arXiv subjects

Maksim Kitsak

Publications and source records attributed to Maksim Kitsak.

At least 19 recordsLinked to original sources

arXiv2026

Hyperbolic Graph Embedders for Link Prediction and Topology Reconstruction

Hyperbolic embeddings provide compact geometric representations of complex networks in hyperbolic spaces, but systematic comparisons of methods developed in machine learning, network science, and algorithmics remain rare. We benchmark 13 unsupervised hyperbolic graph embedders under a unified protocol for link prediction and topology reconstruction on synthetic and empirical networks. The protocol captures both missing-link recovery and the preservation of local and global network structure. Maximum-likelihood and representation-learning-based approaches, including hybrid variants, achieve the strongest overall performance, although no method dominates across all tasks and structural regimes. Performance is more strongly associated with embedding paradigm than with disciplinary origin. We identify the network regimes in which different paradigms succeed or fail and provide practical guidance for method selection in downstream applications.

cs.LG↗

arXiv2026

Geometric Organization and Inference of Shortest Path Nodes in Soft Random Geometric Graphs

The shortest path problem is related to many dynamic processes on networks, ranging from routing in communication networks to signaling in molecular interaction networks. When the network is fully known, the shortest path problem can be solved precisely and in polynomial time. If, however, the network of interest is only partially observable, the shortest path problem is no longer straightforward. Inspired by the shortest path problem in partially observable networks, we investigate the geometric properties of shortest paths in {\it Euclidean} Soft Random Geometric Graphs (SRGGs). We find that shortest paths are aligned along geodesic curves connecting shortest path endpoints. The strength of the shortest path alignment, as quantified by the average distance to geodesic from shortest path nodes and the average path stretch, is higher for larger SRGGs with short-range connections. In addition, we find that the strength of the shortest path alignment is non-monotonic with respect to the average degree of the SRGG. Based on these observations, we establish the conditions under which the alignment of shortest paths may be sufficiently strong to allow the identification of shortest path nodes based on their proximity to geodesic curves. We show that in partially observable networks with uncertain node positions, our geometric approach can outperform network-based shortest-path algorithms. In practical settings, our findings may have applications to navigation, wireless routing, and flow characterization in infrastructure networks.

physics.soc-ph↗

arXiv2025

Contested logistics: Resilience of strategic highways and railways

Military logistics rely heavily on public infrastructure, such as highways and railways, to transport troops, equipment, and supplies, linking critical installations through the Department of Defense's Strategic Highway Network and Strategic Rail Corridor Network. However, these networks are vulnerable to disruptions that can jeopardize operational readiness, particularly in contested environments where adversaries employ non-traditional threats to disrupt logistics, even within the homeland. This paper presents a contested logistics model that utilizes network science and Geographic Information System (GIS) to evaluate the robustness and resilience of strategic transportation networks under various disruption scenarios. By integrating GIS data to model logistics networks, simulating disruptions, and quantifying their impacts, we identified vulnerabilities in US power projection routes and assessed the resilience and robustness of highways and railways. Our findings reveal that highways are more resilient than railways, with greater capacity to absorb targeted disruptions. These findings underscore the importance of prioritizing investments in highway infrastructure and reinforcing vulnerable road and rail segments, particularly in high-risk regions, to enhance the resilience of military logistics and maintain operational effectiveness in contested conditions.

physics.soc-ph↗

arXiv2025

Connecting Independently Trained Modes via Layer-Wise Connectivity

Empirical studies have shown that continuous low-loss paths can be constructed between independently trained neural network models. This phenomenon, known as mode connectivity, refers to the existence of such paths between distinct modes-i.e., well-trained solutions in parameter space. However, existing empirical methods do not reliably connect independently trained modes and have been evaluated mainly on a narrow set of architectures (e.g., basic CNNs, VGG, and ResNet), leaving their effectiveness on newer models unclear. In this work, we propose a new empirical algorithm for connecting independently trained modes that generalizes beyond traditional architectures and supports a broader range of networks, including MobileNet, ShuffleNet, EfficientNet, RegNet, Deep Layer Aggregation (DLA), and Compact Convolutional Transformers (CCT). In addition to broader applicability, the proposed method yields more consistent connectivity paths across independently trained mode pairs and supports connecting modes obtained with different training hyperparameters.

