Tutorial Seminars

T1. High Altitude Platform Stations (HAPS) and 6G & Beyond Wireless Communication Systems

Dr. Metin Öztürk (Ankara Yıldırım Beyazıt University), Dr. Cihan Emre Kement (ASELSAN)

• Session 1: Introduction to HAPS Technology and Application Areas (75 mins)
  • 1.1. What is HAPS? Technical infrastructure, operating principles, and key concepts
  • 1.2. Relationship between HAPS and 6G: Integration with satellite and terrestrial systems
  • 1.3. Application Areas of HAPS: Rural communication, IoT, and disaster management
  • 1.4. Participant questions and discussion
• Session 2: 6G Vision, Future of HAPS, and Challenges (75 mins)
  • 2.1. ITU 6G Vision and European Union Perspective: Goals of digital equality, sustainability, and energy efficiency
  • 2.2. Role of HAPS in the 6G Ecosystem: Ultra-wideband and low-latency applications

T2. Centralized and Distributed Learning Techniques for Millimeter Wave and Terahertz-band

Dr. Ahmet Elbir (Ä°stinye University)

• Session 1: Deep Learning for Hybrid Beamforming and Channel Estimation (75 mins)
  • 1.1. Motivation for DL
  • 1.2. Learning Schemes
  • 1.3. Types of Learning Models
  • 1.4. Data Generation and Training
  • 1.5. Examples: Single/Multi-User Hybrid Beamforming
  • 1.6. Examples: Narrow/Wideband Hybrid Beamforming
  • 1.7. Implementation in MATLAB (data generation and model training)
• Session 2: Advanced Learning Schemes for Communications (75 mins)
  • 2.1. DL for Joint Antenna Selection and Hybrid Beamforming
  • 2.2. Online Learning for Joint Channel Estimation and Hybrid Beamforming
  • 2.3. Transfer Learning for Antenna Selection
  • 2.4. Distributed Learning for Channel Estimation and Beamforming
  • 2.5. DL for Direction of Arrival Estimation
  • 2.6. DL for Joint Radar-Communication Hybrid Beamforming
  • 2.7. DL for Distributed Beamforming
  • 2.8. Future Research Directions
  • 2.9. Discussion/Q&A

T3. Large Language Models and Turkish Adaptation

Dr. Çağrı Toraman (Middle East Technical University)

• Session 1: Fundamentals of Large Language Models (75 mins)
  • 1.1. What are Large Language Models?
  • 1.2. Historical Development of Large Language Models
  • 1.3. Architectures of Large Language Models
  • 1.4. Pre-training and Fine-tuning Techniques
  • 1.5. Advantages and Challenges of Large Language Models
  • 1.6. Q&A
• Session 2: Turkish Adaptation of Large Language Models (75 mins)
  • 2.1. Features and Challenges of the Turkish Language
  • 2.2. Transfer Learning and Multilingual Models
  • 2.3. Tokenization Strategies
  • 2.4. Adaptation Techniques for Turkish LLMs
  • 2.5. Evaluation Metrics and Benchmarks for Turkish LLMs
  • 2.6. Q&A

T4. Cell-Free Massive MIMO: Fundamentals and Energy-Aware C-RAN Implementation

Dr. Özlem Tuğfe Demir (TOBB University of Economics and Technology)

• Session 1: Cell-Free Massive MIMO (75 mins)
  • 1.1. Transition from Cellular to Cell-Free Massive MIMO
  • 1.1.1. Basic terminology, cellular networks and their drawbacks
  • 1.1.2. Key advantages of the cell-free paradigm
  • 1.1.3. Historical background and differences from 4G CoMP
  • 1.2. Fundamentals of Cell-Free Massive MIMO
  • 1.2.1. Definition, motivation, and vision
  • 1.2.2. User-centric operation
  • 1.2.3. Uplink and downlink system model
  • 1.2.4. Network scalability
  • 1.2.5. Uplink operation with different cooperation levels
  • 1.2.6. Downlink operation with different cooperation levels
• Session 2: Cell-Free Massive MIMO in Virtualized C-RAN Architecture (75 mins)
  • 2.1. C-RAN architecture
  • 2.2. Fronthaul transport technologies for Cell-Free Massive MIMO
  • 2.3. End-to-end power consumption modeling with different functional splits
  • 2.4. Joint resource allocation with QoS-awareness for end-to-end power minimization
  • 2.5. Joint total data rate maximization and energy minimization

