Interaction of neuronal and network mechanisms on firing propagation in a feedforward network
Introduction
Neurons convert a stimulus from the external environment or signals from upstream neurons into a train of action potentials (spikes), which are believed to be the fundamental process in the realization of brain function. The brain is the most complex organ in the human body, consisting of approximately 100 billion neurons. In the brain, neurons exhibit great diversity in their shapes and functions [1], [2]. Triggered with an external stimulus, different neurons respond with different firing patterns. Based on the firing patterns observed by intracellular recordings, neurons in the mammalian brain can be divided into various types [3], [4]. Regular spiking, tonic bursting and fast spiking neurons are typical neurons that were discovered in in vivo physiology and laminar network anatomy experiments [3], [5]. In the subiculum, different pyramidal neurons can generate regular spiking and bursting spiking patterns [6]. Neuronal diversity contributes to the emergent properties of neural networks and, consequently, plays an important role in information processing in the nervous system [7], [8], [9]. Therefore, studies about the activities of the neuronal network of a brain region should take into account intrinsic behavioral diversity both within and between neuron types.
Moreover, the nervous system is a highly modular structure [10]. Information processing in the nervous system is related to different functional groups of neurons, by which the information is transferred from one group to its downstream connected groups [11]. Thus, conditions under which spiking activity can be propagated among the neuronal groups are a crucial issue for information processing in the modular brain. In the last decade, a multilayer feedforward network (FFN) has been introduced to exploit the issue.
Multilayer FFNs, which provide important insights into the mechanisms of cortical computation, can mimic the properties of the propagation of spiking activity. Each layer of the network is related to a functional group of neurons, in which neurons in one group receive inputs from many neurons in the previous group, and thus the information is transmitted from one group to the next [11]. Recently, several computational studies have identified two dynamic activity modes that support the propagation of rate coding and temporal coding in the FFN [12], [13], [14]. When neurons in the first layer are subject to white noise, firing rate can be propagated in an FFN by the synchronized firings of Hodgkin–Huxley (HH) neurons [15], [16]. When weak signals are input to the neurons in layer 1, successive layers are able to propagate and amplify the signals while the neurons are subject to intrinsic or external noise above a certain level [17], [18]. Current computational studies also reveal that the propagation of spiking activities in an FFN depends on the interlayer connection probability, synaptic intensity, noise, and input signals [15], [16], [19], [20], [21]. Nevertheless, the diversity of neurons is rarely considered in studies of FFNs. Thus, it is not yet clear that the behaviors of spiking activity propagating within an FFN consider the intrinsic electrical properties of different neuronal types.
As mentioned earlier, neuronal diversity plays a crucial role in information processing in the nervous system. In the mammalian brain, the diversity of neuronal and nonneuronal cell types guarantees the execution of high-order cognitive, sensory, and motor behaviors [22]. Networks with diverse neuronal types offer superior decoding ability compared with homogeneous networks [23]. In a hybrid coupled neural network, it was shown that different types of population spiking patterns can emerge [24] and strongly synchronized population spiking events lead to complete activity cessation [25], by changing the types of connections in the network. It has also been found that increasing the diversity of intrinsic neuronal types can enhance the encoding performance of neuronal populations [26]. Andrew Bogaard et al. found that the response of a neuronal network is strongly affected by intrinsic neuronal properties and that the introduction of a small number of cells with different excitability properties can profoundly influence the spatiotemporal activity of the neural network [27].
Considering the importance of neuronal diversity in the nervous system, the effect of the intrinsic electrical properties of different neuronal types on the propagation of spiking activity in an FFN is addressed in the current work. This paper is organized as follows. In Section 2, the neuronal model, the FFN connections and the simulation methods are introduced. Simulation results are shown in Section 3. The conclusions and discussion are presented in Section 4.
Section snippets
Model and method
Incorporating the biological dynamics of Hodgkin–Huxley neuron and the computational efficiency of integration and firing neuron, Eugene M. Izhikevich developed a simple spiking neuron model and its dynamical equations are given as [28],with the auxiliary after spike resetting:here, v is the membrane potential and u is a membrane recovery variable representing ionic currents of inactivation Na+ and activation K+. The variable
Simulation results
Since the basic neuronal dynamics and the FFN structure have been introduced, we first study the spiking activities propagating in the FFN with a single type of neuron and then investigate the propagation in the FFN with a mixture of neuronal types.
