@article{MTMT:35259088, title = {Global Sinkhorn Autoencoder — Optimal Transport on the latent representation of the full dataset}, url = {https://m2.mtmt.hu/api/publication/35259088}, author = {Csiszárik, Adrián and Kiss, Melinda and Maga, Balázs and Matszangosz, Ákos and Varga, Dániel}, journal-iso = {ANN UNIV SCI BP R EÖTVÖS NOM SECT COMPUT}, journal = {ANNALES UNIVERSITATIS SCIENTIARUM BUDAPESTINENSIS DE ROLANDO EOTVOS NOMINATAE SECTIO COMPUTATORICA}, volume = {57}, unique-id = {35259088}, issn = {0138-9491}, abstract = {We propose an Optimal Transport (OT)-based generative model from the Wasserstein Autoencoder (WAE) family of models, with the following innovative property: the optimization of the latent point positions takes place over the full training dataset rather than over a minibatch. Our contributions are the following: 1. We define a new class of global Wasserstein Autoencoder models, and implement an Optimal Transport-based incarnation we call the Global Sinkhorn Autoencoder. 2. We implement several metrics for evaluating such models, both in the unsupervised setting, and in a semi-supervised setting, which are the following: the global OT loss, which measures the OT loss on the full test dataset; the reconstruction error on the full test dataset; a so-called covered area which measures how well the latent points are matched; and two types of clustering measures. 3. We demonstrate on specific complex prior distributions that global optimal transport improves the performance of generative models compared to minibatch-based baselines when evaluated by the previously listed metrics.}, year = {2024}, pages = {101-115} } @article{MTMT:34848051, title = {Box dimension of generic Hölder level sets}, url = {https://m2.mtmt.hu/api/publication/34848051}, author = {Buczolich, Zoltán and Maga, Balázs}, doi = {10.1016/j.indag.2024.03.015}, journal-iso = {INDAGAT MATH NEW SER}, journal = {INDAGATIONES MATHEMATICAE-NEW SERIES}, volume = {35}, unique-id = {34848051}, issn = {0019-3577}, abstract = {Hausdorff dimension of level sets of generic continuous functions defined on fractals can give information about the “thickness/narrow cross-sections” of a “network” corresponding to a fractal set. This leads to the definition of the topological Hausdorff dimension of fractals. Finer information might be obtained by considering the Hausdorff dimension of level sets of generic 1-Hölder- functions, which has a stronger dependence on the geometry of the fractal, as displayed in our previous papers (Buczolich et al., 2022 [9,10]). In this paper, we extend our investigations to the lower and upper box-counting dimensions as well: while the former yields results highly resembling the ones about the Hausdorff dimension of level sets, the latter exhibits a different behavior. Instead of “finding narrow-cross sections”, results related to upper box-counting dimension “measure” how much level sets can spread out on the fractal, and how widely the generic function can “oscillate” on it. Key differences are illustrated by giving estimates concerning the Sierpiński triangle.}, year = {2024}, eissn = {1872-6100}, pages = {531-554}, orcid-numbers = {Buczolich, Zoltán/0000-0001-5481-8797} } @article{MTMT:32750764, title = {Strong one-sided density without uniform density}, url = {https://m2.mtmt.hu/api/publication/32750764}, author = {Buczolich, Zoltán and Hanson, Bruce and Maga, Balázs and Vértesy, Gáspár}, doi = {10.1007/s10998-022-00455-9}, journal-iso = {PERIOD MATH HUNG}, journal = {PERIODICA MATHEMATICA HUNGARICA}, volume = {86}, unique-id = {32750764}, issn = {0031-5303}, year = {2023}, eissn = {1588-2829}, pages = {13-23}, orcid-numbers = {Buczolich, Zoltán/0000-0001-5481-8797} } @article{MTMT:33124160, title = {Generic Hölder level sets and fractal conductivity}, url = {https://m2.mtmt.hu/api/publication/33124160}, author = {Buczolich, Zoltán and Maga, Balázs and Vértesy, Gáspár}, doi = {10.1016/j.chaos.2022.112696}, journal-iso = {CHAOS SOLITON FRACT}, journal = {CHAOS SOLITONS & FRACTALS}, volume = {164}, unique-id = {33124160}, issn = {0960-0779}, year = {2022}, eissn = {1873-2887}, orcid-numbers = {Buczolich, Zoltán/0000-0001-5481-8797} } @CONFERENCE{MTMT:33099660, title = {Valós analízisbeli problémák}, url = {https://m2.mtmt.hu/api/publication/33099660}, author = {Maga, Balázs}, booktitle = {Intézményi ÚNKP Konferencia 2022}, unique-id = {33099660}, year = {2022}, pages = {165} } @article{MTMT:33072066, title = {Generic Hölder level