arXiv subjects

Douglas Zhou

Publications and source records attributed to Douglas Zhou.

At least 19 recordsLinked to original sources

arXiv2025

Overcoming the Curse of Dimensionality: Structural Connectivity Reconstruction via Pairwise Information Flow in Nonlinear Networks

Inferring structural connectivity from observed dynamics remains a fundamental open problem in complex systems, particularly for nonlinear networks where direct measurements are unavailable, and existing methodological approaches each incur characteristic limitations. Model-based methods require prior knowledge of the mechanistic form of the underlying dynamics, while model-free approaches often lack quantitative correspondence to network structural connectivity, and suffer from the curse of dimensionality as the size and complexity of the system increases. Here we show that pairwise time-delayed information flow is sufficient to recover, without high-dimensional conditioning, structural connectivity in general nonlinear networks. We introduce a pairwise delayed information flow (PDIF) as an information-theoretic framework and derive a theoretical quadratic relationship between PDIF and coupling strength, establishing a direct correspondence between information flow and network architecture. We further show that indirect interaction contributions are suppressed at leading order, enabling accurate reconstruction solely from pairwise measurements. Combining binary state representations, pairwise inference, and time-delayed statistics, PDIF overcomes the dimensionality barrier while remaining model-agnostic and scalable. Validated across nonlinear dynamical systems, neuronal network models, and large-scale electrophysiological recordings, PDIF achieves high reconstruction accuracy and robustness to noise, outperforming existing methods. These results establish a principled, efficient and model-agnostic framework for connectivity reconstruction, and reveal a general mechanism by which pairwise observable statistics encode network structure in nonlinear systems.

q-bio.NC↗

arXiv2025

Network structural change point detection and reconstruction for balanced neuronal networks

Understanding brain dynamics and functions critically depends on knowledge of the network connectivity among neurons. However, the complexity of brain structural connectivity, coupled with continuous modifications driven by synaptic plasticity, makes its direct experimental measurement particularly challenging. Conventional connectivity inference methods based on neuronal recordings often assumes a static underlying structural connectivity and requires stable statistical features of neural activities, making them unsuitable for reconstructing structural connectivity that undergoes changes. To fulfill the needs of reconstructing networks undergoing potential structural changes, we propose a unified network reconstruction framework that combines connectivity-induced change point detection (CPD) with pairwise time-delayed correlation coefficient (TDCC) method. For general neuronal networks in balanced regimes, we develop a theoretical analysis for discriminating changes in structural connectivity based on the fluctuation of neuronal voltage time series. We then demonstrate a pairwise TDCC method to reconstruct the network using spike train recordings segmented at the detected change points. We show the effectiveness of our CPD-TDCC network reconstruction using large-scale network simulations with multiple neuronal models. Crucially, our method accommodates networks with changes in both network topologies and synaptic coupling strengths while retaining accuracy even with sparsely sampled subnetwork data, achieving a critical advancement for practical applications in real experimental situations. Our CPD-TDCC framework addresses the critical gap in network reconstruction by accounting connectivity-induced changes points, potentially offering a valuable tool for studying structure and dynamics in the cortical brain.

q-bio.NC↗

arXiv2024

A Generalized Spiking Locally Competitive Algorithm for Multiple Optimization Problems

We introduce a generalized Spiking Locally Competitive Algorithm (LCA) that is biologically plausible and exhibits adaptability to a large variety of neuron models and network connectivity structures. In addition, we provide theoretical evidence demonstrating the algorithm's convergence in optimization problems of signal recovery. Furthermore, our algorithm demonstrates superior performance over traditional optimization methods, such as FISTA, particularly by achieving faster early convergence in practical scenarios including signal denoising, seismic wave detection, and computed tomography reconstruction. Notably, our algorithm is compatible with neuromorphic chips, such as Loihi, facilitating efficient multitasking within the same chip architecture - a capability not present in existing algorithms. These advancements make our generalized Spiking LCA a promising solution for real-world applications, offering significant improvements in execution speed and flexibility for neuromorphic computing systems.

math.OC↗

arXiv2023

Dendritic Integration Based Quadratic Neural Networks Outperform Traditional Aritificial Ones

Incorporating biological neuronal properties into Artificial Neural Networks (ANNs) to enhance computational capabilities poses a formidable challenge in the field of machine learning. Inspired by recent findings indicating that dendrites adhere to quadratic integration rules for synaptic inputs, we propose a novel ANN model, Dendritic Integration-Based Quadratic Neural Network (DIQNN). This model shows superior performance over traditional ANNs in a variety of classification tasks. To reduce the computational cost of DIQNN, we introduce the Low-Rank DIQNN, while we find it can retain the performance of the original DIQNN. We further propose a margin to characterize the generalization error and theoretically prove this margin will increase monotonically during training. And we show the consistency between generalization and our margin using numerical experiments. Finally, by integrating this margin into the loss function, the change of test accuracy is indeed accelerated. Our work contributes a novel, brain-inspired ANN model that surpasses traditional ANNs and provides a theoretical framework to analyze the generalization error in classification tasks.

