Jack Murdoch Moore
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Jack Murdoch Moore has a total of 38 co-authors. Here are the top co-authors by number of shared publications:
- Gang Yan
Tongji University
12 shared publications - Michael Small
University of Western Australia
5 shared publications - Xiaolei Ru
Tongji University
5 shared publications - Haiying Wang
University of Shanghai for Science and Technology
4 shared publications - Huijie Yang
Business School, University of Shanghai for Science and Technology, Shanghai 200093, China
4 shared publications - Changgui Gu
University of Shanghai for Science and Technology
3 shared publications
Latest Publications
- Open Access
- 28 October 2026
- Published by American Physical Society (APS) in Physical Review E
- Vol. 114 (4) , 044327
Constrained surrogates for arbitrary families of continuous probability distributions
Empirical regularities described by parametrized families of probability distributions can shape how complex systems are interpreted and explained. A recurring problem is to determine the extent to which these descriptions are supported by observations, and how they can be used to characterize the underlying system. Here we introduce a general framework for constructing constrained surrogates capable of representing arbitrary families of continuous distributions. We demonstrate its utility by providing explicit algorithms for several canonical distributions in statistical physics, including continuous power-law, lognormal, exponential, and truncated Gaussian distributions. These constrained surrogates enable unbiased estimation and accurate tests of model validity based on arbitrary (physically meaningful) sample statistics, without requiring specification or estimation of model parameters. We show how these theoretical advantages translate into more informative and nuanced conclusions in applications ranging from bushfire magnitudes and word usage in texts to brain connectomes and city populations.
- 28 October 2026
- Published by American Physical Society (APS) in Physical Review E
- Vol. 114 (4) , 044327
Open AccessConstrained surrogates for arbitrary families of continuous probability distributions
Empirical regularities described by parametrized families of probability distributions can shape how complex systems are interpreted and explained. A recurring problem is to determine the extent to which these descriptions are supported by observations, and how they can be used to characterize the underlying system. Here we introduce a general framework for constructing constrained surrogates capable of representing arbitrary families of continuous distributions. We demonstrate its utility by providing explicit algorithms for several canonical distributions in statistical physics, including continuous power-law, lognormal, exponential, and truncated Gaussian distributions. These constrained surrogates enable unbiased estimation and accurate tests of model validity based on arbitrary (physically meaningful) sample statistics, without requiring specification or estimation of model parameters. We show how these theoretical advantages translate into more informative and nuanced conclusions in applications ranging from bushfire magnitudes and word usage in texts to brain connectomes and city populations.
- Open Access
- 1 October 2026
- Published by AIP Publishing in Chaos: An Interdisciplinary Journal of Nonlinear Science
- Vol. 36 (10)
Inferring stochastic dynamics from incomplete observations of network spreading processes
Inferring the dynamics of infectious diseases enables prediction and intervention. During early monitoring, available information may comprise digital contact data that partly resolve the interaction network and intermittent screening data that leave the states of some nodes unresolved. Learning-based methods can introduce further information loss by compressing available network and state information into summary statistics. To retain the correspondence between node states and network positions, we develop Graph-Sequence Transformer Attention (GTA), which uses observed node states, network topology, and observation times to infer infection and recovery probabilities for susceptible–infectious–recovered spreading. We compare GTA with summary-based XGBoost and a one-dimensional convolutional neural network and with maximum-likelihood estimation of the recovery probability. Across four synthetic network classes and five empirical topologies, the three learning methods often perform comparably and respond similarly to changes in network size, mean degree, and observation timing. GTA is usually most accurate for estimating infection on sparse, heterogeneous networks, whereas summary-based methods often match or outperform it on denser, more homogeneous networks. For a fixed number of observations, wider spacing can initially improve inference by covering more of the spreading trajectory, but eventually increases error as intermediate transitions are missed. Maximum-likelihood estimation can outperform learning methods when recovery transitions are well resolved, but deteriorates as observations become more widely separated. Losing information about node states can compromise performance more than losing information about edges. Overall, the results show how inference from incomplete observations depends on network structure, temporal sampling, and the information retained in the observations.