sets on fractals}, url = {https://m2.mtmt.hu/api/publication/33072066}, author = {Buczolich, Zoltán and Maga, Balázs and Vértesy, Gáspár}, doi = {10.1016/j.jmaa.2022.126543}, journal-iso = {J MATH ANAL APPL}, journal = {JOURNAL OF MATHEMATICAL ANALYSIS AND APPLICATIONS}, volume = {516}, unique-id = {33072066}, issn = {0022-247X}, abstract = {Hausdorff dimensions of level sets of generic continuous functions defined on fractals were considered in two papers by R. Balka, Z. Buczolich and M. Elekes. In those papers the topological Hausdorff dimension of fractals was defined. In this paper we start to study level sets of generic 1-Hölder-α functions defined on fractals. This is related to some sort of “thickness”, “conductivity” properties of fractals. The main concept of our paper is D⁎(α,F) which is the essential supremum of the Hausdorff dimensions of the level sets of a generic 1-Hölder-α function defined on the fractal F. We prove some basic properties of D⁎(α,F), we calculate its value for an example of a “thick fractal sponge”, we show that for connected self similar sets D⁎(α,F) it equals the Hausdorff dimension of almost every level in the range of a generic 1-Hölder-α function. © 2022 The Author(s)}, year = {2022}, eissn = {1096-0813}, orcid-numbers = {Buczolich, Zoltán/0000-0001-5481-8797} } @article{MTMT:32704811, title = {Generic power series on subsets of the unit disk}, url = {https://m2.mtmt.hu/api/publication/32704811}, author = {Maga, Balázs and Maga, Péter}, doi = {10.21136/CMJ.2022.0021-21}, journal-iso = {CZECH MATH J}, journal = {CZECHOSLOVAK MATHEMATICAL JOURNAL}, volume = {72}, unique-id = {32704811}, issn = {0011-4642}, year = {2022}, eissn = {1572-9141}, pages = {637-652} } @inproceedings{MTMT:32627793, title = {Global Sinkhorn Autoencoder - Optimal transport on the latent representation of the full dataset}, url = {https://m2.mtmt.hu/api/publication/32627793}, author = {Gellert, Karolyi and Kiss, Melinda and Csiszárik, Adrián and Matszangosz, Ákos and Maga, Balázs and Varga, Dániel}, booktitle = {Conference on Developments in Computer Science}, unique-id = {32627793}, year = {2021}, pages = {199-202} } @article{MTMT:31981091, title = {Big and little Lipschitz one sets}, url = {https://m2.mtmt.hu/api/publication/31981091}, author = {Buczolich, Zoltán and Hanson, Bruce and Maga, Balázs and Vértesy, Gáspár}, doi = {10.1007/s40879-021-00458-9}, journal-iso = {EUR J MATH}, journal = {EUROPEAN JOURNAL OF MATHEMATICS}, volume = {7}, unique-id = {31981091}, issn = {2199-675X}, year = {2021}, eissn = {2199-6768}, pages = {464-488}, orcid-numbers = {Buczolich, Zoltán/0000-0001-5481-8797} } @CONFERENCE{MTMT:31849361, title = {Attention U-net based adversarial architectures for chest X-ray lung segmentation}, url = {https://m2.mtmt.hu/api/publication/31849361}, author = {Gaál, G. and Maga, Balázs and Lukács, András}, booktitle = {2020 Workshop on Applied Deep Generative Networks, ADGN 2020}, volume = {2692}, unique-id = {31849361}, abstract = {X-ray is by far the most common among medical imaging modalities, being faster, more accessible, and more cost-effective compared to other radiographic methods. Chest X-ray (CXR) is the most commonly requested test due to its contribution to the early detection of lung cancer. The most important biomarker in detecting cancer of the lung are nodules, and in finding those, lung segmentation of chest X-rays is essential. Another goal is interpretability, helping radiologists integrate computer-aided detection methods into their diagnostic pipeline, greatly reducing their workload. For this reason, a robust algorithm to perform this otherwise arduous segmentation task is much desired in the field of medical imaging. In this work, we present a novel deep learning approach that uses state-of-the-art fully convolutional neural networks in conjunction with an adversarial critic model. Our network generalized well to CXR images of unseen datasets with different patient profiles, achieving a final DSC of 97.5% on the JSRT CXR dataset. Copyright © 2020 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0)}, keywords = {diagnosis; DISEASES; Medical imaging; Cost effectiveness; X rays; Biological organs; Cost effective; Interpretability; X ray radiography; State of the art; LEARNING APPROACH; Computer aided detection; Convolutional neural networks; Deep learning; Imaging modality; Robust algorithm; lung segmentation}, year = {2020}, orcid-numbers = {Lukács, András/0000-0003-3955-9824} }