arXiv subjects

Péter Pollner

Publications and source records attributed to Péter Pollner.

16 recordsLinked to original sources

arXiv2026

Million-scale multimodal pollen microscopy with expert-guided foundation models

Automated pollen identification from microscopy remains a bottleneck in aerobiology, palaeoecology and biodiversity monitoring, because scalable systems must generalise across specimen preparation, scanner settings and geographic origins while retaining palynological interpretability. To address this gap, we present a million-scale multimodal pollen microscopy resource, Pollen AI Atlas, assembled from pure-species whole-slide bright-field images spanning four geographic origins, four scanner settings, 45 genera, and one family-only taxon across 31 botanical families. Seeded by one manually selected exemplar per source slide, token-level mining and filtering produced 1,511,390 released grain detections with 99.6\% proposal precision in expert-curated test regions. Each detection was paired with machine-generated grain-level morphological captions from five open-weight vision--language models, guided by expert-verified palynological anchors, yielding structured descriptions of aperture systems, wall ornamentation, shape and size. Among the evaluated models, Gemma4 provided the most controlled primary caption set, combining tight length control, no detected taxon-name or numeric-size leakage and the strongest text-retrieval performance. Baseline benchmarks with frozen visual features reached 88.16\% top-1 accuracy, while cross-regional retrieval showed that caption-derived text embeddings remained robust when image similarity degraded (mAP@20 0.811 versus 0.262). Released data, annotations, captions, splits, code, and weights provide a benchmark for pollen recognition, cross-regional domain adaptation and domain-specific multimodal microscopy learning.

cs.CV↗

arXiv2020

Generalized entropies, density of states, and non-extensivity

The concept of entropy connects the number of possible configurations with the number of variables in large stochastic systems. Independent or weakly interacting variables render the number of configurations scale exponentially with the number of variables, making the Boltzmann-Gibbs-Shannon entropy extensive. In systems with strongly interacting variables, or with variables driven by history-dependent dynamics, this is no longer true. Here we show that contrary to the generally held belief, not only strong correlations or history-dependence, but skewed-enough distribution of visiting probabilities, that is, first-order statistics, also play a role in determining the relation between configuration space size and system size, or, equivalently, the extensive form of generalized entropy. We present a macroscopic formalism describing this interplay between first-order statistics, higher-order statistics, and configuration space growth. We demonstrate that knowing any two strongly restricts the possibilities of the third. We believe that this unified macroscopic picture of emergent degrees of freedom constraining mechanisms provides a step towards finding order in the zoo of strongly interacting complex systems.

cond-mat.stat-mech↗

arXiv2019

Time evolution of the hierarchical networks between PubMed MeSH terms

Hierarchical organisation is a prevalent feature of many complex networks appearing in nature and society. A relating interesting, yet less studied question is how does a hierarchical network evolve over time? Here we take a data driven approach and examine the time evolution of the network between the Medical Subject Headings (MeSH) provided by the National Center for Biotechnology Information (NCBI, part of the U. S. National Library of Medicine). The network between the MeSH terms is organised into 16 different, yearly updated hierarchies such as "Anatomy", "Diseases", "Chemicals and Drugs", etc. The natural representation of these hierarchies is given by directed acyclic graphs, composed of links pointing from nodes higher in the hierarchy towards nodes in lower levels. Due to the yearly updates, the structure of these networks is subject to constant evolution: new MeSH terms can appear, terms becoming obsolete can be deleted or be merged with other terms, and also already existing parts of the network may be rewired. We examine various statistical properties of the time evolution, with a special focus on the attachment and detachment mechanisms of the links, and find a few general features that are characteristic for all MeSH hierarchies. According to the results, the hierarchies investigated display an interesting interplay between non-uniform preference with respect to multiple different topological and hierarchical properties.

physics.soc-ph↗

arXiv2019

Generalised thresholding of hidden variable network models with scale-free property