- 1 October 2026
- Published by AIP Publishing in Chaos: An Interdisciplinary Journal of Nonlinear Science
- Vol. 36 (10)
Open AccessInferring stochastic dynamics from incomplete observations of network spreading processes
Inferring the dynamics of infectious diseases enables prediction and intervention. During early monitoring, available information may comprise digital contact data that partly resolve the interaction network and intermittent screening data that leave the states of some nodes unresolved. Learning-based methods can introduce further information loss by compressing available network and state information into summary statistics. To retain the correspondence between node states and network positions, we develop Graph-Sequence Transformer Attention (GTA), which uses observed node states, network topology, and observation times to infer infection and recovery probabilities for susceptible–infectious–recovered spreading. We compare GTA with summary-based XGBoost and a one-dimensional convolutional neural network and with maximum-likelihood estimation of the recovery probability. Across four synthetic network classes and five empirical topologies, the three learning methods often perform comparably and respond similarly to changes in network size, mean degree, and observation timing. GTA is usually most accurate for estimating infection on sparse, heterogeneous networks, whereas summary-based methods often match or outperform it on denser, more homogeneous networks. For a fixed number of observations, wider spacing can initially improve inference by covering more of the spreading trajectory, but eventually increases error as intermediate transitions are missed. Maximum-likelihood estimation can outperform learning methods when recovery transitions are well resolved, but deteriorates as observations become more widely separated. Losing information about node states can compromise performance more than losing information about edges. Overall, the results show how inference from incomplete observations depends on network structure, temporal sampling, and the information retained in the observations.
- Open Access
- 29 May 2026
- Published by American Physical Society (APS) in Physical Review Research
- Vol. 8 (2)
Herglotz Lagrangian neural networks: Learning dissipative dynamics via contact variational principles
Physics-informed machine learning has proven effective for learning conservative dynamical systems. However, most real-world systems are nonconservative, and current approaches typically handle dissipation through modifications or require access to full-state information. Here, we introduce the discrete Herglotz Lagrangian neural network (HLNN), a structure-preserving framework for learning a class of smooth dissipative dynamics from data. Grounded in the Herglotz variational principle, HLNN extends classical variational learning to dissipative systems by modeling dissipation intrinsically within a contact-geometric framework, without prescribing an explicit empirical dissipation law, and can be trained from position-only observations without explicit velocity supervision. Across benchmarks spanning state-dependent damping, time-modulated damping, and external driving, HLNN reliably recovers long-term trajectories and dissipation rates, outperforming baseline models on the considered benchmarks. Our results establish a unified variational and geometric framework for learning smooth dissipative dynamics and open avenues for data-driven modeling of nonequilibrium and open systems.
- 29 May 2026
- Published by American Physical Society (APS) in Physical Review Research
- Vol. 8 (2)
Open AccessHerglotz Lagrangian neural networks: Learning dissipative dynamics via contact variational principles
Physics-informed machine learning has proven effective for learning conservative dynamical systems. However, most real-world systems are nonconservative, and current approaches typically handle dissipation through modifications or require access to full-state information. Here, we introduce the discrete Herglotz Lagrangian neural network (HLNN), a structure-preserving framework for learning a class of smooth dissipative dynamics from data. Grounded in the Herglotz variational principle, HLNN extends classical variational learning to dissipative systems by modeling dissipation intrinsically within a contact-geometric framework, without prescribing an explicit empirical dissipation law, and can be trained from position-only observations without explicit velocity supervision. Across benchmarks spanning state-dependent damping, time-modulated damping, and external driving, HLNN reliably recovers long-term trajectories and dissipation rates, outperforming baseline models on the considered benchmarks. Our results establish a unified variational and geometric framework for learning smooth dissipative dynamics and open avenues for data-driven modeling of nonequilibrium and open systems.
- 31 March 2026
- Published by IOP Publishing in Physica Scripta
- Vol. 101 (15)
Adaptive neural network control to achieve bipartite consensus in multiple uncertain Euler-Lagrange systems with time-varying delays
Bipartite consensus in networks of multiple uncertain Euler-Lagrange systems plays a vital role in collective physical tasks such as cooperative robotic manipulation and distributed formation control. However, many existing control strategies assume perfect global knowledge of the system and environment, and often lack robustness against external disturbances and antagonistic interactions, especially under time-varying communication delays. To overcome these challenges, this paper introduces an adaptive control scheme based on neural network approximation. The proposed approach incorporates sliding mode techniques for both leaderless and leader-follower architectures to characterize information exchange among agents, while neural networks are integrated with robust control elements to systematically compensate for uncertainties and disturbances. A rigorous stability analysis of the closed-loop system is conducted by combining Lyapunov theory with the Halanay inequality for delayed dynamical systems. Numerical simulations confirm the efficacy and performance of the proposed method, supporting the theoretical findings.
