Güleser Kalaycı Demir
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Güleser Kalaycı Demir has a total of 16 co-authors. Here are the top co-authors by number of shared publications:
- Hatice Doğan
Department of Electrical and Electronics Engineering, Dokuz Eylül University, 35390 İzmir, Turkey
2 shared publications - Nasuf Sonmez
Department of Physics Faculty of Science Ege University 35040 Izmir Turkey
2 shared publications Alkım Şükrü Özkan
Department of Electrical and Electronics Engineering, Dokuz Eylül University, 35390 İzmir, Turkey
1 shared publication- Berk Ağın
Dokuz Eylul University
1 shared publication - Cüneyt Güzeliş
İzmir University of Economics
1 shared publication - Durmuş Demir
Faculty of Engineering and Natural Sciences, Sabancı University, Tuzla, Turkey
1 shared publication
Latest Publications
- Open Access
- 23 May 2025
- Published by Deu Muhendislik Fakultesi Fen ve Muhendislik in Dokuz Eylül Üniversitesi Mühendislik Fakültesi Fen ve Mühendislik Dergisi
- Vol. 27 (80) , 224-233
Indoor Visual Navigation Based on Deep Reinforcement Learning with Neural Ordinary Differential Equations
Recently, Deep Reinforcement Learning (DRL) has gained attention as a promising approach to tackle the challenging problem of mobile robot navigation. This study proposes reinforcement learning utilizing Neural Ordinary Differential Equations (NODEs), which offer effective training and memory capacities, and applies it to model-free point-to-point navigation task. Through the use of NODEs, we achieved improvements in navigation performance as well as enhancements in resource optimization and adaptation. Extensive simulation studies were conducted using real-world indoor scenes to validate our approach. Results effectively demonstrated the effectiveness of our proposed NODEs-based methodology in enhancing navigation performance compared to traditional ResNet and CNN architectures. Furthermore, curriculum learning strategies were integrated into our study to enable the agent to learn through progressively more complex navigation scenarios. The results obtained indicate that this approach facilitates faster and more robust reinforcement learning.
- 23 May 2025
- Published by Deu Muhendislik Fakultesi Fen ve Muhendislik in Dokuz Eylül Üniversitesi Mühendislik Fakültesi Fen ve Mühendislik Dergisi
- Vol. 27 (80) , 224-233
Open AccessIndoor Visual Navigation Based on Deep Reinforcement Learning with Neural Ordinary Differential Equations
Recently, Deep Reinforcement Learning (DRL) has gained attention as a promising approach to tackle the challenging problem of mobile robot navigation. This study proposes reinforcement learning utilizing Neural Ordinary Differential Equations (NODEs), which offer effective training and memory capacities, and applies it to model-free point-to-point navigation task. Through the use of NODEs, we achieved improvements in navigation performance as well as enhancements in resource optimization and adaptation. Extensive simulation studies were conducted using real-world indoor scenes to validate our approach. Results effectively demonstrated the effectiveness of our proposed NODEs-based methodology in enhancing navigation performance compared to traditional ResNet and CNN architectures. Furthermore, curriculum learning strategies were integrated into our study to enable the agent to learn through progressively more complex navigation scenarios. The results obtained indicate that this approach facilitates faster and more robust reinforcement learning.