cs.LG↗

arXiv2024

Vanishing Variance Problem in Fully Decentralized Neural-Network Systems

Federated learning and gossip learning are emerging methodologies designed to mitigate data privacy concerns by retaining training data on client devices and exclusively sharing locally-trained machine learning (ML) models with others. The primary distinction between the two lies in their approach to model aggregation: federated learning employs a centralized parameter server, whereas gossip learning adopts a fully decentralized mechanism, enabling direct model exchanges among nodes. This decentralized nature often positions gossip learning as less efficient compared to federated learning. Both methodologies involve a critical step: computing a representation of received ML models and integrating this representation into the existing model. Conventionally, this representation is derived by averaging the received models, exemplified by the FedAVG algorithm. Our findings suggest that this averaging approach inherently introduces a potential delay in model convergence. We identify the underlying cause and refer to it as the "vanishing variance" problem, where averaging across uncorrelated ML models undermines the optimal variance established by the Xavier weight initialization. Unlike federated learning where the central server ensures model correlation, and unlike traditional gossip learning which circumvents this problem through model partitioning and sampling, our research introduces a variance-corrected model averaging algorithm. This novel algorithm preserves the optimal variance needed during model averaging, irrespective of network topology or non-IID data distributions. Our extensive simulation results demonstrate that our approach enables gossip learning to achieve convergence efficiency comparable to that of federated learning.

cs.LG↗

arXiv2024

Random hyperbolic graphs in $d+1$ dimensions

We consider random hyperbolic graphs in hyperbolic spaces of any dimension $d+1\geq 2$. We present a rescaling of model parameters that casts the random hyperbolic graph model of any dimension to a unified mathematical framework, leaving the degree distribution invariant with respect to the dimension. Unlike the degree distribution, clustering does depend on the dimension, decreasing to 0 at $d \rightarrow \infty$. We analyze all of the other limiting regimes of the model, and we release a software package that generates random hyperbolic graphs and their limits in hyperbolic spaces of any dimension.

physics.soc-ph↗

arXiv2024

Access to Emergency Services: A New York City Case Study

Emergency services play a crucial role in safeguarding human life and property within society. In this paper, we propose a network-based methodology for calculating transportation access between emergency services and the broader community. Using New York City as a case study, this study identifies 'emergency service deserts' based on the National Fire Protection Association (NFPA) guidelines, where accessibility to Fire, Emergency Medical Services, Police, and Hospitals are compromised. The results show that while 95% of NYC residents are well-served by emergency services, the residents of Staten Island are disproportionately underserved. By quantifying the relationship between first responder travel time, Emergency Services Sector (ESS) site density, and population density, we discovered a negative power law relationship between travel time and ESS site density. This relationship can be used directly by policymakers to determine which parts of a community would benefit the most from providing new ESS locations. Furthermore, this methodology can be used to quantify the resilience of emergency service infrastructure by observing changes in accessibility in communities facing threats.

physics.soc-ph↗

arXiv2023

Topological properties and organizing principles of semantic networks

Interpreting natural language is an increasingly important task in computer algorithms due to the growing availability of unstructured textual data. Natural Language Processing (NLP) applications rely on semantic networks for structured knowledge representation. The fundamental properties of semantic networks must be taken into account when designing NLP algorithms, yet they remain to be structurally investigated. We study the properties of semantic networks from ConceptNet, defined by 7 semantic relations from 11 different languages. We find that semantic networks have universal basic properties: they are sparse, highly clustered, and many exhibit power-law degree distributions. Our findings show that the majority of the considered networks are scale-free. Some networks exhibit language-specific properties determined by grammatical rules, for example networks from highly inflected languages, such as e.g. Latin, German, French and Spanish, show peaks in the degree distribution that deviate from a power law. We find that depending on the semantic relation type and the language, the link formation in semantic networks is guided by different principles. In some networks the connections are similarity-based, while in others the connections are more complementarity-based. Finally, we demonstrate how knowledge of similarity and complementarity in semantic networks can improve NLP algorithms in missing link inference.

cs.CL↗

arXiv2023

Reporting delays: a widely neglected impact factor in COVID-19 forecasts

Epidemic forecasts are only as good as the accuracy of epidemic measurements. Is epidemic data, particularly COVID-19 epidemic data, clean and devoid of noise? Common sense implies the negative answer. While we cannot evaluate the cleanliness of the COVID-19 epidemic data in a holistic fashion, we can assess the data for the presence of reporting delays. In our work, through the analysis of the first COVID-19 wave, we find substantial reporting delays in the published epidemic data. Motivated by the desire to enhance epidemic forecasts, we develop a statistical framework to detect, uncover, and remove reporting delays in the infectious, recovered, and deceased epidemic time series. Our framework can uncover and analyze reporting delays in 8 regions significantly affected by the first COVID-19 wave. Further, we demonstrate that removing reporting delays from epidemic data using our statistical framework may decrease the error in epidemic forecasts. While our statistical framework can be used in combination with any epidemic forecast method that intakes infectious, recovered, and deceased data, to make a basic assessment, we employed the classical SIRD epidemic model. Our results indicate that the removal of reporting delays from the epidemic data may decrease the forecast error by up to 50. We anticipate that our framework will be indispensable in the analysis of novel COVID-19 strains and other existing or novel infectious diseases.