- 31 March 2026
- Published by IOP Publishing in Physica Scripta
- Vol. 101 (15)
Adaptive neural network control to achieve bipartite consensus in multiple uncertain Euler-Lagrange systems with time-varying delays
Bipartite consensus in networks of multiple uncertain Euler-Lagrange systems plays a vital role in collective physical tasks such as cooperative robotic manipulation and distributed formation control. However, many existing control strategies assume perfect global knowledge of the system and environment, and often lack robustness against external disturbances and antagonistic interactions, especially under time-varying communication delays. To overcome these challenges, this paper introduces an adaptive control scheme based on neural network approximation. The proposed approach incorporates sliding mode techniques for both leaderless and leader-follower architectures to characterize information exchange among agents, while neural networks are integrated with robust control elements to systematically compensate for uncertainties and disturbances. A rigorous stability analysis of the closed-loop system is conducted by combining Lyapunov theory with the Halanay inequality for delayed dynamical systems. Numerical simulations confirm the efficacy and performance of the proposed method, supporting the theoretical findings.
- Open Access
- 1 December 2025
- Published by IOP Publishing in Journal of Physics: Complexity
- Vol. 6 (4) , 045014
Coarse-graining complex networks by optimizing mutual information estimates of random walks
We leverage information theory to develop a novel method of structural coarse-graining for unfolding complex networks across multiple scales, enabling further investigations into multi-scale phenomena. The approach utilizes slow dynamics of random walks on graphs, in which the short-term behavior manifests as movement between individual nodes whereas the long-term behavior, governed by the slow-timescale features, comprises moving between clusters. Specifically, the slow-timescale features are obtained by embedding the walk trajectory into a hidden space to maximize mutual information between historical observations separated by a given time-span. The coarse-grained structures are obtained through aggregating nodes in this hidden space, and exhibit more macroscopic properties as the time-span increases. Our method provides greater control over the resolution of the resultant network. Our results demonstrate that the proposed approach (a) clearly reveals hierarchical community structures in scientific collaboration networks, and (b) illuminates structural and dynamical self-similarity in the human connectome. In addition, we derive an analytic formula for mutual information between the current and historical state of a random walk on a given complex network, which allows us to demonstrate the superiority of neural estimators in determining the mutual information between discrete time series. As well as providing a flexible tool for structural coarse-graining, our work illustrates the practical advantages of neural network-based estimators of information theoretic quantities for addressing problems in network science.
- 1 December 2025
- Published by IOP Publishing in Journal of Physics: Complexity
- Vol. 6 (4) , 045014
Open AccessCoarse-graining complex networks by optimizing mutual information estimates of random walks
We leverage information theory to develop a novel method of structural coarse-graining for unfolding complex networks across multiple scales, enabling further investigations into multi-scale phenomena. The approach utilizes slow dynamics of random walks on graphs, in which the short-term behavior manifests as movement between individual nodes whereas the long-term behavior, governed by the slow-timescale features, comprises moving between clusters. Specifically, the slow-timescale features are obtained by embedding the walk trajectory into a hidden space to maximize mutual information between historical observations separated by a given time-span. The coarse-grained structures are obtained through aggregating nodes in this hidden space, and exhibit more macroscopic properties as the time-span increases. Our method provides greater control over the resolution of the resultant network. Our results demonstrate that the proposed approach (a) clearly reveals hierarchical community structures in scientific collaboration networks, and (b) illuminates structural and dynamical self-similarity in the human connectome. In addition, we derive an analytic formula for mutual information between the current and historical state of a random walk on a given complex network, which allows us to demonstrate the superiority of neural estimators in determining the mutual information between discrete time series. As well as providing a flexible tool for structural coarse-graining, our work illustrates the practical advantages of neural network-based estimators of information theoretic quantities for addressing problems in network science.