- 23 July 2024
- Published by Bulent Evcevit University in Karaelmas Science and Engineering Journal
- Vol. 14 (2) , 13-28
Siamese Neural Networks Based Ensemble Model for the Prediction of Protein-Protein Interactions
A wide range of biological processes, including signal transmission, immunological responses, and metabolic cycles, are impacted by protein-protein interactions. These interactions have enormous implications for figuring out the origins of diseases and creating treatments. However, experimental methods for identifying PPIs are resource-intensive, time-consuming, and have limited coverage. Thus, computational techniques are essential to help and enhance activities related to protein identification. This study aims to build a deep learning network for predicting protein-protein interactions using only sequence information. Three different encoding methods are used to encode protein sequences: Binary Encoding, Autocovariance, and Position Specific Scoring Matrix. In order to predict protein-protein interactions, a convolutional Siamese neural network is employed to find complex patterns between protein sequence pairs. This network consists of two identical subnetworks with matched parameters. When applied to the human dataset, the suggested technique shows strong prediction performance with an accuracy of 84.07%, sensitivity of 92.45%, and precision of 91.45% for the model using the PSSM protein representation approach. An ensemble approach is suggested to combine the outputs from these three encoders because it is known that different encoding techniques capture various aspects of the same protein sequence. The accuracy obtained increased to 86.27% for the ensemble approach on the test set, with a sensitivity of 93.07% and a precision of 92.15%. The outcome highlights the importance of integrating several encoding methods to benefit from their complementary features and raise the accuracy of protein-protein interaction prediction.
- 23 July 2024
- Published by Bulent Evcevit University in Karaelmas Science and Engineering Journal
- Vol. 14 (2) , 13-28
Siamese Neural Networks Based Ensemble Model for the Prediction of Protein-Protein Interactions
A wide range of biological processes, including signal transmission, immunological responses, and metabolic cycles, are impacted by protein-protein interactions. These interactions have enormous implications for figuring out the origins of diseases and creating treatments. However, experimental methods for identifying PPIs are resource-intensive, time-consuming, and have limited coverage. Thus, computational techniques are essential to help and enhance activities related to protein identification. This study aims to build a deep learning network for predicting protein-protein interactions using only sequence information. Three different encoding methods are used to encode protein sequences: Binary Encoding, Autocovariance, and Position Specific Scoring Matrix. In order to predict protein-protein interactions, a convolutional Siamese neural network is employed to find complex patterns between protein sequence pairs. This network consists of two identical subnetworks with matched parameters. When applied to the human dataset, the suggested technique shows strong prediction performance with an accuracy of 84.07%, sensitivity of 92.45%, and precision of 91.45% for the model using the PSSM protein representation approach. An ensemble approach is suggested to combine the outputs from these three encoders because it is known that different encoding techniques capture various aspects of the same protein sequence. The accuracy obtained increased to 86.27% for the ensemble approach on the test set, with a sensitivity of 93.07% and a precision of 92.15%. The outcome highlights the importance of integrating several encoding methods to benefit from their complementary features and raise the accuracy of protein-protein interaction prediction.
- Open Access
- 5 October 2023
- Published by American Chemical Society (ACS) in ACS Omega
- Vol. 8 (41) , 38452-38458
Mathematical Model for Growth and Rifampicin-Dependent Killing Kinetics of Escherichia coli Cells
Antibiotic resistance is a global health threat. We urgently need better strategies to improve antibiotic use to combat antibiotic resistance. Currently, there are a limited number of antibiotics in the treatment repertoire of existing bacterial infections. Among them, rifampicin is a broad-spectrum antibiotic against various bacterial pathogens. However, during rifampicin exposure, the appearance of persisters or resisters decreases its efficacy. Hence, to benefit more from rifampicin, its current standard dosage might be reconsidered and explored using both computational tools and experimental or clinical studies. In this study, we present the mathematical relationship between the concentration of rifampicin and the growth and killing kinetics of Escherichia coli cells. We generated time-killing curves of E. coli cells in the presence of 4, 16, and 32 μg/mL rifampicin exposures. We specifically focused on the oscillations with decreasing amplitude over time in the growth and killing kinetics of rifampicin-exposed E. coli cells. We propose the solution form of a second-order linear differential equation for a damped oscillator to represent the mathematical relationship. We applied a nonlinear curve fitting solver to time-killing curve data to obtain the model parameters. The results show a high fitting accuracy.
