Computer Science
Loading...
ISSN 1508-2806
e-ISSN: 2300-7036
Issue Date
2019
Volume
Vol. 20
Number
No. 1
Description
Reviewed by: Bartlomiej Sniezynski, Anna Zygmunt, Filip Malawski, Grazyna Slusarczyk, Piotr Breitkopf, Krzysztof Banas, Krzysztof Wolk, Marcin Pietron, Jerzy Proficz, Krzysztof Korcyl, Pawel Rosciszewski, Aleksander Wawer, Agnieszka Landowska
Journal Volume
Computer Science
Vol. 20 (2019)
Projects
Pages
Articles
COMPACT: biometric dataset of face images acquired in uncontrolled indoor environment
(Wydawnictwa AGH, 2019) Włodarczyk, Michał; Kacperski, Damian; Sankowski, Wojciech; Grabowski, Kamil
Biometric databases are important components that help improve the performanceof state-of-the-art recognition applications. The availability of more andmore challenging data is attracting the attention of researchers, who are systematicallydeveloping novel recognition algorithms and increasing the accuracyof identification. Surprisingly, most of the popular face datasets (like LFW orIJBA) are not fully unconstrained. The majority of the available images werenot acquired on-the-move, which reduces the amount of blurring that is causedby motion or incorrect focusing. Therefore, the COMPACT database for studyingless-cooperative face recognition is introduced in this paper. The datasetconsists of high-resolution images of 108 subjects acquired in a fully automatedmanner as people go through the recognition gate. This ensures that the collecteddata contains real-world degradation factors: different distances, expressions,occlusions, pose variations, and motion blur. Additionally, the authorsconducted a series of experiments that verified the face-recognition performanceon the collected data.
Hypergraph grammar based multi-thread multi-frontal direct solver with Galois scheduler
(Wydawnictwa AGH, 2019) Jopek, Konrad; Paszyński, Maciej; Paszyńska, Anna; Hassan, Muhammad Amber; Pingali, Keshav
In this paper, we analyze two-dimensional grids with point and edge singularities in order to develop an eficient parallel hypergraph grammar-based multi- frontal direct solver algorithm. We express these grids by a hypergraph. For these meshes, we define a sequence of hypergraph grammar productions expressing the construction of frontal matrices, eliminating fully assembled nodes, merging the resulting Schur complements, and repeating the process of elimination and merging until a single frontal matrix remains. The dependency relationship between hypergraph grammar productions is analyzed, and a dependency graph is plotted (which is equivalent to the elimination tree of a multi- frontal solver algorithm). We utilize a classical multi-frontal solver algorithm, the hypergraph grammar productions allow us to construct an eficient elimination tree based on the graph representation of the computational mesh (not the global matrix itself). The hypergraph grammar productions are assigned to nodes on a dependency graph, and they are implemented as tasks in the GALOIS parallel environment and scheduled according to the developed dependency graph over the shared memory parallel machine. We show that our hypergraph grammar-based solver outperforms the parallel MUMPS solver.
Towards textual data augmentation for neural networks: synonyms and maximum loss
(Wydawnictwa AGH, 2019) Jungiewicz, Michał; Smywiński-Pohl, Aleksander
Data augmentation is one of the ways to deal with labeled data scarcity and overfitting. Both of these problems are crucial for modern deep-learning algorithms, which require massive amounts of data. The problem is better explored in the context of image analysis than for text, this work is a step forward to help close this gap. We propose a method for augmenting textual data when training convolutional neural networks for sentence classification. The augmentation is based on the substitution of words using a thesaurus as well as Princeton University's WordNet. Our method improves upon the baseline in most of the cases. In terms of accuracy, the best of the variants is 1.2% (pp.) better than the baseline.
Study of OpenCL processing models for FPGA devices
(Wydawnictwa AGH, 2019) Szkotak, Piotr; Russek, Paweł; Wiatr, Kazimierz
In our study, we present the results of the implementation of the SHA-512 algorithm in FPGAs. The distinguished element of our work is that we conducted the work using OpenCL for FPGA, which is a relatively new development method for reconfigurable logic. We examine loop unrolling as an OpenCL performance optimization method and compare the efficiency of the different kernel implementation types: NDRange, Single-Work Item, and SIMD kernels. In our conclusions, we compare the metrics of the created FPGA accelerator to the corresponding GPGPU solutions. Also, our paper is accompanied by a source code repository to allow the reader to follow and extend our survey.
Computational intelligence for predicting biological effects of drug absorption in lungs
(Wydawnictwa AGH, 2019) Pacławski, Adam; Szlęk, Jakub; Mendyk, Aleksander
Recently, the lungs have been extensively examined as a route for delivering drugs (active pharmaceutical ingredients, APIs) into the bloodstream, this is mainly due to the possibility of the noninvasive administration of macromolecules such as proteins and peptides. The absorption mechanisms of chemical compounds in the lungs are still not fully understood, which makes pulmonary formulation composition development challenging. This manuscript presents the development of an empirical model capable of predicting the excipients’ influence on the absorption of drugs in the lungs. Due to the complexity of the problem and the not-fully-understood mechanisms of absorption, computational intelligence tools were applied. As a result, a mathematical formula was established and analyzed. The normalized root-mean-squared error (NRMSE) and $R^2$ of the model were 4.57%, and 0.83, respectively. The presented approach is beneficial both practically by developing an in silico predictive model and theoretically by gaining knowledge of the influence of APIs and excipient structure on absorption in the lungs.