q-bio.PE↗

arXiv2023

Complementarity in Complex Networks

In many networks, including networks of protein-protein interactions, interdisciplinary collaboration networks, and semantic networks, connections are established between nodes with complementary rather than similar properties. While complementarity is abundant in networks, we lack mathematical intuition and quantitative methods to study complementarity mechanisms in these systems. In this work, we close this gap by providing a rigorous definition of complementarity and developing geometric complementarity frameworks for modeling and inference tasks on networks. We demonstrate the utility of complementarity frameworks by learning geometric representations of several real systems. Complementarity not only offers novel practical analysis methods but also enhances our intuition about formation mechanisms in networks on a broader scale and calls for a careful re-evaluation of existing similarity-inspired methods.

physics.soc-ph↗

arXiv2022

Finding shortest and nearly shortest path nodes in large substantially incomplete networks

Dynamic processes on networks, be it information transfer in the Internet, contagious spreading in a social network, or neural signaling, take place along shortest or nearly shortest paths. Unfortunately, our maps of most large networks are substantially incomplete due to either the highly dynamic nature of networks, or high cost of network measurements, or both, rendering traditional path finding methods inefficient. We find that shortest paths in large real networks, such as the network of protein-protein interactions (PPI) and the Internet at the autonomous system (AS) level, are not random but are organized according to latent-geometric rules. If nodes of these networks are mapped to points in latent hyperbolic spaces, shortest paths in them align along geodesic curves connecting endpoint nodes. We find that this alignment is sufficiently strong to allow for the identification of shortest path nodes even in the case of substantially incomplete networks. We demonstrate the utility of latent-geometric path-finding in problems of cellular pathway reconstruction and communication security.

physics.soc-ph↗

arXiv2021

Relationship among state reopening policies, health outcomes and economic recovery through first wave of the COVID-19 pandemic in the U.S

State governments in the U.S. have been facing difficult decisions involving tradeoffs between economic and health-related outcomes during the COVID-19 pandemic. Despite evidence of the effectiveness of government-mandated restrictions mitigating the spread of contagion, these orders are stigmatized due to undesirable economic consequences. This tradeoff resulted in state governments employing mandates in widely different ways. We compare the different policies states implemented during periods of restriction (lockdown) and reopening with indicators of COVID-19 spread and consumer card spending at each state during the first wave of the pandemic in the U.S. between March and August 2020. We find that while some states enacted reopening decisions when the incidence rate of COVID-19 was minimal or sustained in its relative decline, other states relaxed socioeconomic restrictions near their highest incidence and prevalence rates experienced so far. Nevertheless, all states experienced similar trends in consumer card spending recovery, which was strongly correlated with reopening policies following the lockdowns and relatively independent from COVID-19 incidence rates at the time. Our findings suggest that consumer card spending patterns can be attributed to government mandates rather than COVID-19 incidence in the states. We estimate the recovery in states that reopened in late April was more than the recovery in states that did not reopen in the same period - 15% for consumer card spending and 18% for spending by high income households. This result highlights the important role of state policies in minimizing health impacts while promoting economic recovery and helps planning effective interventions in subsequent waves and immunization efforts.

econ.GN↗

arXiv2021

Two-population SIR model and strategies to reduce mortality in pandemics

Despite many studies on the transmission mechanism of the Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), it remains still challenging to efficiently reduce mortality. In this work, we apply a two-population Susceptible-Infected-Removed (SIR) model to investigate the COVID-19 spreading when contacts between elderly and non-elderly individuals are reduced due to the high mortality risk of elderly people. We discover that the reduction of connections between two populations can delay the death curve but cannot well reduce the final mortality. We propose a merged SIR model, which advises elderly individuals to interact less with their non-elderly connections at the initial stage but interact more with their non-elderly relationships later, to reduce the final mortality. Finally, immunizing elderly hub individuals can also significantly decrease mortality.

physics.soc-ph↗

arXiv2020

Link prediction with hyperbolic geometry

Link prediction is a paradigmatic problem in network science with a variety of applications. In latent space network models this problem boils down to ranking pairs of nodes in the order of increasing latent distances between them. The network model with hyperbolic latent spaces has a number of attractive properties suggesting it must be a powerful tool to predict links, but the past work in this direction reported mixed results. Here we perform systematic investigation of the utility of latent hyperbolic geometry for link prediction in networks. We first show that some measures of link prediction accuracy are extremely sensitive with respect to inaccuracies in the inference of latent hyperbolic coordinates of nodes, so that we develop a new coordinate inference method that maximizes the accuracy of such inference. Applying this method to synthetic and real networks, we then find that while there exists a multitude of competitive methods to predict obvious easy-to-predict links, among which hyperbolic link prediction is rarely the best but often competitive, it is the best, often by far, when the task is to predict less obvious missing links that are really hard to predict. These links include missing links in incomplete networks with large fractions of missing links, missing links between nodes that do not have any common neighbors, and missing links between dissimilar nodes at large latent distances. Overall these results suggest that the harder a specific link prediction task is, the more seriously one should consider using hyperbolic geometry.