- 5 October 2023
- Published by American Chemical Society (ACS) in ACS Omega
- Vol. 8 (41) , 38452-38458
Open AccessMathematical Model for Growth and Rifampicin-Dependent Killing Kinetics of Escherichia coli Cells
Antibiotic resistance is a global health threat. We urgently need better strategies to improve antibiotic use to combat antibiotic resistance. Currently, there are a limited number of antibiotics in the treatment repertoire of existing bacterial infections. Among them, rifampicin is a broad-spectrum antibiotic against various bacterial pathogens. However, during rifampicin exposure, the appearance of persisters or resisters decreases its efficacy. Hence, to benefit more from rifampicin, its current standard dosage might be reconsidered and explored using both computational tools and experimental or clinical studies. In this study, we present the mathematical relationship between the concentration of rifampicin and the growth and killing kinetics of Escherichia coli cells. We generated time-killing curves of E. coli cells in the presence of 4, 16, and 32 μg/mL rifampicin exposures. We specifically focused on the oscillations with decreasing amplitude over time in the growth and killing kinetics of rifampicin-exposed E. coli cells. We propose the solution form of a second-order linear differential equation for a damped oscillator to represent the mathematical relationship. We applied a nonlinear curve fitting solver to time-killing curve data to obtain the model parameters. The results show a high fitting accuracy.
- 19 January 2023
- Published by Springer Nature in JBIC Journal of Biological Inorganic Chemistry
- Vol. 28 (2) , 213-224
Quantum tunneling time delay investigation of ${{\varvec{K}}}^{+}$ ion in human telomeric G-quadruplex systemsGuanine-rich quadruplex DNA (G-quadruplex) is of interest both in cell biology and nanotechnology. Its biological functions necessitate a G-quadruplex to be stabilized against escape of the monovalent metal cations. The potassium ion (\({{\varvec{K}}}^{+}\)) is particularly important as it experiences a potential energy barrier while it enters and exits the G-quadruplex systems which are normally found in human telomere. In the present work, we analyzed the time it takes for the \({{\varvec{K}}}^{+}\) cations to get in and out of the G-quadruplex. Our time estimate is based on entropic tunneling time—a time formula which gave biologically relevant results for DNA point mutation by proton tunneling. The potential energy barrier experienced by \({{\varvec{K}}}^{+}\) ions is determined from a quantum mechanical simulation study, Schrodinger equation is solved using MATLAB, and the computed eigenfunctions and eigenenergies are used in the entropic tunneling time formula to compute the time delay and charge accumulation rate during the tunneling of \({{\varvec{K}}}^{+}\) in G-quadruplex. The computations have shown that ion tunneling takes picosecond times. In addition, average \({{\varvec{K}}}^{+}\) accumulation rate is found to be in the picoampere range. Our results show that time delay during the \({{\varvec{K}}}^{+}\) ion tunneling is in the ballpark of the conformational transition times in biological systems, and it could be an important parameter for understanding its biological role in human DNA as well as for the possible applications in biotechnology. To our knowledge, for the first time in the literature, time delay during the ion tunneling from and into G-quadruplexes is computed. Graphical abstract- 19 January 2023
- Published by Springer Nature in JBIC Journal of Biological Inorganic Chemistry
- Vol. 28 (2) , 213-224
Quantum tunneling time delay investigation of ${{\varvec{K}}}^{+}$ ion in human telomeric G-quadruplex systemsGuanine-rich quadruplex DNA (G-quadruplex) is of interest both in cell biology and nanotechnology. Its biological functions necessitate a G-quadruplex to be stabilized against escape of the monovalent metal cations. The potassium ion (\({{\varvec{K}}}^{+}\)) is particularly important as it experiences a potential energy barrier while it enters and exits the G-quadruplex systems which are normally found in human telomere. In the present work, we analyzed the time it takes for the \({{\varvec{K}}}^{+}\) cations to get in and out of the G-quadruplex. Our time estimate is based on entropic tunneling time—a time formula which gave biologically relevant results for DNA point mutation by proton tunneling. The potential energy barrier experienced by \({{\varvec{K}}}^{+}\) ions is determined from a quantum mechanical simulation study, Schrodinger equation is solved using MATLAB, and