Wydawnictwa Uczelniane / TUL Press

Stały URI zbioruhttp://hdl.handle.net/11652/17

Przeglądaj

Filtry

2021 - 20233

Ustawienia

Autorsearch.filters.applied.operator.equals: search.filters.author.Tomczyk, Arkadiusz

Wyniki wyszukiwania

Teraz wyświetlane 1 - 3 z 3
  • Pozycja
    Energy Dissipation Anomalies in Buildings
    (Wydawnictwo Politechniki Łódzkiej, 2023) Morawski, Michał; Tomczyk, Arkadiusz; Idaczyk, Maciej
  • Pozycja
    Improvement of Attention Mechanism Explainability in Prediction of Chemical Molecules’ Properties
    (Wydawnictwo Politechniki Łódzkiej, 2023) Durys, Bartosz; Tomczyk, Arkadiusz
    In this paper, the analysis of selected graph neural network operators is presented. The classic Graph Convolutional Network (GCN) was compared with methods containing trainable attention coefficients: Graph Attention Network (GAT) and Graph Transformer (GT). Moreover, which is an original contribution of this work, training of GT was modified with an additional loss function component enabling easier explainability of the produced model. The experiments were conducted using datasets with chemical molecules where both classification and regression tasks are considered. The results show that additional constraint not only does not make the results worse but, in some cases, it improves predictions.
  • Pozycja
    Synergy of Convolutional Neural Networks and Geometric Active Contours
    (Wydawnictwo Politechniki Łódzkiej, 2021) Tomczyk, Arkadiusz; Pankiv, Oleksandr; Szczepaniak, Piotr S.
    Hybrid approach to machine learning techniques could potentially provide improvements in image segmentation results. In this paper, a model of cooperation of convolutional neural networks and geometric active contours is proposed and developed. The novelty of the approach lies in combining deep neural networks and active contour model in order to improve CNN output results. The method is examined on the image segmentation task and applied to the detection and extraction of nuclei of HL60 cell line. The model had been tested on both 2-D and 3-D images. Because of feature learning characteristics of convolutional neural networks, the proposed solution should perform well in multiple scenarios and can be considered generic.