arXiv subjects

Stanislav Sobolevsky

Publications and source records attributed to Stanislav Sobolevsky.

At least 19 recordsLinked to original sources

arXiv2026

FastODT: A tree-based framework for efficient continual learning

Machine learning models deployed in real-world settings must operate under evolving data distributions and constrained computational resources. This challenge is particularly acute in non-stationary domains such as energy time series, weather monitoring, and environmental sensing. To remain effective, models must support adaptability, continuous learning, and long-term knowledge retention. This paper introduces a oblivious tree-based model with Hoeffding bound controlling its growth. It seamlessly integrates rapid learning and inference with efficient memory management and robust knowledge preservation, thus allowing for online learning. Extensive experiments across energy and environmental sensing time-series benchmarks demonstrate that the proposed framework achieves performance competitive with, and in several cases surpassing, existing online and batch learning methods, while maintaining superior computational efficiency. Collectively, these results demonstrate that the proposed approach fulfills the core objectives of adaptability, continual updating, and efficient retraining without full model retraining. The framework provides a scalable and resource-aware foundation for deployment in real-world non-stationary environments where resources are constrained and sustained adaptation is essential.

cs.LG↗

arXiv2025

Urban delineation through the lens of commute networks: Leveraging graph embeddings to distinguish socioeconomic groups in cities

Delineating areas within metropolitan regions stands as an important focus among urban researchers, shedding light on the urban perimeters shaped by evolving population dynamics. Applications to urban science are numerous, from facilitating comparisons between delineated districts and administrative divisions to informing policymakers of the shifting economic and labor landscapes. In this study, we propose using commute networks sourced from the census for the purpose of urban delineation, by modeling them with a Graph Neural Network (GNN) architecture. We derive low-dimensional representations of granular urban areas (nodes) using GNNs. Subsequently, nodes' embeddings are clustered to identify spatially cohesive communities in urban areas. Our experiments across the U.S. demonstrate the effectiveness of network embeddings in capturing significant socioeconomic disparities between communities in various cities, particularly in factors such as median household income. The role of census mobility data in regional delineation is also noted, and we establish the utility of GNNs in urban community detection, as a powerful alternative to existing methods in this domain. The results offer insights into the wider effects of commute networks and their use in building meaningful representations of urban regions.

cs.SI↗

arXiv2025

Commute Networks as a Signature of Urban Socioeconomic Performance: Evaluating Mobility Structures with Deep Learning Models

Urban socioeconomic modeling has predominantly concentrated on extensive location and neighborhood-based features, relying on the localized population footprint. However, networks in urban systems are common, and many urban modeling methods don't account for network-based effects. In this study, we propose using commute information records from the census as a reliable and comprehensive source to construct mobility networks across cities. Leveraging deep learning architectures, we employ these commute networks across U.S. metro areas for socioeconomic modeling. We show that mobility network structures provide significant predictive performance without considering any node features. Consequently, we use mobility networks to present a supervised learning framework to model a city's socioeconomic indicator directly, combining Graph Neural Network and Vanilla Neural Network models to learn all parameters in a single learning pipeline. Our experiments in 12 major U.S. cities show the proposed model outperforms previous conventional machine learning models. This work provides urban researchers methods to incorporate network effects in urban modeling and informs stakeholders of wider network-based effects in urban policymaking and planning.

cs.LG↗

arXiv2024

Foundation Models for the Electric Power Grid

Foundation models (FMs) currently dominate news headlines. They employ advanced deep learning architectures to extract structural information autonomously from vast datasets through self-supervision. The resulting rich representations of complex systems and dynamics can be applied to many downstream applications. Therefore, FMs can find uses in electric power grids, challenged by the energy transition and climate change. In this paper, we call for the development of, and state why we believe in, the potential of FMs for electric grids. We highlight their strengths and weaknesses amidst the challenges of a changing grid. We argue that an FM learning from diverse grid data and topologies could unlock transformative capabilities, pioneering a new approach in leveraging AI to redefine how we manage complexity and uncertainty in the electric grid. Finally, we discuss a power grid FM concept, namely GridFM, based on graph neural networks and show how different downstream tasks benefit.

cs.LG↗

arXiv2024

Cities Reconceptualized: Unveiling Hidden Uniform Urban Shape through Commute Flow Modeling in Major US Cities

