Nurcin Celik
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Nurcin Celik has a total of 48 co-authors. Here are the top co-authors by number of shared publications:
- Haluk Damgacioglu
Department of Public Health Sciences, Medical University of South Carolina , Charleston, SC
7 shared publications - Chun-Hung Chen
George Mason University
3 shared publications - Temitope Runsewe
University of Miami
3 shared publications Abdurrahman Yavuz
University of Miami
2 shared publications- Abdurrahman Yavuz
University of Miami
2 shared publications Aristotelis E. Thanos
2 shared publications
Latest Publications
- 2 July 2026
- Published by Springer Nature
- p. 797-825
DDDAS @ 5G and Beyond 5G Networks for Resilient Communications Infrastructures and Microgrid Clusters
- 2 July 2026
- Published by Springer Nature
- p. 797-825
DDDAS @ 5G and Beyond 5G Networks for Resilient Communications Infrastructures and Microgrid Clusters
- 2 July 2026
- Published by Springer Nature
- p. 925-947
Dynamic Data Driven Applications Systems (DDDAS) for Cyber Risk Management in Microgrids
- 2 July 2026
- Published by Springer Nature
- p. 925-947
Dynamic Data Driven Applications Systems (DDDAS) for Cyber Risk Management in Microgrids
- 2 July 2026
- Published by Springer Nature
- p. 431-449
Self-Healing of Distributed Microgrids Using DDDAMS
- 2 July 2026
- Published by Springer Nature
- p. 431-449
Self-Healing of Distributed Microgrids Using DDDAMS
- 18 October 2025
- Published by World Scientific Pub Co Pte Ltd in Asia-Pacific Journal of Operational Research
- Vol. 42 (06)
Supercharging Digital Twins with AI
The increasing availability of large data sets has enabled the development of digital twins (DTs), virtual models that accurately replicate real-world systems. This paper explores how AI and machine learning (AI/ML) can enhance the efficiency of stochastic simulations, particularly optimization via simulation, to strengthen the predictive capabilities of DTs and drive better decision-making. To achieve this, we present a structured framework for simulation optimization that formally establishes the role of AI/ML in improving decision-making within DTs. We introduce innovative methodologies, SAMPLE and TRAIN, designed to enhance the efficiency of simulation optimization by integrating AI with simulation tools. These approaches transform complex problems into AI models, reducing the need for extensive observations while enabling efficient solutions for intricate manufacturing systems. They also improve constraint handling by converting chance constraints into conditional value-at-risk (CVaR) constraints and estimating them efficiently through CVaR regression. By leveraging information from both constraints and objective functions, these frameworks facilitate the effective application of general optimization techniques, accelerating the search process within the metamodel. Moreover, they ensure the accuracy of metamodels, strengthening their reliability in decision support. We conclude with insights into future opportunities for leveraging AI/ML techniques to further enhance and expand the capabilities of digital twins, unlocking new pathways for operational efficiency and strategic value creation.
- 18 October 2025
- Published by World Scientific Pub Co Pte Ltd in Asia-Pacific Journal of Operational Research
- Vol. 42 (06)
Supercharging Digital Twins with AI
The increasing availability of large data sets has enabled the development of digital twins (DTs), virtual models that accurately replicate real-world systems. This paper explores how AI and machine learning (AI/ML) can enhance the efficiency of stochastic simulations, particularly optimization via simulation, to strengthen the predictive capabilities of DTs and drive better decision-making. To achieve this, we present a structured framework for simulation optimization that formally establishes the role of AI/ML in improving decision-making within DTs. We introduce innovative methodologies, SAMPLE and TRAIN, designed to enhance the efficiency of simulation optimization by integrating AI with simulation tools. These approaches transform complex problems into AI models, reducing the need for extensive observations while enabling efficient solutions for intricate manufacturing systems. They also improve constraint handling by converting chance constraints into conditional value-at-risk (CVaR) constraints and estimating them efficiently through CVaR regression. By leveraging information from both constraints and objective functions, these frameworks facilitate the effective application of general optimization techniques, accelerating the search process within the metamodel. Moreover, they ensure the accuracy of metamodels, strengthening their reliability in decision support. We conclude with insights into future opportunities for leveraging AI/ML techniques to further enhance and expand the capabilities of digital twins, unlocking new pathways for operational efficiency and strategic value creation.
