The structural aspects of neural dynamics and information flow

Woo, JunHyuk; Choi, Kiri; Kim, Soon Ho; Han, Kyungreem; Choi, M. Y.

doi:10.31083/j.fbl2701015

Cited by 5 publications

(3 citation statements)

References 65 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For example, a recent experimental study has shown that damping behaviours of the dendritic action potential in the neocortical layer 2/3 pyramidal neuron can perform the linearly non-separable XOR operations [9], which have been implemented using multiple neuron layers and summing junctions [10]. This result supports the development of artificial neural networks consisting of not only simple but also complex neuron models with the capabilities of nonlinear computations [11][12][13][14]; some recent models include the mirror neuron (for MirrorBot) [15] and multimodal neurons in the contrastive language-image pre-training (CLIP) algorithm [16].…”

Section: Introductionmentioning

confidence: 67%

Characterization of dynamics and information processing of integrate-and-fire neuron models

Woo¹,

Kim²,

Han³

et al. 2021

J. Phys. A: Math. Theor.

View full text Add to dashboard Cite

The integrate-and-fire (IF) model is the most widely used simple spiking neuron model in neuromorphic computing as well as in artificial neural network algorithms. Here, we characterize the dynamics and information processing of IF models using computer simulations and information-theoretic approaches. Neural dynamics is analysed by means of the time evolution of axonal spikes and the phase-plane portrait, and the coding efficiency of a neuron is estimated by the ratio of mutual information of input and output spikes for the binary hidden neural state. The exponential IF model exhibits higher similarity to the biophysical model in both neural dynamics and coding, compared to other IF type models. Electronic circuit simulations based on the simulation programme with integrated circuit emphasis reveal the nonlinear current–voltage characteristics of the IF neuron models and criticalities in the neural networks. Relevant information-theoretic measures indicate that the computational capabilities of neuromorphic devices largely depend on the neuron models. Such an approach combined with the analysis of neural dynamics provides a useful tool to investigate underlying dynamics of mathematical neuron models; it is applicable to the design and evaluation of neuromorphic models.

show abstract

Section: Introductionmentioning

confidence: 67%

Characterization of dynamics and information processing of integrate-and-fire neuron models

Woo¹,

Kim²,

Han³

et al. 2021

J. Phys. A: Math. Theor.

View full text Add to dashboard Cite

show abstract

“…Since Ramón y Cajal provided evidence that this general hypothesis was also at work in neurons [2], the problem of neuron classification based on their morphology has been a subject of considerable interest in brain research. Along with a causal relationship for the correlation between the neuronal morphologies and spiking patterns of electrophysiological recordings [3], the morphological detail of a neuron has been suggested as one of the key determinants of physiology and functional differentiation of neurons [4][5][6][7][8], engendering a number of morphology-based classification methods [9][10][11][12][13][14][15][16][17][18][19][20][21][22][23].…”

Section: Introductionmentioning

confidence: 99%

Polymer physics-based classification of neurons

Choi

Kim

Hyeon

2022

Preprint

Self Cite

View full text Add to dashboard Cite

Recognizing that diverse morphologies of neurons are reminiscent of structures of branched polymers, we put forward a principled and systematic way of classifying neurons that employs the ideas of polymer physics. In particular, we use 3D coordinates of individual neurons to calculate the form factor, F(q), a Fourier transform of density-density correlation of particles comprising an object of interest. For a polymer-like object consisting of n monomers, F(q) scales with the wavenumber q as F(q)~ q-D at an intermediate range of q, where D is the fractal dimension or the inverse scaling exponent (D=ν-1) characterizing the geometrical feature (r~ nν) of the object over a length scale r(= 2π/q). F(q) can be used to describe a neuron morphology in terms of its size (Rn) and the extent of branching quantified by D. By assessing the similarity between F(q)s calculated for a set of neurons, we tackle the neuron classification problem. Our F(q)-based classification, applied to publicly available neuron datasets from three different organisms, not only complements other classification methods but also offers a physical picture of how the dendritic and axonal branches of an individual neuron fill the space of dense neural networks inside the brain.

show abstract

Section: Introductionmentioning

confidence: 99%

Polymer Physics-Based Classification of Neurons

2022

Self Cite

View full text Add to dashboard Cite

Recognizing that diverse morphologies of neurons are reminiscent of structures of branched polymers, we put forward a principled and systematic way of classifying neurons that employs the ideas of polymer physics. In particular, we use 3D coordinates of individual neurons, which are accessible in recent neuron reconstruction datasets from electron microscope images. We numerically calculate the form factor, F (q), a Fourier transform of the distance distribution of particles comprising an object of interest, which is routinely measured in scattering experiments to quantitatively characterize the structure of materials. For a polymer-like object consisting of n monomers spanning over a length scale of r, F (q) scales with the wavenumber q(= 2π/r) as F (q) ∼ q −D at an intermediate range of q, where D is the fractal dimension or the inverse scaling exponent (D = ν −1 ) characterizing the geometrical feature (r ∼ n ν ) of the object. F (q) can be used to describe a neuron morphology in terms of its size (R n ) and the extent of branching quantified by D. By defining the distance between F (q)s as a measure of similarity between two neuronal morphologies, we tackle the neuron classification problem. In comparison with other existing classification methods for neuronal morphologies, our F (q)-based classification rests solely on 3D coordinates of neurons with no prior knowledge of morphological features. When applied to publicly available neuron datasets from three different organisms, our method not only complements other methods but also offers a physical picture of how the dendritic and axonal branches of an individual neuron fill the space of dense neural networks inside the brain.

show abstract

The structural aspects of neural dynamics and information flow

Cited by 5 publications

References 65 publications

Characterization of dynamics and information processing of integrate-and-fire neuron models

Characterization of dynamics and information processing of integrate-and-fire neuron models

Polymer physics-based classification of neurons

Polymer Physics-Based Classification of Neurons

Contact Info

Product

Resources

About