arXiv subjects

Maksym Bondarenko

Publications and source records attributed to Maksym Bondarenko.

4 recordsLinked to original sources

arXiv2026

Geospatial foundation-model embeddings improve population estimation unevenly across space and scale

Reliable subnational population estimates are essential for applications, yet remain difficult where censuses are sparse, outdated or spatially coarse. Existing population-mapping workflows rely on hand-built geospatial covariates, such as settlement extent, night-time lights, and environmental conditions, which must be assembled and harmonised across scales and geographies. Geospatial foundation models offer an alternative by learning reusable representations of place from more multifaceted and heterogeneous data sources. Here, we benchmark Population Dynamics Foundation Model (PDFM) embeddings against the harmonised geospatial covariates for subnational population estimation in Brazil, Nigeria and the United States. Under geographically structured validation, PDFM increased predictive fit by a median of 20.1% (IQR: 10.0-33.2%, across country-model comparisons) reduction in unexplained variance, and reduced Kullback-Leibler divergence by 23.2% (9.2-26.2%). However, these gains were uneven. PDFM was most advantageous where the geospatial covariates weakly characterised settlement context, such as larger and less-developed subnational areas. Moreover, PDFM performance was scale-coupled with embeddings providing less flexible transfer across spatial aggregations than geospatial covariates. These findings showed that geospatial foundation-model representations of place can improve population estimation in data poor settings, but their benefits break down predictably under spatial scale mismatch, revealing a fundamental limitation of current geospatial AI.

cs.LG↗

arXiv2024

EraseDraw: Learning to Draw Step-by-Step via Erasing Objects from Images

Creative processes such as painting often involve creating different components of an image one by one. Can we build a computational model to perform this task? Prior works often fail by making global changes to the image, inserting objects in unrealistic spatial locations, and generating inaccurate lighting details. We observe that while state-of-the-art models perform poorly on object insertion, they can remove objects and erase the background in natural images very well. Inverting the direction of object removal, we obtain high-quality data for learning to insert objects that are spatially, physically, and optically consistent with the surroundings. With this scalable automatic data generation pipeline, we can create a dataset for learning object insertion, which is used to train our proposed text conditioned diffusion model. Qualitative and quantitative experiments have shown that our model achieves state-of-the-art results in object insertion, particularly for in-the-wild images. We show compelling results on diverse insertion prompts and images across various domains.In addition, we automate iterative insertion by combining our insertion model with beam search guided by CLIP.

cs.CV↗

arXiv2020

The COVID-19 pandemic as experienced by the individual

The ongoing COVID-19 pandemic has progressed with varying degrees of intensity in individual countries, suggesting it is important to analyse factors that vary between them. We study measures of `population-weighted density', which capture density as perceived by a randomly chosen individual. These measures of population density can significantly explain variation in the initial rate of spread of COVID-19 between countries within Europe. However, such measures do not explain differences on a global scale, particularly when considering countries in East Asia, or looking later into the epidemics. Therefore, to control for country-level differences in response to COVID-19 we consider the cross-cultural measure of individualism proposed by Hofstede. This score can significantly explain variation in the size of epidemics across Europe, North America, and East Asia. Using both our measure of population-weighted density and the Hofstede score we can significantly explain half the variation in the current size of epidemics across Europe and North America. By controlling for country-level responses to the virus and population density, our analysis of the global incidence of COVID-19 can help focus attention on epidemic control measures that are effective for individual countries.

physics.soc-ph↗

arXiv2008

Master-modes in 3D turbulent channel flow

Turbulent flow fields can be expanded into a series in a set of basic functions. The terms of such series are often called modes. A master- (or determining) mode set is a subset of these modes, the time history of which uniquely determines the time history of the entire turbulent flow provided that this flow is developed. In the present work the existence of the master-mode-set is demonstrated numerically for turbulent channel flow. The minimal size of a master-mode set and the rate of the process of the recovery of the entire flow from the master-mode set history are estimated. The velocity field corresponding to the minimal master-mode set is found to be a good approximation for mean velocity in the entire flow field. Mean characteristics involving velocity derivatives deviate in a very close vicinity to the wall, while master-mode two-point correlations exhibit unrealistic oscillations. This can be improved by using a larger than minimal master-mode set. The near-wall streaks are found to be contained in the velocity field corresponding to the minimal master-mode set, and the same is true at least for the large-scale part of the longitudinal vorticity structure. A database containing the time history of a master-mode set is demonstrated to be an efficient tool for investigating rare events in turbulent flows. In particular, a travelling-wave-like object was identified on the basis of the analysis of the database. Two master-mode-set databases of the time history of a turbulent channel flow are made available online at http://www.dnsdata.afm.ses.soton.ac.uk/. The services provided include the facility for the code uploaded by a user to be run on the server with an access to the data.

physics.flu-dyn↗