- 1 November 2025
- Published by AIP Publishing in Chaos: An Interdisciplinary Journal of Nonlinear Science
- Vol. 35 (11)
Asymmetric information and epidemic coupled dynamics regulated by heterogeneous dynamic behaviors in multilayer networks
To investigate the complex interactions among epidemic spreading, multi-information diffusion, and individual behaviors in multiplex networks, we propose a novel coupled awareness–epidemic spreading model that integrates three key components: three types of awareness driven by multi-information, i.e., proactive protective awareness, negligent protective awareness, and global information-induced awareness; two types of heterogeneous local perception rates that capture individual differences in awareness acquisition and risk perception; and a threshold-triggered iterative accumulation mechanism that models the dynamic reinforcement of protective behaviors when local epidemic risks exceed the critical threshold θ2. These mechanisms enable time-varying heterogeneous awareness propagation under global-local coupling and adaptive disease transmission regulated by heterogeneous perception and iterative behavioral accumulation. Using a microscopic Markov chain approach, we theoretically determine the epidemic threshold βc. Results indicate that global information dissemination more effectively suppresses epidemic spread than the behavioral iteration accumulation mechanism. The global information trigger threshold θ1 induces a two-stage phase transition in βc, whereas the behavioral iteration trigger threshold θ2 generates a three-stage transition, particularly pronounced at high θ1. The initial stress-response intensity α1 positively correlates with βc, suppressing the epidemic more effectively when θ1 is low. Proactive awareness significantly reduces infection risk, while negligent awareness exhibits more complex nonlinear effects due to competitive interactions. Under low public awareness, behavioral response decay plays a critical role in epidemic dynamics.
- 1 November 2025
- Published by AIP Publishing in Chaos: An Interdisciplinary Journal of Nonlinear Science
- Vol. 35 (11)
Asymmetric information and epidemic coupled dynamics regulated by heterogeneous dynamic behaviors in multilayer networks
To investigate the complex interactions among epidemic spreading, multi-information diffusion, and individual behaviors in multiplex networks, we propose a novel coupled awareness–epidemic spreading model that integrates three key components: three types of awareness driven by multi-information, i.e., proactive protective awareness, negligent protective awareness, and global information-induced awareness; two types of heterogeneous local perception rates that capture individual differences in awareness acquisition and risk perception; and a threshold-triggered iterative accumulation mechanism that models the dynamic reinforcement of protective behaviors when local epidemic risks exceed the critical threshold θ2. These mechanisms enable time-varying heterogeneous awareness propagation under global-local coupling and adaptive disease transmission regulated by heterogeneous perception and iterative behavioral accumulation. Using a microscopic Markov chain approach, we theoretically determine the epidemic threshold βc. Results indicate that global information dissemination more effectively suppresses epidemic spread than the behavioral iteration accumulation mechanism. The global information trigger threshold θ1 induces a two-stage phase transition in βc, whereas the behavioral iteration trigger threshold θ2 generates a three-stage transition, particularly pronounced at high θ1. The initial stress-response intensity α1 positively correlates with βc, suppressing the epidemic more effectively when θ1 is low. Proactive awareness significantly reduces infection risk, while negligent awareness exhibits more complex nonlinear effects due to competitive interactions. Under low public awareness, behavioral response decay plays a critical role in epidemic dynamics.
- Open Access
- 23 December 2024
- Published by Oxford University Press (OUP) in PNAS Nexus
- Vol. 4 (1) , pgae580
Brain-inspired wiring economics for artificial neural networks
Wiring patterns of brain networks embody a trade-off between information transmission, geometric constraints, and metabolic cost, all of which must be balanced to meet functional needs. Geometry and wiring economy are crucial in the development of brains, but their impact on artificial neural networks (ANNs) remains little understood. Here, we adopt a wiring cost-controlled training framework that simultaneously optimizes wiring efficiency and task performance during structural evolution of sparse ANNs whose nodes are located at arbitrary but fixed positions. We show that wiring cost control improves performance across a wide range of tasks, ANN architectures and training methods, and can promote task-specific structural modules. An optimal wiring cost range provides both enhanced predictive performance and high values of topological properties, such as modularity and clustering, which are observed in real brain networks and known to improve robustness, interpretability, and performance of ANNs. In addition, ANNs trained using wiring cost can emulate the connection distance distribution observed in the brains of real organisms (such as Ciona intestinalis and Caenorhabditis elegans), especially when achieving high task performance, offering insights into biological organizing principles. Our results shed light on the relationship between topology and task specialization of ANNs trained within biophysical constraints, and their geometric resemblance to real neuronal-level brain maps.