The hidden variable formalism (based on the assumption of some intrinsic node parameters) turned out to be a remarkably efficient and powerful approach in describing and analyzing the topology of complex networks. Owing to one of its most advantageous property - namely proven to be able to reproduce a wide range of different degree distribution forms - it has become a standard tool for generating networks having the scale-free property. One of the most intensively studied version of this model is based on a thresholding mechanism of the exponentially distributed hidden variables associated to the nodes (intrinsic vertex weights), which give rise to the emergence of a scale-free network where the degree distribution $p(k)\sim k^{-γ}$ is decaying with an exponent of $γ=2$. Here we propose a generalization and modification of this model by extending the set of connection probabilities and hidden variable distributions that lead to the aforementioned degree distribution, and analyze the conditions leading to the above behavior analytically. In addition, we propose a relaxation of the hard threshold in the connection probabilities, which opens up the possibility for obtaining sparse scale free networks with arbitrary scaling exponent.

physics.soc-ph↗

arXiv2019

Emergence of leader-follower hierarchy among players in an on-line experiment

Hierarchical networks are prevalent in nature and society, corresponding to groups of actors - animals, humans or even robots - organised according to a pyramidal structure with decision makers at the top and followers at the bottom. While this phenomenon is seemingly universal, the underlying governing principles are poorly understood. Here we study the emergence of hierarchies in groups of people playing a simple dot guessing game in controlled experiments, lasting for about 40 rounds, conducted over the Internet. During the games, the players had the possibility to look at the answer of a limited number of other players of their choice. This act of asking for advice defines a directed connection between the involved players, and according to our analysis, the initial random configuration of the emerging networks became more structured overt time, showing signs of hierarchy towards the end of the game. In addition, the achieved score of the players appeared to be correlated with their position in the hierarchy. These results indicate that under certain conditions imitation and limited knowledge about the performance of other actors is sufficient for the emergence of hierarchy in a social group.

physics.soc-ph↗

arXiv2018

Phase space volume scaling of generalized entropies and anomalous diffusion scaling governed by corresponding non-linear Fokker-Planck equations

Many physical, biological or social systems are governed by history-dependent dynamics or are composed of strongly interacting units, showing an extreme diversity of microscopic behaviour. Macroscopically, however, they can be efficiently modeled by generalizing concepts of the theory of Markovian, ergodic and weakly interacting stochastic processes. In this paper, we model stochastic processes by a family of generalized Fokker-Planck equations whose stationary solutions are equivalent to the maximum entropy distributions according to generalized entropies. We show that at asymptotically large times and volumes, the scaling exponent of the anomalous diffusion process described by the generalized Fokker-Planck equation and the phase space volume scaling exponent of the generalized entropy bijectively determine each other via a simple algebraic relation. This implies that these basic measures characterizing the transient and the stationary behaviour of the processes provide the same information regarding the asymptotic regime, and consequently, the classification of the processes given by these two exponents coincide.

physics.gen-ph↗

arXiv2017

High Quality Queueing Information from Accelerated Active Network Tomography

Monitoring network state can be crucial in Future Internet infrastructures. Passive monitoring of all the routers is expensive and prohibitive. Storing, accessing and sharing the data is a technological challenge among networks with conflicting economic interests. Active monitoring methods can be attractive alternatives as they are free from most of these issues. Here we demonstrate that it is possible to improve the active network tomography methodology to such extent that the quality of the extracted link or router level delay is comparable to the passively measurable information. We show that the temporal precision of the measurements and the performance of the data analysis should be simultaneously improved to achieve this goal. In this paper we not only introduce a new efficient message-passing based algorithm but we also show that it is applicable for data collected by the ETOMIC high precision active measurement infrastructure. The measurements are conducted in the GEANT2 high speed academic network connecting the sites, which is an ideal test ground for such Future Internet applications.