- Open Access
STABLE: Gated Continual Learning for Large Language Models
Large language models (LLMs) increasingly require mechanisms for continual adaptation without full retraining. However, sequential updates can lead to catastrophic forgetting, where new edits degrade previously acquired knowledge. This work presents STABLE, a gated continual self editing framework that constrains forgetting during sequential updates using parameter efficient fine tuning via Low Rank Adaptation (LoRA; see arXiv:2106.09685). Each candidate edit is evaluated against a stability budget using one of three metrics: (i) Exact Match (EM) drop, capturing factual accuracy loss; (ii) bits increase, reflecting reduced model confidence; and (iii) KL divergence, quantifying distributional drift between the base and adapted models. If a threshold is exceeded, the LoRA update is rescaled through a clipping procedure or rejected. Experiments on the Qwen-2.5-7B model show that gating effectively mitigates forgetting while preserving adaptability. EM based gating achieved the highest cumulative performance in short continual learning sequences. Our results show that different gating strategies can achieve comparable distribution shift (measured by KL divergence) while producing different accuracy outcomes, highlighting the importance of gating design in continual adaptation. This approach offers a principled method for continual model editing, enabling LLMs to integrate new knowledge while maintaining reliability. Code: https://github.com/Bhoy1/STABLE
Open AccessSTABLE: Gated Continual Learning for Large Language Models
Large language models (LLMs) increasingly require mechanisms for continual adaptation without full retraining. However, sequential updates can lead to catastrophic forgetting, where new edits degrade previously acquired knowledge. This work presents STABLE, a gated continual self editing framework that constrains forgetting during sequential updates using parameter efficient fine tuning via Low Rank Adaptation (LoRA; see arXiv:2106.09685). Each candidate edit is evaluated against a stability budget using one of three metrics: (i) Exact Match (EM) drop, capturing factual accuracy loss; (ii) bits increase, reflecting reduced model confidence; and (iii) KL divergence, quantifying distributional drift between the base and adapted models. If a threshold is exceeded, the LoRA update is rescaled through a clipping procedure or rejected. Experiments on the Qwen-2.5-7B model show that gating effectively mitigates forgetting while preserving adaptability. EM based gating achieved the highest cumulative performance in short continual learning sequences. Our results show that different gating strategies can achieve comparable distribution shift (measured by KL divergence) while producing different accuracy outcomes, highlighting the importance of gating design in continual adaptation. This approach offers a principled method for continual model editing, enabling LLMs to integrate new knowledge while maintaining reliability. Code: https://github.com/Bhoy1/STABLE
- 26 August 2025
- Published by Springer Nature
- p. 383-390
Adaptive Data Driven Network Slicing and Resource Blocks Assignment Using Deep Reinforcement Learning
- 26 August 2025
- Published by Springer Nature
- p. 383-390
Adaptive Data Driven Network Slicing and Resource Blocks Assignment Using Deep Reinforcement Learning
- 27 February 2024
- Published by Springer Nature
- p. 135-143
DDDAS-Based Learning for Edge Computing at 5G and Beyond 5G
- 27 February 2024
- Published by Springer Nature
- p. 135-143
DDDAS-Based Learning for Edge Computing at 5G and Beyond 5G
- 6 September 2023
- Published by Springer Nature
- p. 743-764
Manufacturing the Future via Dynamic Data Driven Applications Systems (DDDAS)
- 6 September 2023
- Published by Springer Nature
- p. 743-764
Manufacturing the Future via Dynamic Data Driven Applications Systems (DDDAS)
- 26 April 2023
- Published by Elsevier in Journal of Environmental Management
- Vol. 340, 117855
Machine learning models for estimating contamination across different curbside collection strategies
- 26 April 2023
- Published by Elsevier in Journal of Environmental Management
- Vol. 340, 117855
Machine learning models for estimating contamination across different curbside collection strategies