cs.NE↗

arXiv2023

Mapping Biological Neuron Dynamics into an Interpretable Two-layer Artificial Neural Network

Dendrites are crucial structures for computation of an individual neuron. It has been shown that the dynamics of a biological neuron with dendrites can be approximated by artificial neural networks (ANN) with deep structure. However, it remains unclear whether a neuron can be further captured by a simple, biologically plausible ANN. In this work, we develop a two-layer ANN, named as dendritic bilinear neural network (DBNN), to accurately predict both the sub-threshold voltage and spike time at the soma of biological neuron models with dendritic structure. Our DBNN is found to be interpretable and well captures the dendritic integration process of biological neurons including a bilinear rule revealed in previous works. In addition, we show DBNN is capable of performing diverse tasks including direction selectivity, coincidence detection, and image classification. Our work proposes a biologically interpretable ANN that characterizes the computation of biological neurons, which can be potentially implemented in the deep learning framework to improve computational ability.

q-bio.NC↗

arXiv2021

Quantitative relations among causality measures with applications to nonlinear pulse-output network reconstruction

The causal connectivity of a network is often inferred to understand the network function. It is arguably acknowledged that the inferred causal connectivity relies on causality measure one applies, and it may differ from the network's underlying structural connectivity. However, the interpretation of causal connectivity remains to be fully clarified, in particular, how causal connectivity depends on causality measures and how causal connectivity relates to structural connectivity. Here, we focus on nonlinear networks with pulse signals as measured output, $e.g.$, neural networks with spike output, and address the above issues based on four intensively utilized causality measures, $i.e.$, time-delayed correlation, time-delayed mutual information, Granger causality, and transfer entropy. We theoretically show how these causality measures are related to one another when applied to pulse signals. Taking the simulated Hodgkin-Huxley neural network and the real mouse brain network as two illustrative examples, we further verify the quantitative relations among the four causality measures and demonstrate that the causal connectivity inferred by any of the four well coincides with the underlying network structural connectivity, therefore establishing a direct link between the causal and structural connectivity. We stress that the structural connectivity of networks can be reconstructed pairwisely without conditioning on the global information of all other nodes in a network, thus circumventing the curse of dimensionality. Our framework provides a practical and effective approach for pulse-output network reconstruction.

q-bio.NC↗

arXiv2021

Improved effective linearization of nonlinear Schrödinger waves by increasing nonlinearity

From among the waves whose dynamics are governed by the nonlinear Schrödinger (NLS) equation, we find a robust, spatiotemporally disordered family, in which waves initialized with increasing amplitudes, on average, over long time scales, effectively evolve as ever more weakly coupled collections of plane waves. In particular, the relative amount of energy contained in their coupling decays to zero with increasing wave amplitude.

nlin.PS↗

arXiv2021

Design Fast Algorithms For Hodgkin-Huxley Neuronal Networks

The stiffness of the Hodgkin-Huxley (HH) equations during an action potential (spike) limits the use of large time steps. We observe that the neurons can be evolved independently between spikes, $i.e.,$ different neurons can be evolved with different methods and different time steps. This observation motivates us to design fast algorithms to raise efficiency. We present an adaptive method, an exponential time differencing (ETD) method and a library-based method to deal with the stiff period. All the methods can use time steps one order of magnitude larger than the regular Runge-Kutta methods to raise efficiency while achieving precise statistical properties of the original HH neurons like the largest Lyapunov exponent and mean firing rate. We point out that the ETD and library methods can stably achieve maximum 8 and 10 times of speedup, respectively.

q-bio.NC↗

arXiv2021

Library-based Fast Algorithm for Simulating the Hodgkin-Huxley Neuronal Networks

We present a modified library-based method for simulating the Hodgkin-Huxley (HH) neuronal networks. By pre-computing a high resolution data library during the interval of an action potential (spike), we can avoid evolving the HH equations during the spike and can use a large time step to raise efficiency. The library method can stably achieve at most 10 times of speedup compared with the regular Runge-Kutta method while capturing most statistical properties of HH neurons like the distribution of spikes which data is widely used in the statistical analysis like transfer entropy and Granger causality. The idea of library method can be easily and successfully applied to other HH-type models like the most prominent \textquotedblleft regular spiking\textquotedblright , \textquotedblleft fast spiking\textquotedblright , \textquotedblleft intrinsically bursting\textquotedblright{} and \textquotedblleft low-threshold spike\textquotedblright{} types of HH models.

