Page 56 - tech fest 2025 ב״ש חוברת תקצירים

56
Simulations of Approximation Algorithms for MAX-3SAT in Various Instances
CS-E-33
Lidor Tubul; lidortu0@gmail.com Ram Nagid; ramnagid@walla.com
Advisor: Dr. Dina Barak-Pelleg
SCE - Shamoon College of Engineering, Be’er-Sheva
The satisfiability problem (SAT) and its optimization variant, MAX-SAT, are fundamental in theoretical computer science. SAT was the first NP-complete problem. Several approximation algorithms exist for MAX-3SAT, including randomized approximation, derandomization, linear programming, and hybrid approaches. Our project implemented and compared the approximation algorithms for MAX-3SAT, evaluating their performance on random and structured industrial instances. Industrial instances exhibit a community structure, where variables belong to disjoint groups (communities). Clauses in these instances are more likely to contain variables from the same community. A simulation tool was developed to facilitate the comparisons, allowing users to analyze different algorithms on fixed instances, or to test a specific algorithm across various instance types.
Keywords: algorithms, applied probability, MAX-3SAT, simulations
Deep Learning-Based Identification of Facial Asymmetry for Medical Diagnosis
CS-E-34
Mahmoud Abu Hamd; mahmaab1@ac.sce.ac.il
Advisors: Dr. Moshe Davidian1, Prof. Shlomo Greenberg1 1SCE - Shamoon College of Engineering, Be’er-Sheva
Facial asymmetry identification is crucial for diagnosing and rehabilitating medical conditions such as facial paralysis and Bell’s palsy. Recent advancements in deep learning, particularly convolutional neural networks (CNNs), enable highly accurate and efficient object detection methods. This project explored CNN-based models for detecting facial asymmetry, comparing different network architectures to determine the most effective approach. By applying multiple CNN structures to the same dataset, we evaluated their performance and selected the optimal model for accurate and reliable asymmetry detection. In addition, this project compared different areas of the face to choose the parts that best identify facial asymmetry. Our findings contribute to the development of automated, deep learning-based diagnostic tools that can enhance medical assessment and treatment planning.
Keywords: convolutional neural networks, deep learning, facial asymmetry detection, medical image analysis