physics.soc-ph↗

arXiv2020

Weighted hypersoft configuration model

Maximum entropy null models of networks come in different flavors that depend on the type of constraints under which entropy is maximized. If the constraints are on degree sequences or distributions, we are dealing with configuration models. If the degree sequence is constrained exactly, the corresponding microcanonical ensemble of random graphs with a given degree sequence is the configuration model per se. If the degree sequence is constrained only on average, the corresponding grand-canonical ensemble of random graphs with a given expected degree sequence is the soft configuration model. If the degree sequence is not fixed at all but randomly drawn from a fixed distribution, the corresponding hypercanonical ensemble of random graphs with a given degree distribution is the hypersoft configuration model, a more adequate description of dynamic real-world networks in which degree sequences are never fixed but degree distributions often stay stable. Here, we introduce the hypersoft configuration model of weighted networks. The main contribution is a particular version of the model with power-law degree and strength distributions, and superlinear scaling of strengths with degrees, mimicking the properties of some real-world networks. As a byproduct, we generalize the notions of sparse graphons and their entropy to weighted networks.

physics.soc-ph↗

arXiv2019

Lack of Resilience in Transportation Networks: Economic Implications

Disruptions to transportation networks are inevitable. Currently, most mandated development-related transportation planning is intended to prepare for frequently occurring and observable disruptions while low probability events that have not yet materialized attract less attention. When road networks are not resilient, these unpredictable events can cause significant delays that may not be proportional to the extent of the disruption. Enhancing resilience can help in mitigating consequences of disruptions but requires financial investment that is difficult to justify given that low probability event may not have materialized. This paper highlights economic implications of unmitigated random disruptions in urban road systems and makes the case for investment in transportation network resilience. We utilized a model of urban transportation network performance that quantifies resilience and demonstrated its integration with microeconomic transportation planning model REMI TranSight. The model was applied to 10 cities in the USA to calculate several economic indicators under baseline scenario where economic impact was assumed to be proportional to the magnitude of disruptive events and under a test scenario where the magnitude of disruption was used to calculate additional delays in transportation networks and these additional delays were explicitly integrated in REMI model. Results show that GDP losses suffered as a result of disruption may be far more significant in the case scenario and economic output is not necessarily back to normal in the year the following disruptive event. We thus conclude that support for investment decisions on network efficiency and resilience management should be based on a framework that utilizes resilience, quantified in terms that are compatible with standard practice, and scenarios to test the implications of topological attributes.

physics.soc-ph↗

arXiv2018

Stability of a Giant Connected Component in a Complex Network

We analyze the stability of the network's giant connected component under impact of adverse events, which we model through the link percolation. Specifically, we quantify the extent to which the largest connected component of a network consists of the same nodes, regardless of the specific set of deactivated links. Our results are intuitive in the case of single-layered systems: the presence of large degree nodes in a single-layered network ensures both its robustness and stability. In contrast, we find that interdependent networks that are robust to adverse events have unstable connected components. Our results bring novel insights to the design of resilient network topologies and the reinforcement of existing networked systems.

physics.soc-ph↗

arXiv2017

Resilience and efficiency in transportation networks

Urban transportation systems are vulnerable to congestion, accidents, weather, special events, and other costly delays. Whereas typical policy responses prioritize reduction of delays under normal conditions to improve the efficiency of urban road systems, analytic support for investments that improve resilience (defined as system recovery from additional disruptions) is still scarce. In this effort, we represent paved roads as a transportation network by mapping intersections to nodes and road segments between the intersections to links. We built road networks for 40 of the urban areas defined by the U.S. Census Bureau. We developed and calibrated a model to evaluate traffic delays using link loads. The loads may be regarded as traffic-based centrality measures, estimating the number of individuals using corresponding road segments. Efficiency was estimated as the average annual delay per peak-period auto commuter, and modeled results were found to be close to observed data, with the notable exception of New York City. Resilience was estimated as the change in efficiency resulting from roadway disruptions and was found to vary between cities, with increased delays due to a 5% random loss of road linkages ranging from 9.5% in Los Angeles to 56.0% in San Francisco. The results demonstrate that many urban road systems that operate inefficiently under normal conditions are nevertheless resilient to disruption, whereas some more efficient cities are more fragile. The implication is that resilience, not just efficiency, should be considered explicitly in roadway project selection and justify investment opportunities related to disaster and other disruptions.

physics.soc-ph↗