSX. The bottom inset shows in detail the connection scheme involving the analyzed plasticity. (B) Scheme of synaptic plasticity when potentiation appears; the S event (the maximum of the spike received by X neuron from the slave one), denoted by a red vertical line, precedes the M event (blue vertical line) inside the temporal window (gray lines) due to spike anticipation. (C) When there is no anticipation the M event precedes the S one and the synapse is depressed. (D) The relationship between anticipation and ISI (Figure 5C) reveals that the wider the window, the larger the number of spikes that can coincide inside the window (denoted by magenta points), since the corresponding anticipations lie in the range of the window width. This panel corresponds to w = 0.5. (E) When w = 0.3 the range of anticipations less than or equal to w is smaller, so there are fewer spikes that can potentiate the synapse. (F–H) Present the time evolution of the synaptic weight kSX evaluated for three different windows w = 0.5, w = 0.4, and w = 0.3, respectively. The increasing and decreasing rates of the synaptic reinforcement (i.e., the change of the synaptic weight) are ri = 0.02 and rd = ri/7, respectively.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4663270&req=5

Figure 6: Synaptic plasticity under spike anticipation. (A) Neural network with a loop composed of three intermediate neurons between master (M) and slave (S), and where information flux converges into the neuron X. Red arrow points the synapse whose plasticity is analyzed by monitoring the evolution of its coupling weight kSX. The bottom inset shows in detail the connection scheme involving the analyzed plasticity. (B) Scheme of synaptic plasticity when potentiation appears; the S event (the maximum of the spike received by X neuron from the slave one), denoted by a red vertical line, precedes the M event (blue vertical line) inside the temporal window (gray lines) due to spike anticipation. (C) When there is no anticipation the M event precedes the S one and the synapse is depressed. (D) The relationship between anticipation and ISI (Figure 5C) reveals that the wider the window, the larger the number of spikes that can coincide inside the window (denoted by magenta points), since the corresponding anticipations lie in the range of the window width. This panel corresponds to w = 0.5. (E) When w = 0.3 the range of anticipations less than or equal to w is smaller, so there are fewer spikes that can potentiate the synapse. (F–H) Present the time evolution of the synaptic weight kSX evaluated for three different windows w = 0.5, w = 0.4, and w = 0.3, respectively. The increasing and decreasing rates of the synaptic reinforcement (i.e., the change of the synaptic weight) are ri = 0.02 and rd = ri/7, respectively.

Mentions: Here we study synaptic plasticity under spike anticipation focusing on the previous idea of causal potentiation/depression of neural connectivity. For doing that we study the neural network presented in Figure 2, paying attention to the synapse connecting slave and X neurons (red arrow in Figure 6A) and showing how plasticity can be affected by the complex structure of the relationship between spike anticipation and ISIs (Figure 5). The key factor of the network in Figure 2 is the input provided to X neuron by master (M) and slave (S) neurons since their causal correlation will be the factor for potentiating or depressing the synapse: when the X neuron receives a spike from the S neuron before the spike from the M neuron arrives the S–X synapse is potentiated, being depressed on the contrary. This mechanism is inspired by spike timing dependent plasticity or STDP models (Friedel and van Hemmen, 2008; Butts and Kanold, 2010) with the difference that they focus on the input–output spike correlation and we analyse the causal relationship between spikes of different inputs, assuming that this is the reflection of a causal correlation between input and output. Figure 6B shows a scheme of such a process, where red and blue vertical lines denote the maxima of the S and M spikes, respectively, and gray vertical lines denote the temporal window where both spikes must coincide in the correct order to elicit potentiation. In the studied neural network spike anticipation provides this preceding behavior and so the synaptic potentiation. On the contrary, when no anticipation exists the M spike precedes the S spike and the synapse is depressed (Figure 6C).


Effects of Spike Anticipation on the Spiking Dynamics of Neural Networks.

de Santos-Sierra D, Sanchez-Jimenez A, Garcia-Vellisca MA, Navas A, Villacorta-Atienza JA - Front Comput Neurosci (2015)

Synaptic plasticity under spike anticipation. (A) Neural network with a loop composed of three intermediate neurons between master (M) and slave (S), and where information flux converges into the neuron X. Red arrow points the synapse whose plasticity is analyzed by monitoring the evolution of its coupling weight kSX. The bottom inset shows in detail the connection scheme involving the analyzed plasticity. (B) Scheme of synaptic plasticity when potentiation appears; the S event (the maximum of the spike received by X neuron from the slave one), denoted by a red vertical line, precedes the M event (blue vertical line) inside the temporal window (gray lines) due to spike anticipation. (C) When there is no anticipation the M event precedes the S one and the synapse is depressed. (D) The relationship between anticipation and ISI (Figure 5C) reveals that the wider the window, the larger the number of spikes that can coincide inside the window (denoted by magenta points), since the corresponding anticipations lie in the range of the window width. This panel corresponds to w = 0.5. (E) When w = 0.3 the range of anticipations less than or equal to w is smaller, so there are fewer spikes that can potentiate the synapse. (F–H) Present the time evolution of the synaptic weight kSX evaluated for three different windows w = 0.5, w = 0.4, and w = 0.3, respectively. The increasing and decreasing rates of the synaptic reinforcement (i.e., the change of the synaptic weight) are ri = 0.02 and rd = ri/7, respectively.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4663270&req=5

