2025-02-07 | | Total: 5
Graph generation is a critical yet challenging task as empirical analyses require a deep understanding of complex, non-Euclidean structures. Although diffusion models have recently made significant achievements in graph generation, these models typically adapt from the frameworks designed for image generation, making them ill-suited for capturing the topological properties of graphs. In this work, we propose a novel Higher-order Guided Diffusion (HOG-Diff) model that follows a coarse-to-fine generation curriculum and is guided by higher-order information, enabling the progressive generation of plausible graphs with inherent topological structures. We further prove that our model exhibits a stronger theoretical guarantee than classical diffusion frameworks. Extensive experiments on both molecular and generic graph generation tasks demonstrate that our method consistently outperforms or remains competitive with state-of-the-art baselines. Our code is available at https://github.com/Yiminghh/HOG-Diff.
During the last decade, scholars have studied the behavior of decision-making in newsvendor settings and have identified numerous behavior patterns for deviating from normative behavior. However, there is a dearth of research which have examined the influence of individual heterogeneity on decision-making in newsvendor settings. This study examines the level of construal (Abstract and concrete) using construal level theory (CLT) on performance in newsvendor games. In addition, this study measures the cognitive reflection of individuals using cognitive reflection test (CRT) ex-ante to analyze the true impact of how people construe a problem and its impact on their decision-making.
Aging, understood as the tendency to remain in a given state the longer the persistence time in that state, plays a crucial role in the dynamics of complex systems. In this paper, we explore the influence of aging on coevolution models, that is, models in which the dynamics of the states of the nodes in a complex network is coupled to the dynamics of the structure of the network. In particular we consider the coevolving voter model, and we introduce two versions of this model that include aging effects: the Link Aging Model (LAM) and the Node Aging Model (NAM). In the LAM, aging is associated with the persistence time of a link in the evolving network, while in the NAM, aging is associated with the persistence time of a node in a given state. We show that aging significantly affects the absorbing phase transition of the coevolution voter model, shifting the transition point in opposite directions for the LAM and NAM. We also show that the generic absorbing phase transition can disappear due to aging effects.
This paper presents a versatile model for generating fractal complex networks that closely mirror the properties of real-world systems. By combining features of reverse renormalization and evolving network models, the proposed approach introduces several tunable parameters, offering exceptional flexibility in capturing the diverse topologies and scaling behaviors found in both natural and man-made networks. The model effectively replicates their key characteristics such as fractal dimensions, power-law degree distributions, and densities. Unlike traditional deterministic models, it incorporates stochasticity into the network growth process, overcoming limitations like discontinuities in degree distributions and rigid size constraints. The model's applicability is demonstrated through its ability to reproduce the structural features of real-world fractal networks, including the Internet, the World Wide Web, and co-authorship networks.
The global energy transition relies heavily on the large-scale deployment of PV capacity with deployment targets typically defined for 2050. However, sustaining the PV system beyond 2050 will require continuous renewal. This research explores the overlooked industrial transition from the initial implementation phase to the long-term renewal phase, emphasizing the consequences of this dynamic shift. Using streamlined modelling we estimate the annual production needed to both expand and maintain the global PV system. Our results indicate that PV panel production dynamics during this transition are very sensitive to two key factors: deployment speed and panel lifespan. If deployment occurs over a shorter period than the average panel lifespan, production initially overshoots and exhibits an endogenous damped oscillatory behavior due to a succession of installation and replacement cycles. Conversely, if deployment is more gradual, production increases smoothly before stabilizing at the renewal rate. Given the current deployment scenarios and lifespan estimates, the PV industry is likely to face significant production damped oscillations, up to 60%. These oscillations, corresponding to over/underproduction, are further amplified by the increasingly ambitious energy transition targets, which accelerate deployment rates. Panel lifespan conversly remains a less flexible parameter. This study discusses oscillations from a systemic perspective and how they could exacerbate challenges for the long-term sustainability of PV, including industrial, workforce, economic and geopolitical dimensions. Beyond the case of PV, this study underscores a broader issue in the energy transition: the shift from infrastructure expansion to long-term maintenance through renewal. Addressing this often-overlooked phase is essential for ensuring the sustainability of renewable energy systems beyond 2050.