- 23 December 2024
- Published by Oxford University Press (OUP) in PNAS Nexus
- Vol. 4 (1) , pgae580
Open AccessBrain-inspired wiring economics for artificial neural networks
Wiring patterns of brain networks embody a trade-off between information transmission, geometric constraints, and metabolic cost, all of which must be balanced to meet functional needs. Geometry and wiring economy are crucial in the development of brains, but their impact on artificial neural networks (ANNs) remains little understood. Here, we adopt a wiring cost-controlled training framework that simultaneously optimizes wiring efficiency and task performance during structural evolution of sparse ANNs whose nodes are located at arbitrary but fixed positions. We show that wiring cost control improves performance across a wide range of tasks, ANN architectures and training methods, and can promote task-specific structural modules. An optimal wiring cost range provides both enhanced predictive performance and high values of topological properties, such as modularity and clustering, which are observed in real brain networks and known to improve robustness, interpretability, and performance of ANNs. In addition, ANNs trained using wiring cost can emulate the connection distance distribution observed in the brains of real organisms (such as Ciona intestinalis and Caenorhabditis elegans), especially when achieving high task performance, offering insights into biological organizing principles. Our results shed light on the relationship between topology and task specialization of ANNs trained within biophysical constraints, and their geometric resemblance to real neuronal-level brain maps.
- Open Access
- 9 December 2024
- Published by American Physical Society (APS) in Physical Review E
- Vol. 110 (6) , 064311
Susceptible-infected-recovered-susceptible processes competing on simplicial complexes
Contagions do not exist in isolation but in the presence of other propagating quantities, with which they may compete. Furthermore, higher-order interactions (interactions involving more than two nodes) are widely represented in real-world systems. We propose a stochastic model for susceptible-infected-recovered-susceptible (SIRS) processes competing in a simplicial complex that accommodates higher-order interactions. We also propose deterministic microscopic Markov chain (MMC) and mean-field (MF) forms of this model, which we analyze to reveal conditions for the persistence of each infection. We verify our analysis via numerical simulations, which also unveil critical mass effects, discontinuous phase transitions, and bistability. The stochastic model exhibits eight distinct classes of dependence of final state on initial conditions. All eight of these classes can be approximately reproduced with MMC, but only seven are seen under MF. The class observed in MMC and approximated in the stochastic model, but missing from MF, involves three distinct steady states for a single set of parameter values. However, in the absence of higher-order interactions MF exhibits an additional class, not observed in MMC or the stochastic model, corresponding to an infinite spectrum of steady states, in which the ratio of infection levels remains at its initial level. Both MMC and MF models can exhibit sustained oscillations in which one contagion disappears while the other tends to a limit cycle. Published by the American Physical Society 2024
- 9 December 2024
- Published by American Physical Society (APS) in Physical Review E
- Vol. 110 (6) , 064311
Open AccessSusceptible-infected-recovered-susceptible processes competing on simplicial complexes
Contagions do not exist in isolation but in the presence of other propagating quantities, with which they may compete. Furthermore, higher-order interactions (interactions involving more than two nodes) are widely represented in real-world systems. We propose a stochastic model for susceptible-infected-recovered-susceptible (SIRS) processes competing in a simplicial complex that accommodates higher-order interactions. We also propose deterministic microscopic Markov chain (MMC) and mean-field (MF) forms of this model, which we analyze to reveal conditions for the persistence of each infection. We verify our analysis via numerical simulations, which also unveil critical mass effects, discontinuous phase transitions, and bistability. The stochastic model exhibits eight distinct classes of dependence of final state on initial conditions. All eight of these classes can be approximately reproduced with MMC, but only seven are seen under MF. The class observed in MMC and approximated in the stochastic model, but missing from MF, involves three distinct steady states for a single set of parameter values. However, in the absence of higher-order interactions MF exhibits an additional class, not observed in MMC or the stochastic model, corresponding to an infinite spectrum of steady states, in which the ratio of infection levels remains at its initial level. Both MMC and MF models can exhibit sustained oscillations in which one contagion disappears while the other tends to a limit cycle. Published by the American Physical Society 2024