Urban development is shaped by historical, geographical, and economic factors, presenting challenges for planners in understanding urban form. This study models commute flows across multiple U.S. cities, uncovering consistent patterns in urban population distributions and commuting behaviors. By embedding urban locations to reflect mobility networks, we observe that population distributions across redefined urban spaces tend to approximate log-normal distributions, in contrast to the often irregular distributions found in geographical space. This divergence suggests that natural and historical constraints shape spatial population patterns, while, under ideal conditions, urban organization may naturally align with log-normal distribution. A theoretical model using preferential attachment and random walks supports the emergence of this distribution in urban settings. These findings reveal a fundamental organizing principle in urban systems that, while not always visible geographically, consistently governs population flows and distributions. This insight into the underlying urban structure can inform planners seeking to design efficient, resilient cities.

physics.soc-ph↗

arXiv2023

Subnetwork Constraints for Tighter Upper Bounds and Exact Solution of the Clique Partitioning Problem

We consider a variant of the clustering problem for a complete weighted graph. The aim is to partition the nodes into clusters maximizing the sum of the edge weights within the clusters. This problem is known as the clique partitioning problem, being NP-hard in the general case of having edge weights of different signs. We propose a new method of estimating an upper bound of the objective function that we combine with the classical branch-and-bound technique to find the exact solution. We evaluate our approach on a broad range of random graphs and real-world networks. The proposed approach provided tighter upper bounds and achieved significant convergence speed improvements compared to known alternative methods.

cs.SI↗

arXiv2021

Hierarchical Graph Neural Networks

Over the recent years, Graph Neural Networks have become increasingly popular in network analytic and beyond. With that, their architecture noticeable diverges from the classical multi-layered hierarchical organization of the traditional neural networks. At the same time, many conventional approaches in network science efficiently utilize the hierarchical approaches to account for the hierarchical organization of the networks, and recent works emphasize their critical importance. This paper aims to connect the dots between the traditional Neural Network and the Graph Neural Network architectures as well as the network science approaches, harnessing the power of the hierarchical network organization. A Hierarchical Graph Neural Network architecture is proposed, supplementing the original input network layer with the hierarchy of auxiliary network layers and organizing the computational scheme updating the node features through both - horizontal network connections within each layer as well as the vertical connection between the layers. It enables simultaneous learning of the individual node features along with the aggregated network features at variable resolution and uses them to improve the convergence and stability of the individual node feature learning. The proposed Hierarchical Graph Neural network architecture is successfully evaluated on the network embedding and modeling as well as network classification, node labeling, and community tasks and demonstrates increased efficiency in those.

cs.LG↗

arXiv2021

Recurrent Graph Neural Network Algorithm for Unsupervised Network Community Detection

Network community detection often relies on optimizing partition quality functions, like modularity. This optimization appears to be a complex problem traditionally relying on discrete heuristics. And although the problem could be reformulated as continuous optimization, direct application of the standard optimization methods has limited efficiency in overcoming the numerous local extrema. However, the rise of deep learning and its applications to graphs offers new opportunities. And while graph neural networks have been used for supervised and unsupervised learning on networks, their application to modularity optimization has not been explored yet. This paper proposes a new variant of the recurrent graph neural network algorithm for unsupervised network community detection through modularity optimization. The new algorithm's performance is compared against a popular and fast Louvain method and a more efficient but slower Combo algorithm recently proposed by the author. The approach also serves as a proof-of-concept for the broader application of recurrent graph neural networks to unsupervised network optimization.

cs.SI↗

arXiv2021

Pattern Ensembling for Spatial Trajectory Reconstruction

Digital sensing provides an unprecedented opportunity to assess and understand mobility. However, incompleteness, missing information, possible inaccuracies, and temporal heterogeneity in the geolocation data can undermine its applicability. As mobility patterns are often repeated, we propose a method to use similar trajectory patterns from the local vicinity and probabilistically ensemble them to robustly reconstruct missing or unreliable observations. We evaluate the proposed approach in comparison with traditional functional trajectory interpolation using a case of sea vessel trajectory data provided by The Automatic Identification System (AIS). By effectively leveraging the similarities in real-world trajectories, our pattern ensembling method helps to reconstruct missing trajectory segments of extended length and complex geometry. It can be used for locating mobile objects when temporary unobserved as well as for creating an evenly sampled trajectory interpolation useful for further trajectory mining.

physics.data-an↗

arXiv2020

Transportation Interventions Reshaping NYC Commute: the Probabilistic Simulation Framework Assessing the Impacts of Ridesharing and Manhattan Congestion Surcharge