the computed eigenfunctions and eigenenergies are used in the entropic tunneling time formula to compute the time delay and charge accumulation rate during the tunneling of \({{\varvec{K}}}^{+}\) in G-quadruplex. The computations have shown that ion tunneling takes picosecond times. In addition, average \({{\varvec{K}}}^{+}\) accumulation rate is found to be in the picoampere range. Our results show that time delay during the \({{\varvec{K}}}^{+}\) ion tunneling is in the ballpark of the conformational transition times in biological systems, and it could be an important parameter for understanding its biological role in human DNA as well as for the possible applications in biotechnology. To our knowledge, for the first time in the literature, time delay during the ion tunneling from and into G-quadruplexes is computed. Graphical abstract- 20 June 2022
- Published by World Scientific Pub Co Pte Ltd in International Journal of Modern Physics A
- Vol. 37 (17)
Light charged Higgs search with deviation neural networks
- 20 June 2022
- Published by World Scientific Pub Co Pte Ltd in International Journal of Modern Physics A
- Vol. 37 (17)
Light charged Higgs search with deviation neural networks
- 26 July 2019
- Published by The Scientific and Technological Research Council of Turkey (TUBITAK-ULAKBIM) - DIGITAL COMMONS JOURNALS in Turkish Journal of Electrical Engineering and Computer Sciences
- Vol. 27 (4) , 3247-3257
Importance-based signal detection and parameter estimation with applications tonew particle search
- 26 July 2019
- Published by The Scientific and Technological Research Council of Turkey (TUBITAK-ULAKBIM) - DIGITAL COMMONS JOURNALS in Turkish Journal of Electrical Engineering and Computer Sciences
- Vol. 27 (4) , 3247-3257
Importance-based signal detection and parameter estimation with applications tonew particle search
- Open Access
Deep-Learning in Search of Light Charged Higgs
In this work, we deep-learn light charged Higgs signal in top quark decays which poses difficulties due to strong W boson contamination. We construct Deep Neural Networks (DNN) with appropriate architecture and determine signal extraction efficiency by considering various features (kinematical and human engineered parameters). Results show that DNN gives better performance than the classical neural networks and has ability to find regions of high efficiency even the input features are not human-engineered. In a sense, human-engineered high-level features are offset by DNNs with different combinations of the low-level kinematical features. Additionally, it is shown that increasing the number of processing units in DNNs does not necessarily cause an increase in efficiency due mainly to increased complexity. Our method and results can set an example of signal extraction from strong backgrounds.
Open AccessDeep-Learning in Search of Light Charged Higgs
In this work, we deep-learn light charged Higgs signal in top quark decays which poses difficulties due to strong W boson contamination. We construct Deep Neural Networks (DNN) with appropriate architecture and determine signal extraction efficiency by considering various features (kinematical and human engineered parameters). Results show that DNN gives better performance than the classical neural networks and has ability to find regions of high efficiency even the input features are not human-engineered. In a sense, human-engineered high-level features are offset by DNNs with different combinations of the low-level kinematical features. Additionally, it is shown that increasing the number of processing units in DNNs does not necessarily cause an increase in efficiency due mainly to increased complexity. Our method and results can set an example of signal extraction from strong backgrounds.
- 3 October 2016
- Published by Wiley in International Transactions on Electrical Energy Systems
- Vol. 27 (4) , e2284
Quadratic approximation-based hybrid Artificial Cooperative Search algorithm for economic emission load dispatch problems
- 3 October 2016
- Published by Wiley in International Transactions on Electrical Energy Systems
- Vol. 27 (4) , e2284
Quadratic approximation-based hybrid Artificial Cooperative Search algorithm for economic emission load dispatch problems
- 1 January 2015
- Published by Springer Nature
- p. 77-86
Coexistence of Deterministic and Stochastic Bistability in a 1-D Birth-Death Process with Hill Type Nonlinear Birth Rates
- 1 January 2015
- Published by Springer Nature
- p. 77-86
Coexistence of Deterministic and Stochastic Bistability in a 1-D Birth-Death Process with Hill Type Nonlinear Birth Rates