Peter Janku

Peter Janku

Auto-generated profile, based on publicly available information

Peter Janku

Tomas Bata University in Zlín, Zlin, Czech Republic

Research Focus & Metrics

Academic Career Length

13Years (2014-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
21 October 2016 in MATEC Web of Conferences
Verified
Journal Article
Open access
PDF file
PJPeter Janku
Peter Janku
KKKarel Koplik
Karel Koplik
TDTomas Dulik
Tomas Dulik
ISIstvan Szabo
Istvan Szabo
MATEC Web of Conferences, open access proceedings in Materials science, Engineering and Chemistry

21 December 2018 in MENDEL
Verified
Journal Article
Open access
PDF file
PJPeter Janku
Peter Janku
ZOZuzana Komínková Oplatková
Zuzana Komínková Oplatková
TDTomas Dulik
Tomas Dulik
PSPetr Snopek
Petr Snopek
JLJiri Liba
Jiri Liba
This paper deals with the preliminary research of the fire detection in a video stream. Early fire detection can save lives and properties from huge losses and damages. Therefore the surveillance of the areas is necessary. Early fire discovery with high accuracy, i.e. a low number of false positive or false negative cases, is essential in any environment, especially in places with the high motion of people. The traditional fire detection sensors have some drawbacks: they need separate systems and infrastructure to be implemented, to use sensors in the case of the industrial environment with open fire technologies is often impossible, and others. The fire detection in a video stream is one of the possible and feasible solutions suitable for replacement or supplement of conventional fire detection sensors without a need for installation a huge infrastructure. The paper provides the state of the art in the fire detection. The following part of the paper proposes the new system of feature extraction and describes the feedforward neural network which was used for the training and testing of the proposed idea. The promising results are presented with over 93% accuracy on a selected dataset of movies which consist of more and highly varied instances than published by other researchers involved in the fire detection field. The structure of the neural networks promises higher computational speed than currently implemented deep learning systems.

14 July 2024
Verified
Conference Paper
Open access
MPMichal Pluhacek
Michal Pluhacek
JKJozef Kovac
Jozef Kovac
AVAdam Viktorin
Adam Viktorin
PJPeter Janku
Peter Janku
TKTomas Kadavy
Tomas Kadavy
RSRoman Senkerik
Roman Senkerik
This study investigates the use of the GPT-4 Turbo, a large language model, to enhance the Self-Organizing Migrating Algorithm (SOMA), specifically its All to All variant (SOMA-ATA). Utilizing the model's extensive context capacity for iterative prompting without feedback, we sought to autonomously generate superior algorithmic versions. Contrary to our initial hypothesis, the improvements did not progress linearly. Nevertheless, one iteration stood out significantly, consistently outperforming the baseline across various pairwise comparisons and showing a robust performance profile. This iteration's structure deviated substantially from traditional SOMA principles, underscoring the potential of large language models to create distinctive and effective algorithmic strategies. The results affirm the methodology's ability to produce high-performing algorithms without expert intervention, setting the stage for future research to integrate feedback mechanisms and conduct detailed code analyses to understand further the modifications made by the Large Language Models.

14 July 2024
Verified
Conference Paper
Open access
MPMichal Pluhacek
Michal Pluhacek
JKJozef Kovac
Jozef Kovac
PJPeter Janku
Peter Janku
TKTomas Kadavy
Tomas Kadavy
RSRoman Senkerik
Roman Senkerik
AVAdam Viktorin
Adam Viktorin
In this study, we investigate the applicability, challenges, and effectiveness of the advanced large language model GPT 4 Turbo in enhancing the selected metaheuristic algorithm, which is Differential Evolution. Our research primarily examines whether iterative, repetitive prompting could lead to progressive improvements in algorithm performance. We also explore the potential of developing enhanced algorithms through this process that markedly surpass the established baseline in terms of performance. In addition, the impact of the model's temperature parameter on these improvements is evaluated. As part of our diverse testing approach, we conduct an experiment where the best-performing algorithm from the initial phase is used as a new baseline. This step is to determine if further refinement via GPT 4 Turbo can achieve even better algorithmic efficiency. Finally, we have performed the benchmarking comparison of selected enhanced variants against the top three algorithms from the CEC 2022 competition.


1 May 2014
Verified
Conference Paper
Closed access
DPDaniel Pohuba
Daniel Pohuba
TDTomas Dulik
Tomas Dulik
PJPeter Janku
Peter Janku
This paper introduces a system for automatic evaluation of correctness and originality of source codes submitted by students enrolled in courses dealing with computer programming. Automatic correctness checking consists of searching for plagiarisms in assignments submitted earlier and checking the correct implementation of algorithms. User interface is implemented as a Moodle module using its Plagiarism API. The complete system is published with GPLv3 license; therefore other learning institutions can use it as well.