- Open Access
- 21 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network Science and Engineering
- Vol. 11 (6) , 5943-5953
Mixup in Latent Geometry for Graph Classification
Mixup is a data augmentation method which can interpolate between existing data to create new samples. By enlarging the training distribution, it reduces the risk of overfitting and improves generalization. Mixup is relatively straightforward to apply to image samples because pixels with equivalent coordinates in different images can be associated. However, alignment of distinct graphs with different sizes is non-trivial, thereby hindering the application of Mixup to graph data. Here we develop a novel algorithm to address this issue by exploiting the latent hyperbolic geometry which has been shown to underlie many real-world graphs. By considering global graph structure similarity and several fundamental structural features of graph models, we demonstrate that our mixup scheme leads to synthetic graphs whose structural features approximate the linear interpolation of parent graphs, a property important for avoiding the generation of mislabeled synthetic data. We apply the proposed algorithm to classify empirical graphs, and the results show that it improves classification performance on all six benchmark datasets and significantly enhances the generalization ability and robustness of graph neural networks.
- 21 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network Science and Engineering
- Vol. 11 (6) , 5943-5953
Open AccessMixup in Latent Geometry for Graph Classification
Mixup is a data augmentation method which can interpolate between existing data to create new samples. By enlarging the training distribution, it reduces the risk of overfitting and improves generalization. Mixup is relatively straightforward to apply to image samples because pixels with equivalent coordinates in different images can be associated. However, alignment of distinct graphs with different sizes is non-trivial, thereby hindering the application of Mixup to graph data. Here we develop a novel algorithm to address this issue by exploiting the latent hyperbolic geometry which has been shown to underlie many real-world graphs. By considering global graph structure similarity and several fundamental structural features of graph models, we demonstrate that our mixup scheme leads to synthetic graphs whose structural features approximate the linear interpolation of parent graphs, a property important for avoiding the generation of mislabeled synthetic data. We apply the proposed algorithm to classify empirical graphs, and the results show that it improves classification performance on all six benchmark datasets and significantly enhances the generalization ability and robustness of graph neural networks.
- 23 September 2024
- Published by American Physical Society (APS) in Physical Review Letters
- Vol. 133 (13) , 138401
Geometric Scaling Law in Real Neuronal Networks
We investigate the synapse-resolution connectomes of fruit flies across different developmental stages, revealing a consistent scaling law in neuronal connection probability relative to spatial distance. This power-law behavior significantly differs from the exponential distance rule previously observed in coarse-grained brain networks. We demonstrate that the geometric scaling law carries functional significance, aligning with the maximum entropy of information communication and the functional criticality balancing integration and segregation. Perturbing either the empirical probability model’s parameters or its type results in the loss of these advantageous properties. Furthermore, we derive an explicit quantitative predictor for neuronal connectivity, incorporating only interneuronal distance and neurons’ in and out degrees. Our findings establish a direct link between brain geometry and topology, shedding lights on the understanding of how the brain operates optimally within its confined space.
- 23 September 2024
- Published by American Physical Society (APS) in Physical Review Letters
- Vol. 133 (13) , 138401
Geometric Scaling Law in Real Neuronal Networks
We investigate the synapse-resolution connectomes of fruit flies across different developmental stages, revealing a consistent scaling law in neuronal connection probability relative to spatial distance. This power-law behavior significantly differs from the exponential distance rule previously observed in coarse-grained brain networks. We demonstrate that the geometric scaling law carries functional significance, aligning with the maximum entropy of information communication and the functional criticality balancing integration and segregation. Perturbing either the empirical probability model’s parameters or its type results in the loss of these advantageous properties. Furthermore, we derive an explicit quantitative predictor for neuronal connectivity, incorporating only interneuronal distance and neurons’ in and out degrees. Our findings establish a direct link between brain geometry and topology, shedding lights on the understanding of how the brain operates optimally within its confined space.