Figure 6: Synaptic plasticity under spike anticipation. (A) Neural network with a loop composed of three intermediate neurons between master (M) and slave (S), and where information flux converges into the neuron X. Red arrow points the synapse whose plasticity is analyzed by monitoring the evolution of its coupling weight kSX. The bottom inset shows in detail the connection scheme involving the analyzed plasticity. (B) Scheme of synaptic plasticity when potentiation appears; the S event (the maximum of the spike received by X neuron from the slave one), denoted by a red vertical line, precedes the M event (blue vertical line) inside the temporal window (gray lines) due to spike anticipation. (C) When there is no anticipation the M event precedes the S one and the synapse is depressed. (D) The relationship between anticipation and ISI (Figure 5C) reveals that the wider the window, the larger the number of spikes that can coincide inside the window (denoted by magenta points), since the corresponding anticipations lie in the range of the window width. This panel corresponds to w = 0.5. (E) When w = 0.3 the range of anticipations less than or equal to w is smaller, so there are fewer spikes that can potentiate the synapse. (F–H) Present the time evolution of the synaptic weight kSX evaluated for three different windows w = 0.5, w = 0.4, and w = 0.3, respectively. The increasing and decreasing rates of the synaptic reinforcement (i.e., the change of the synaptic weight) are ri = 0.02 and rd = ri/7, respectively.
Mentions: Here we study synaptic plasticity under spike anticipation focusing on the previous idea of causal potentiation/depression of neural connectivity. For doing that we study the neural network presented in Figure 2, paying attention to the synapse connecting slave and X neurons (red arrow in Figure 6A) and showing how plasticity can be affected by the complex structure of the relationship between spike anticipation and ISIs (Figure 5). The key factor of the network in Figure 2 is the input provided to X neuron by master (M) and slave (S) neurons since their causal correlation will be the factor for potentiating or depressing the synapse: when the X neuron receives a spike from the S neuron before the spike from the M neuron arrives the S–X synapse is potentiated, being depressed on the contrary. This mechanism is inspired by spike timing dependent plasticity or STDP models (Friedel and van Hemmen, 2008; Butts and Kanold, 2010) with the difference that they focus on the input–output spike correlation and we analyse the causal relationship between spikes of different inputs, assuming that this is the reflection of a causal correlation between input and output. Figure 6B shows a scheme of such a process, where red and blue vertical lines denote the maxima of the S and M spikes, respectively, and gray vertical lines denote the temporal window where both spikes must coincide in the correct order to elicit potentiation. In the studied neural network spike anticipation provides this preceding behavior and so the synaptic potentiation. On the contrary, when no anticipation exists the M spike precedes the S spike and the synapse is depressed (Figure 6C).

Bottom Line: We show that the introduction of intermediary neurons in the network enhances spike anticipation and analyse how these variations in spike anticipation can significantly change the firing regime of the neural network according to its functional and structural properties.In addition we show that the interspike interval (ISI), one of the main features of the neural response associated with the information coding, can be closely related to spike anticipation by each spike, and how synaptic plasticity can be modulated through that relationship.This study has been performed through numerical simulation of a coupled system of Hindmarsh-Rose neurons.

View Article: PubMed Central - PubMed

Affiliation: Group of Biometrics, Biosignals and Security, Research Centre for Smart Buildings and Energy Efficiency (CeDInt), Technical University of Madrid Madrid, Spain ; Laboratory of Computational System Biology, Center for Biomedical Technology, Technical University of Madrid Madrid, Spain.

ABSTRACT
Synchronization is one of the central phenomena involved in information processing in living systems. It is known that the nervous system requires the coordinated activity of both local and distant neural populations. Such an interplay allows to merge different information modalities in a whole processing supporting high-level mental skills as understanding, memory, abstraction, etc. Though, the biological processes underlying synchronization in the brain are not fully understood there have been reported a variety of mechanisms supporting different types of synchronization both at theoretical and experimental level. One of the more intriguing of these phenomena is the anticipating synchronization, which has been recently reported in a pair of unidirectionally coupled artificial neurons under simple conditions (Pyragiene and Pyragas, 2013), where the slave neuron is able to anticipate in time the behavior of the master one. In this paper, we explore the effect of spike anticipation over the information processing performed by a neural network at functional and structural level. We show that the introduction of intermediary neurons in the network enhances spike anticipation and analyse how these variations in spike anticipation can significantly change the firing regime of the neural network according to its functional and structural properties. In addition we show that the interspike interval (ISI), one of the main features of the neural response associated with the information coding, can be closely related to spike anticipation by each spike, and how synaptic plasticity can be modulated through that relationship. This study has been performed through numerical simulation of a coupled system of Hindmarsh-Rose neurons.

No MeSH data available.


Related in: MedlinePlus