Understanding holistic impact of planned transportation solutions and interventions on urban systems is challenged by their complexity but critical for decision making. The cornerstone for such impact assessments is estimating the transportation mode-shift resulting from the intervention. And while transportation planning has well-established models for the mode-choice assessment such as the nested multinomial logit model, an individual choice simulation could be better suited for addressing the mode-shift allowing to consistently account for individual preferences. In addition, no model perfectly represents the reality while the available ground truth data on the actual transportation choices needed to infer the model is often incomplete or inconsistent. The present paper addresses those challenges by offering an individual mode-choice and mode-shift simulation model and the Bayesian inference framework. It accounts for uncertainties in the data as well as the model estimate and translates them into uncertainties of the resulting mode-shift and the impacts. The framework is evaluated on the two intervention cases: introducing ride-sharing for-hire-vehicles in NYC as well the recent introduction of the Manhattan Congestion Surcharge. Being successfully evaluated on the cases above, the framework can be used for assessing mode-shift and resulting economic, social and environmental implications for any future urban transportation solutions and policies being considered by decision-makers or transportation companies.

stat.AP↗

arXiv2019

Transfer Learning from an Auxiliary Discriminative Task for Unsupervised Anomaly Detection

Unsupervised anomaly detection from high dimensional data like mobility networks is a challenging task. Study of different approaches of feature engineering from such high dimensional data have been a focus of research in this field. This study aims to investigate the transferability of features learned by network classification to unsupervised anomaly detection. We propose use of an auxiliary classification task to extract features from unlabelled data by supervised learning, which can be used for unsupervised anomaly detection. We validate this approach by designing experiments to detect anomalies in mobility network data from New York and Taipei, and compare the results to traditional unsupervised feature learning approaches of PCA and autoencoders. We find that our feature learning approach yields best anomaly detection performance for both datasets, outperforming other studied approaches. This establishes the utility of this approach to feature engineering, which can be applied to other problems of similar nature.

cs.LG↗

arXiv2019

Pattern and Anomaly Detection in Urban Temporal Networks

Broad spectrum of urban activities including mobility can be modeled as temporal networks evolving over time. Abrupt changes in urban dynamics caused by events such as disruption of civic operations, mass crowd gatherings, holidays and natural disasters are potentially reflected in these temporal mobility networks. Identification and early detecting of such abnormal developments is of critical importance for transportation planning and security. Anomaly detection from high dimensional network data is a challenging task as edge level measurements often have low values and high variance resulting in high noise-to-signal ratio. In this study, we propose a generic three-phase pipeline approach to tackle curse of dimensionality and noisiness of the original data. Our pipeline consists of i) initial network aggregation leveraging community detection ii) unsupervised dimensionality reduction iii) clustering of the resulting representations for outlier detection. We perform extensive experiments to evaluate the proposed approach on mobility data collected from two major cities, New York City and Taipei. Our results empirically prove that proposed methodology outperforms traditional approaches for anomaly detection. We further argue that the proposed anomaly detection framework is potentially generalizable to various other types of temporal networks e.g. social interactions, information propagation and epidemic spread.

physics.soc-ph↗

arXiv2018

On The Painleve Property For A Class Of Quasilinear Partial Differential Equations

The last decades saw growing interest across multiple disciplines in nonlinear phenomena described by partial differential equations (PDE). Integrability of such equations is tightly related with the Painleve property - solutions being free from moveable critical singularities. The problem of Painleve classification of ordinary and partial nonlinear differential equations lasting since the end of XIX century saw significant advances for the equation of lower (mainly up to fourth with rare exceptions) order, however not that much for the equations of higher orders. Recent works of the author have completed the Painleve classification for several broad classes of ordinary differential equations of arbitrary order, advancing the methodology of the Panleve analysis. This paper transfers one of those results on a broad class of nonlinear partial differential equations - quasilinear equations of an arbitrary order three or higher, algebraic in the dependent variable and including only the highest order derivatives of it. Being a first advance in Painleve classification of broad classes of arbitrary order nonlinear PDE's known to the author, this work highlights the potential in building classifications of that kind going beyond specific equations of a limited order, as mainly considered so far.