T5. Visual Object Tracking

Dr. Hakan Çevikalp

• 1.1. Introduction
• 1.1.1. Importance of object tracking in computer vision
• 1.1.2. Real-world applications
• 1.1.2.1. Surveillance
• 1.1.2.2. Autonomous driving
• 1.1.2.3. Robotics
• 1.1.2.4. Augmented reality
• 1.2. Overview of tracking paradigms
• 1.2.1. Single Object Tracking (SOT)
• 1.2.2. Multiple Object Tracking (MOT)
• 1.3. Key challenges in object tracking
• 1.3.1. Occlusions
• 1.3.2. Appearance changes
• 1.3.3. Real-time processing requirements

• 2.1. Current methods
• 2.1.1. Deep learning approaches
• 2.1.2. Correlation filters
• 2.1.3. Siamese networks
• 2.2. Comparison of SOT and MOT
• 2.2.1. Methodological and practical differences
• 2.2.2. Commonalities and related techniques
• 2.3. Developments in the field
• 2.3.1. Recent advances in tracking algorithms
• 2.3.2. Evaluation metrics
• 2.3.3. Benchmark datasets
• 2.4. Future directions
• 2.4.1. Open research problems
• 2.4.2. Impact on real-world applications

T6. Biomedical Image Reconstruction and Image Processing Algorithms

Dr. Emine Ülkü Sarıtaş (Bilkent University)

•   1.1. Biomedical imaging modalities
•   1.2. Difference between image reconstruction and image processing
•   1.3. Image quality measurement metrics
•   1.4. Multidimensional Fourier transform and systems
•   1.5. Radon transform and Fourier slice theorem
•   1.6. Sampling in multidimensional space

•   2.1. Inverse problems
•   2.2. Iterative methods
•   2.3. Reconstruction algorithms for undersampling
•   2.4. Reconstruction algorithms for compressed sensing
•   2.5. Deep learning-based methods

•   3.1. Noise removal
•   3.2. Super-resolution
•   3.3. Artifact correction
•   3.4. Deep learning-based methods

T7.a. Artificial Intelligence Applications in Model-Based Design

Sebahattün Babür (FİGES)

Artificial Intelligence (AI) applications in Model-Based Design (MBD) accelerate the design cycle and increase accuracy by integrating a data-driven intelligence layer into the system development process. After modeling physical systems with tools like Simulink and Simscape, AI-supported decision mechanisms can be integrated into these models to develop smart controllers, predictive maintenance algorithms, and autonomous system behaviors. With deep learning, machine learning, and statistical analysis techniques, optimization based on real-world data becomes possible, and the developed AI models can be applied both in simulation environments and real-time test systems. Thus, the MBD process is not limited to modeling physical behaviors but enables systems to become more flexible and predictable through learning and adaptation capabilities.

T7.b. 5G O-RAN Modeling and Simulation with MATLAB

Fatih Genç (FİGES)

Open Radio Access Network (O-RAN) presents a flexible, vendor-independent, and software-based architecture in 5G systems, distinguishing it from traditional RAN structures. This study focuses on modeling and simulating the 5G O-RAN architecture in the MATLAB environment. Using MATLAB’s 5G Toolbox and Simulink support, the simulation infrastructure enables modeling O-RAN components such as Central Unit (CU), Distributed Unit (DU), and Radio Unit (RU) under various functional split scenarios (e.g., 7.2x and 8). Additionally, standardized fronthaul interfaces and control plane signaling are considered to analyze performance metrics such as latency, data rate, and reliability. The study also demonstrates the integration of AI and machine learning algorithms developed in MATLAB into the RAN Intelligent Controller (RIC) component. This setup provides a valuable tool for both academic and industrial research by enabling controlled and repeatable testing of the 5G O-RAN architecture.