Conclusion and discussion
The mammalian nervous system is a laminar structure containing many areas and neuronal types that form a complex neuronal network [31]. Information propagates within and across the many networks, which could be easily simulated through an FFN [11]. There are many anatomical pieces of evidence for the rationality of FFN, such as the sequential brainstem nuclei in sensory pathways or the cortical layers through which thalamic inputs from layer IV propagate [32]. In the last decade, several
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
This work was supported by the Weigao Young Talent Launch Project.
Hongfang Tan received her B.S. and the M.S. degree in theoretical physics from the Shaanxi normal university, Xi’an, China. She is now teaching in the physics department of Weigao school, Weinan, China. Her current research interests include nonlinear dynamics, complex network, and computational neuroscience.
References (47)
- et al.
Epigenomic signatures of neuronal diversity in the mammalian brain
Neuron
(2015) - et al.
Synaptic excitability of the burst firing neurons in cat sensorimotor cortex in vitro
Brain Res.
(1999) - et al.
Propagation of firing rate by synchronization in a feed-forward multilayer HindmarshRose neural network
Neurocomputing
(2018) - et al.
Decoding neuronal diversity in the developing cerebral cortex: from single cells to functional networks
Curr. Opin. Neurobiol.
(2018) Firing dynamics in hybrid coupled populations of bistable neurons
Neurocomputing
(2019)- et al.
Synchronization-induced spike termination in networks of bistable neurons
Neural Networks
(2019) - et al.
Frequency-difference-dependent stochastic resonance in neural systems
Phys. Rev. E
(2017) - et al.
Theta-gamma-modulated synaptic currents in hippocampal granule cells invivo define a mechanism for network oscillations
Neuron
(2014) First node of ranvier facilitates high-frequency burst encoding
Neuron
(2011)- et al.
A method for detecting high-frequency oscillations using semi-supervised k-means and mean shift clustering
Neurocomputing
(2019)
High-frequency (600 Hz) population spikes in human EEG delineate thalamic and cortical fMRI activation sites
NeuroImage
Dynamic origin of spike and wave discharges in the brain
NeuroImage
The cerebellar mossy fiber synapse as a model for high-frequency transmission in the mammalian CNS
Trends Neurosci.
Ultrafast action potentials mediate kilohertz signaling at a central synapse
Neuron
Generating neuronal diversity in the mammalian cerebral cortex
Annu. Rev. Cell Dev. Biol.
Intrinsic firing patterns of diverse neocortical neurons
Trends Neurosci.
Simple model of spiking neurons
IEEE Trans. Neural Networks
Target-specific output patterns are predicted by the distribution of regular-spiking and bursting pyramidal neurons in the subiculum
Hippocampus
Automated high-throughput characterization of single neurons by means of simplified spiking models
PLoS Comput. Biol.
Beta-rhythm oscillations and synchronization transition in network models of Izhikevich neurons: effect of topology and synaptic type
Front. Comput. Neurosci.
Autapse-induced firing patterns transitions in the Morris-Lecar neuron model
Nonlinear Dyn.
Modular and hierarchically modular organization of brain networks
Front. Neurosci.
Spiking activity propagation in neuronal networks: reconciling different perspectives on neural coding
Nat. Rev. Neurosci.
Cited by (1)
Hongfang Tan received her B.S. and the M.S. degree in theoretical physics from the Shaanxi normal university, Xi’an, China. She is now teaching in the physics department of Weigao school, Weinan, China. Her current research interests include nonlinear dynamics, complex network, and computational neuroscience.
Liqiang Wang received the Ph.D. degree in theoretical physics from the Shaanxi normal university, Xi’an, China, in 2016. He is currently an Associate Professor in School of Science, Xi’an University of Posts and Telecommunications, Xi’an, China. His current research interests include nonlinear dynamics, complex systems, and machine learning.