cs.NI↗

arXiv2017

Detecting and classifying lesions in mammograms with Deep Learning

In the last two decades Computer Aided Diagnostics (CAD) systems were developed to help radiologists analyze screening mammograms. The benefits of current CAD technologies appear to be contradictory and they should be improved to be ultimately considered useful. Since 2012 deep convolutional neural networks (CNN) have been a tremendous success in image recognition, reaching human performance. These methods have greatly surpassed the traditional approaches, which are similar to currently used CAD solutions. Deep CNN-s have the potential to revolutionize medical image analysis. We propose a CAD system based on one of the most successful object detection frameworks, Faster R-CNN. The system detects and classifies malignant or benign lesions on a mammogram without any human intervention. The proposed method sets the state of the art classification performance on the public INbreast database, AUC = 0.95 . The approach described here has achieved the 2nd place in the Digital Mammography DREAM Challenge with AUC = 0.85 . When used as a detector, the system reaches high sensitivity with very few false positive marks per image on the INbreast dataset. Source code, the trained model and an OsiriX plugin are availaible online at https://github.com/riblidezso/frcnn_cad .

cs.CV↗

arXiv2016

Comparing the hierarchy of keywords in on-line news portals

The tagging of on-line content with informative keywords is a widespread phenomenon from scientific article repositories through blogs to on-line news portals. In most of the cases, the tags on a given item are free words chosen by the authors independently. Therefore, relations among keywords in a collection of news items is unknown. However, in most cases the topics and concepts described by these keywords are forming a latent hierarchy, with the more general topics and categories at the top, and more specialised ones at the bottom. Here we apply a recent, cooccurrence-based tag hierarchy extraction method to sets of keywords obtained from four different on-line news portals. The resulting hierarchies show substantial differences not just in the topics rendered as important (being at the top of the hierarchy) or of less interest (categorised low in the hierarchy), but also in the underlying network structure. This reveals discrepancies between the plausible keyword association frameworks in the studied news portals.

physics.soc-ph↗

arXiv2016

Partial order similarity based on mutual information

Comparing the ranking of candidates by different voters is an important topic in social and information science with a high relevance from the point of view of practical applications. In general, ties and pairs of incomparable candidates may occur, thus, the alternative rankings are described by partial orders. Various distance measures between partial orders have already been introduced, where zero distance is corresponding to a perfect match between a pair of partial orders, and larger values signal greater differences. Here we take a different approach and propose a similarity measure based on adjusted mutual information. In general, the similarity value of unity is corresponding to exactly matching partial orders, while a low similarity is associated to a pair of independent partial orders. The time complexity of the computation of this similarity measure is $\mathcal{O}(\left|{\mathcal C}\right|^3)$ in the worst case, and $\mathcal{O}(\left|{\mathcal C}\right|^2\ln \left|{\mathcal C}\right|)$ in the typical case of partial orders corresponding to trees with constant branching number, where $\left|{\mathcal C}\right|$ denotes the number of candidates. An interesting feature of our approach is that the similarity measure is sensitive to the position of the disagreements in the ranking: Differences at the highly ranked candidates induce larger similarity drop compared to disagreements at the bottom candidates.

stat.AP↗

arXiv2015

Hierarchical networks of scientific journals

Scientific journals are the repositories of the gradually accumulating knowledge of mankind about the world surrounding us. Just as our knowledge is organised into classes ranging from major disciplines, subjects and fields to increasingly specific topics, journals can also be categorised into groups using various metrics. In addition to the set of topics characteristic for a journal, they can also be ranked regarding their relevance from the point of overall influence. One widespread measure is impact factor, but in the present paper we intend to reconstruct a much more detailed description by studying the hierarchical relations between the journals based on citation data. We use a measure related to the notion of m-reaching centrality and find a network which shows the level of influence of a journal from the point of the direction and efficiency with which information spreads through the network. We can also obtain an alternative network using a suitably modified nested hierarchy extraction method applied to the same data. The results are weakly methodology-dependent and reveal non-trivial relations among journals. The two alternative hierarchies show large similarity with some striking differences, providing together a complex picture of the intricate relations between scientific journals.