q-bio.NC↗

arXiv2020

Network Mechanism for Insect Olfaction

Early olfactory pathway responses to the presentation of an odor exhibit remarkably similar dynamical behavior across phyla from insects to mammals, and frequently involve transitions among quiescence, collective network oscillations, and asynchronous firing. We hypothesize that the time scales of fast excitation and fast and slow inhibition present in these networks may be the essential element underlying this similar behavior, and design an idealized, conductance-based integrate-and-fire (I&F) model to verify this hypothesis via numerical simulations. To better understand the mathematical structure underlying the common dynamical behavior across species, we derive a firing-rate (FR) model and use it to extract a slow passage through a saddle-node-on-an-invariant-circle (SNIC) bifurcation structure. We expect this bifurcation structure to provide new insights into the understanding of the dynamical behavior of neuronal assemblies and that a similar structure can be found in other sensory systems.

q-bio.NC↗

arXiv2019

Design Efficient Exponential Time Differencing method For Hodgkin-Huxley Neural Networks

The exponential time differencing (ETD) method allows using a large time step to efficiently evolve the stiff system such as Hodgkin-Huxley (HH) neural networks. For pulse-coupled HH networks, the synaptic spike times cannot be predetermined and are convoluted with neuron's trajectory itself. This presents a challenging issue for the design of an efficient numerical simulation algorithm. The stiffness in the HH equations are quite different between the spike and non-spike regions. Here, we design a second-order adaptive exponential time differencing algorithm (AETD2) for the numerical evolution of HH neural networks. Compared with the regular second-order Runge-Kutta method (RK2), our AETD2 method can use time steps one order of magnitude larger and improve computational efficiency more than ten times while excellently capturing accurate traces of membrane potentials of HH neurons. This high accuracy and efficiency can be robustly obtained and do not depend on the dynamical regimes, connectivity structure or the network size.

q-bio.NC↗

arXiv2019

Asymmetric behavior of surface waves induced by an underlying interfacial wave

We develop a weakly nonlinear model to study the spatiotemporal manifestation and the dynamical behavior of surface waves in the presence of an underlying interfacial solitary wave in a two-layer fluid system. We show that interfacial solitary-wave solutions of this model can capture the ubiquitous broadening of large-amplitude internal waves in the ocean. In addition, the model is capable of capturing three asymmetric behaviors of surface waves: (i) Surface waves become short in wavelength at the leading edge and long at the trailing edge of an underlying interfacial solitary wave. (ii) Surface waves propagate towards the trailing edge with a relatively small group velocity, and towards the leading edge with a relatively large group velocity. (iii) Surface waves become high in amplitude at the leading edge and low at the trailing edge. These asymmetric behaviors can be well quantified in the theoretical framework of ray-based theories. Our model is relatively easily tractable both theoretically and numerically, thus facilitating the understanding of the surface signature of the observed internal waves.

nlin.PS↗

arXiv2019

Modulation-resonance mechanism for surface waves in a two-layer fluid system

We propose a Boussinesq-type model to study the surface/interfacial wave manifestation of an underlying, slowly-varying, long-wavelength, baroclinic flow in a two-layer, density-stratified system. The results of our model show numerically that, under strong nonlinearity, surface waves, with their typical wavenumber being the resonant $k_{\mathrm{res}}$, can be generated locally at the leading edge of the underlying slowly-varying, long-wavelength baroclinic flow. Here, the resonant $k_{\mathrm{res}}$ satisfies the class 3 triad resonance condition among two short-mode waves and one long-mode wave in which all waves propagate in the same direction. Moreover, when the slope of the baroclinic flow is sufficiently small, only one spatially-localized large-amplitude surface wave packet can be generated at the leading edge. This localized surface wave packet becomes high in amplitude and large in group velocity after the interaction with its surrounding waves. These results are qualitatively consistent with various experimental observations including resonant surface waves at the leading edge of an internal wave. Subsequently, we propose a mechanism, referred to as the modulation-resonance mechanism, underlying these surface phenomena, based on our numerical simulations. The proposed modulation-resonance mechanism combines the linear modulation (ray-based) theory for the spatiotemporal asymmetric behavior of surface waves and the nonlinear class 3 triad resonance theory for the energy focusing of surface waves around the resonant wavenumber $k_{\mathrm{res}}$ in Fourier space.