nlin.SI↗

arXiv2018

Digital Urban Sensing: A Multi-layered Approach

Studies of human mobility increasingly rely on digital sensing, the large-scale recording of human activity facilitated by digital technologies. Questions of variability and population representativity, however, in patterns seen from these sources, remain major challenges for interpreting any outcomes gleaned from these records. The present research explores these questions by providing a comparison of the spatial and temporal activity distributions seen from taxi, subway and Citi Bike trips, mobile app records, geo-tagged Twitter data as well as 311 service requests in the five boroughs of New York City. The comparison reveals substantially different spatial and temporal patterns amongst these datasets, emphasizing limitations in the capacity of individual datasets to represent urban dynamics in their entirety. We further provide interpretations on these differences by decomposing the spatial distributions with working-residential balance and different propensities for demographic groups to use different activities. Nevertheless, the differences also highlight the opportunity to leverage the plurality to create multi-layered models of urban dynamics. We demonstrate the capacity of such models to advance urban zoning and socio-economic modeling - two common applications of digital urban sensing.

physics.soc-ph↗

arXiv2018

House Price Modeling with Digital Census

Urban house prices are strongly associated with local socioeconomic factors. In literature, house price modeling is based on socioeconomic variables from traditional census, which is not real-time, dynamic and comprehensive. Inspired by the emerging concept of "digital census" - using large-scale digital records of human activities to measure urban population dynamics and socioeconomic conditions, we introduce three typical datasets, namely 311 complaints, crime complaints and taxi trips, into house price modeling. Based on the individual housing sales data in New York City, we provide comprehensive evidence that these digital census datasets can substantially improve the modeling performances on both house price levels and changes, regardless whether traditional census is included or not. Hence, digital census can serve as both effective alternatives and complements to traditional census for house price modeling.

econ.EM↗

arXiv2018

Impact Of Bike Sharing In New York City

The Citi Bike deployment changes the landscape of urban mobility in New York City and provides an example of a scalable solution that many other large cities are already adopting around the world. Urban stakeholders who are considering a similar deployment would largely benefit from a quantitative assessment of the impact of bike sharing on urban transportation, as well as associated economic, social and environmental implications. While the Citi Bike usage data is publicly available, the main challenge of such an assessment is to provide an adequate baseline scenario of what would have happened in the city without the Citi Bike system. Existing efforts, including the reports of Citi Bike itself, largely imply arbitrary and often unrealistic assumptions about the alternative transportation mode people would have used otherwise (e.g. by comparing bike trips against driving). The present paper offers a balanced baseline scenario based on a transportation choice model to describe projected customer behavior in the absence of the Citi Bike system. The model also acknowledges the fact that Citi Bike might be used for recreational purposes and, therefore, not all the trips would have been actually performed, if Citi Bike would not be available. The model is trained using open Citi Bike and other urban transportation data and it is applied to assess direct benefits of Citi Bike trips for the end users, as well as for urban stakeholders across different boroughs of New York City and the nearby Jersey City. Besides estimating the travel time and cost savings, the model also reports the associated gas savings, emissions cut and additional exercise for the customers, covering all three areas of anticipated impacts - economic, social and environmental.

physics.soc-ph↗

arXiv2018

Complex network representation through multi-dimensional node projection

Complex network topology might get pretty complicated challenging many network analysis objectives, such as community detection for example. This however makes common emergent network phenomena such as scale-free topology or small-world property even more intriguing. In the present proof-of-concept paper we propose a simple model of network representation inspired by a signal transmission physical analogy, which is apparently capable of reproducing both of the above phenomena. The model appears to be general enough to represent and/or approximate arbitrary complex networks. We propose an approach constructing such a representation by projecting each node into a multi-dimensional space of signal spectrum vectors, where network topology is induced by their overlaps. As one of the implications this enables reducing community detection in complex networks to a straightforward clustering over the projection space, for which multiple efficient approaches are available. We believe such a network representation could turn out to be a useful tool for multiple network analysis objectives.

cs.SI↗

arXiv2017

Inferring hierarchical structure of spatial and generic complex networks through a modeling framework

Our recent paper [Grauwin et al. Sci. Rep. 7 (2017)] demonstrates that community and hierarchical structure of the networks of human interactions largely determines the least and should be taken into account while modeling them. In the present proof-of-concept pre-print the opposite question is considered: could the hierarchical structure itself be inferred to be best aligned with the network model? The inference mechanism is provided for both - spatial networks as well as complex networks in general - through a model based on hierarchical and (if defined) geographical distances. The mechanism allows to discover hierarchical and community structure at any desired resolution in complex networks and in particular - the space-independent structure of the spatial networks. The approach is illustrated on the example of the interstate people migration network in USA.

cs.SI↗