physics.soc-ph↗

arXiv2015

Comparing the hierarchy of author given tags and repository given tags in a large document archive

Folksonomies - large databases arising from collaborative tagging of items by independent users - are becoming an increasingly important way of categorizing information. In these systems users can tag items with free words, resulting in a tripartite item-tag-user network. Although there are no prescribed relations between tags, the way users think about the different categories presumably has some built in hierarchy, in which more special concepts are descendants of some more general categories. Several applications would benefit from the knowledge of this hierarchy. Here we apply a recent method to check the differences and similarities of hierarchies resulting from tags given by independent individuals and from tags given by a centrally managed repository system. The results from out method showed substantial differences between the lower part of the hierarchies, and in contrast, a relatively high similarity at the top of the hierarchies.

physics.soc-ph↗

arXiv2014

Extracting tag hierarchies

Tagging items with descriptive annotations or keywords is a very natural way to compress and highlight information about the properties of the given entity. Over the years several methods have been proposed for extracting a hierarchy between the tags for systems with a "flat", egalitarian organization of the tags, which is very common when the tags correspond to free words given by numerous independent people. Here we present a complete framework for automated tag hierarchy extraction based on tag occurrence statistics. Along with proposing new algorithms, we are also introducing different quality measures enabling the detailed comparison of competing approaches from different aspects. Furthermore, we set up a synthetic, computer generated benchmark providing a versatile tool for testing, with a couple of tunable parameters capable of generating a wide range of test beds. Beside the computer generated input we also use real data in our studies, including a biological example with a pre-defined hierarchy between the tags. The encouraging similarity between the pre-defined and reconstructed hierarchy, as well as the seemingly meaningful hierarchies obtained for other real systems indicate that tag hierarchy extraction is a very promising direction for further research with a great potential for practical applications.

cs.IR↗

arXiv2014

Universal hierarchical behavior of citation networks

Many of the essential features of the evolution of scientific research are imprinted in the structure of citation networks. Connections in these networks imply information about the transfer of knowledge among papers, or in other words, edges describe the impact of papers on other publications. This inherent meaning of the edges infers that citation networks can exhibit hierarchical features, that is typical of networks based on decision-making. In this paper, we investigate the hierarchical structure of citation networks consisting of papers in the same field. We find that the majority of the networks follow a universal trend towards a highly hierarchical state, and i) the various fields display differences only concerning their phase in life (distance from the "birth" of a field) or ii) the characteristic time according to which they are approaching the stationary state. We also show by a simple argument that the alterations in the behavior are related to and can be understood by the degree of specialization corresponding to the fields. Our results suggest that during the accumulation of knowledge in a given field, some papers are gradually becoming relatively more influential than most of the other papers.

physics.soc-ph↗

arXiv2002

Tunable Lyapunov exponent in inverse magnetic billiards

The stability properties of the classical trajectories of charged particles are investigated in a two dimensional stadium-shaped inverse magnetic domain, where the magnetic field is zero inside the stadium domain and constant outside. In the case of infinite magnetic field the dynamics of the system is the same as in the Bunimovich billiard, i.e., ergodic and mixing. However, for weaker magnetic fields the phase space becomes mixed and the chaotic part gradually shrinks. The numerical measurements of the Lyapunov exponent (performed with a novel method) and the integrable/chaotic phase space volume ratio show that both quantities can be smoothly tuned by varying the external magnetic field. A possible experimental realization of the arrangement is also discussed.

cond-mat.mes-hall↗

arXiv1996

A New Method for Computing Topological Pressure

The topological pressure introduced by Ruelle and similar quantities describe dynamical multifractal properties of dynamical systems. These are important characteristics of mesoscopic systems in the classical regime. Original definition of these quantities are based on the symbolic description of the dynamics. It is hard or impossible to find symbolic description and generating partition to a general dynamical system, therefore these quantities are often not accessible for further studies. Here we present a new method by which the symbolic description can be omitted. We apply the method for a mixing and an intermittent system.

chao-dyn↗