physics.flu-dyn↗

arXiv2019

Digital System Reconstruction by Pairwise Transfer Entropy

Transfer entropy (TE) is an attractive model-free method to detect causality and infer structural connectivity of general digital systems. However it relies on high dimensions used in its definition to clearly remove the memory effect and distinguish the direct causality from the indirect ones which makes it almost inoperable in practice. In this work, we try to use a low order and pairwise TE framework with binary data suitably filtered from the recorded signals to avoid the high dimensional problem. Under this setting, we find and explain that the TE values from the connected and unconnected pairs have a significant difference of magnitude, which can be easily classified by cluster methods. This phenomenon widely and robustly holds over a wide range of systems and dynamical regimes. In addition, we find the TE value is quadratically related to the coupling strength and thus we can establish a quantitative mapping between the causal and structural connectivity.

q-bio.QM↗

arXiv2018

Swift Two-sample Test on High-dimensional Neural Spiking Data

To understand how neural networks process information, it is important to investigate how neural network dynamics varies with respect to different stimuli. One challenging task is to design efficient statistical approaches to analyze multiple spike train data obtained from a short recording time. Based on the development of high-dimensional statistical methods, it is able to deal with data whose dimension is much larger than the sample size. However, these methods often require statistically independent samples to start with, while neural data are correlated over consecutive sampling time bins. We develop an approach to pretreat neural data to become independent samples over time by transferring the correlation of dynamics for each neuron in different sampling time bins into the correlation of dynamics among different dimensions within each sampling time bin. We verify the method using simulation data generated from Integrate-and-fire neuron network models and a large-scale network model of primary visual cortex within a short time, e.g., a few seconds. Our method may offer experimenters to use the advantage of the development of statistical methods to analyze high-dimensional neural data.

q-bio.NC↗

arXiv2018

Maximum Entropy Principle Analysis in Network Systems with Short-time Recordings

In many realistic systems, maximum entropy principle (MEP) analysis provides an effective characterization of the probability distribution of network states. However, to implement the MEP analysis, a sufficiently long-time data recording in general is often required, e.g., hours of spiking recordings of neurons in neuronal networks. The issue of whether the MEP analysis can be successfully applied to network systems with data from short recordings has yet to be fully addressed. In this work, we investigate relationships underlying the probability distributions, moments, and effective interactions in the MEP analysis and then show that, with short recordings of network dynamics, the MEP analysis can be applied to reconstructing probability distributions of network states under the condition of asynchronous activity of nodes in the network. Using spike trains obtained from both Hodgkin-Huxley neuronal networks and electrophysiological experiments, we verify our results and demonstrate that MEP analysis provides a tool to investigate the neuronal population coding properties, even for short recordings.

physics.bio-ph↗

arXiv2018

Dynamical and Coupling Structure of Pulse-Coupled Networks in Maximum Entropy Analysis

Maximum entropy principle (MEP) analysis with few non-zero effective interactions successfully characterizes the distribution of dynamical states of pulse-coupled networks in many experiments, e.g., in neuroscience. To better understand the underlying mechanism, we found a relation between the dynamical structure, i.e., effective interactions in MEP analysis, and the coupling structure of pulse-coupled network to understand how a sparse coupling structure could lead to a sparse coding by effective interactions. This relation quantitatively displays how the dynamical structure is closely related to the coupling structure.

physics.bio-ph↗

arXiv2017

A New Framework for Determination of Excitatory and Inhibitory Conductances Using Somatic Clamp

The interaction between excitation and inhibition is crucial for brain computation. To understand synaptic mechanisms underlying brain function, it is important to separate excitatory and inhibitory inputs to a target neuron. In the traditional method, after applying somatic current or voltage clamp, the excitatory and inhibitory conductances are determined from the synaptic current-voltage (I-V) relation --- the slope corresponds to the total conductance and the intercept corresponds to the reversal current. Because of the space clamp effect, the measured conductance in general deviates substantially from the local conductance on the dendrite. Therefore, the interpretation of the conductance measured by the traditional method remains to be clarified. In this work, based on the investigation of an idealized ball-and-stick neuron model and a biologically realistic pyramidal neuron model, we first demonstrate both analytically and numerically that the conductance determined by the traditional method has no clear biological interpretation due to the neglect of a nonlinear interaction between the clamp current and the synaptic current across the spatial dendrites. As a consequence, the traditional method can induce an arbitrarily large error of conductance measurement, sometimes even leads to unphysically negative conductance. To circumvent the difficulty of elucidating synaptic impact on neuronal computation using the traditional method, we then propose a framework to determine the effective conductance that reflects directly the functional impact of synaptic inputs on action potential initiation and thereby neuronal information processing. Our framework has been further verified in realistic neuron simulations, thus greatly improves upon the traditional approach by providing a reliable and accurate assessment of the role of synaptic activity in neuronal computation.

q-bio.NC↗