Estimating comparable distances to tipping points across mutualistic systems by scaled recovery rates

Biodiversity-induced opposing shifts of tipping points in mutualistic ecological networks

While biodiversity is recognized as crucial for ecosystem stability, the mechanisms governing its dual role in collapse and restoration dynamics remain unclear. By analyzing ten empirical plant-pollinator mutualistic networks, we uncover a biodiversity paradox: increased biodiversity lowers the collapse threshold while enhancing restoration potential. This counterintuitive phenomenon is quantitatively linked to a significant negative correlation between biodiversity levels and hysteresis loop width. To understand this paradox, we develop a refined degree-weighted mean-field framework, reducing high-dimensional dynamics to a tractable two-dimensional system. By integrating potential landscape theory from nonequilibrium statistical mechanics, we uncover the physical basis of biodiversity-driven threshold shifts. Systematic modulation of mutualistic interaction degrees across stochastic networks further confirms the universal regulatory role of reduced biodiversity in collapse-restoration tipping points. Our findings provide a quantitative framework for predicting ecosystem resilience and optimizing restoration strategies across biodiversity gradients.

Deep learning resilience inference for complex networked systems

Resilience, the ability to maintain fundamental functionality amidst failures and errors, is crucial for complex networked systems. Most analytical approaches rely on predefined equations for node activity dynamics and simplifying assumptions on network topology, limiting their applicability to real-world systems. Here, we propose ResInf, a deep learning framework integrating transformers and graph neural networks to infer resilience directly from observational data. ResInf learns representations of node activity dynamics and network topology without simplifying assumptions, enabling accurate resilience inference and low-dimensional visualization. Experimental results show that ResInf significantly outperforms analytical methods, with an F1-score improvement of up to 41.59% over Gao-Barzel-Barabási framework and 14.32% over spectral dimension reduction. It also generalizes to unseen topologies and dynamics and maintains robust performance despite observational disturbances. Our findings suggest that ResInf addresses an important gap in resilience inference for real-world systems, offering a fresh perspective on incorporating data-driven approaches to complex network modeling.

Quantifying the Resilience of Coal Energy Supply in China Toward Carbon Neutrality

Facing the challenge of achieving the goal of carbon neutrality, China is decoupling the currently close dependence of its economy on coal use. The energy supply and demand decarbonization has substantial influence on the resilience of the coal supply. However, a general understanding of the precise impact of energy decarbonization on the resilience of the coal energy supply is still lacking. Here, from the perspective of network science, we propose a theoretical framework to explore the resilience of the coal market of China. We show that the processes of increasing the connectivity and the competition between the coal enterprises, which are widely believed to improve the resilience of the coal market, can undermine the sustainability of the coal supply. Moreover, our results reveal that the policy of closing small-sized coal mines may not only reduce the safety accidents in the coal production but also improve the resilience of the coal market network. Using our model, we also suggest a few practical policies for minimizing the systemic risk of the coal energy supply.

Network properties determine neural network performance

Machine learning influences numerous aspects of modern society, empowers new technologies, from Alphago to ChatGPT, and increasingly materializes in consumer products such as smartphones and self-driving cars. Despite the vital role and broad applications of artificial neural networks, we lack systematic approaches, such as network science, to understand their underlying mechanism. The difficulty is rooted in many possible model configurations, each with different hyper-parameters and weighted architectures determined by noisy data. We bridge the gap by developing a mathematical framework that maps the neural network's performance to the network characters of the line graph governed by the edge dynamics of stochastic gradient descent differential equations. This framework enables us to derive a neural capacitance metric to universally capture a model's generalization capability on a downstream task and predict model performance using only early training results. The numerical results on 17 pre-trained ImageNet models across five benchmark datasets and one NAS benchmark indicate that our neural capacitance metric is a powerful indicator for model selection based only on early training results and is more efficient than state-of-the-art methods.

Optimal sampling of spatial patterns improves deep learning-based early warning signals of critical transitions

Complex spatio-temporal systems like lakes, forests and climate systems exhibit alternative stable states. In such systems, as the threshold value of the driver is crossed, the system may experience a sudden (discontinuous) transition or smooth (continuous) transition to an undesired steady state. Theories predict that changes in the structure of the underlying spatial patterns precede such transitions. While there has been a large body of research on identifying early warning signals of critical transitions, the problem of forecasting the type of transitions (sudden versus smooth) remains an open challenge. We address this gap by developing an advanced machine learning (ML) toolkit that serves as an early warning indicator of spatio-temporal critical transitions, Spatial Early Warning Signal Network (S-EWSNet). ML models typically resemble a black box and do not allow envisioning what the model learns in discerning the labels. Here, instead of naively relying upon the deep learning model, we let the deep neural network learn the latent features characteristic of transitions via an optimal sampling strategy (OSS) of spatial patterns. The S-EWSNet is trained on data from a stochastic cellular automata model deploying the OSS, providing an early warning indicator of transitions while detecting its type in simulated and empirical samples.

Efficient parameter inference in networked dynamical systems via steady states: A surrogate objective function approach integrating mean-field and nonlinear least squares

In networked dynamical systems, inferring governing parameters is crucial for predicting nodal dynamics, such as gene expression levels, species abundance, or population density. While many parameter estimation techniques rely on time-series data, particularly systems that converge over extreme time ranges, only noisy steady-state data is available, requiring a new approach to infer dynamical parameters from noisy observations of steady states. However, the traditional optimization process is computationally demanding, requiring repeated simulation of coupled ordinary differential equations. To overcome these limitations, we introduce a surrogate objective function that leverages decoupled equations to compute steady states, significantly reducing computational complexity. Furthermore, by optimizing the surrogate objective function, we obtain steady states that more accurately approximate the ground truth than noisy observations and predict future equilibria when topology changes. We empirically demonstrate the effectiveness of the proposed method across ecological, gene regulatory, and epidemic networks. Our approach provides an efficient and effective way to estimate parameters from steady-state data and has the potential to improve predictions in networked dynamical systems.

Resilience of hybrid herbivore-plant-pollinator networks

The concept of network resilience has gained increasing attention in the last few decades owing to its great potential in strengthening and maintaining complex systems. From network-based approaches, researchers have explored resilience of real ecological systems comprising diverse types of interactions, such as mutualism, antagonist, and predation, or mixtures of them. In this paper, we propose a dimension-reduction method for analyzing the resilience of hybrid herbivore-plant-pollinator networks. We qualitatively evaluate the contribution of species toward maintaining resilience of networked systems, as well as the distinct roles played by different categories of species. Our findings demonstrate that the strong contributors to network resilience within each category are more vulnerable to extinction. Notably, among the three types of species in consideration, plants exhibit a higher likelihood of extinction, compared to pollinators and herbivores.