Thesis Defense - Onur Kerem Özmen (MSPHYS)

Onur Kerem Özmen – M.Sc. in Physics

Prof. Taylan Akodoğan – Advisor

Prof. Bora Işıldak –  Co. Advisor

Date: 02.07.2024

Time: 10.00

Location: AB1 245

“Utilizing Convolutional Neural Networks for Temperature-Based Analysis of Two Dimensional Ising Model Configurations”

Prof. Taylan Akdoğan, Özyeğin University

Prof. Bora Işıldak, Yıldız Technical University

Asst. Prof. Hale Sert , Istanbul University

Prof. M. Burçin Ünlü, Özyeğin University

Prof. Fatih Uğurdağ, Özyeğin University

Abstract                                                      

This study investigates the application of Convolutional Neural Networks (CNNs) to classify data associated with complex physical systems, specifically focusing on the Two-Dimensional Ising Model and its temperature-dependent ferromagnetic behaviour. The motivation behind this work is to leverage CNNs for distinguishing between phases of the system under different temperature regions: below critical, at critical, and above critical temperatures. Three distinct neural network architectures were implemented, and their performance was evaluated using a dataset of simulated configurations. The results demonstrate that the models achieved high accuracy in predicting the temperature-related phases, indicating the effectiveness of CNNs in capturing physical patterns. These findings contribute to understanding CNNs in physical sciences and suggest that machine learning techniques could be applied to other complex systems. Future work might focus on refining the model architectures and extending the analysis to other lattice structures.

Bio:

Onur Kerem Özmen received his B.Sc degree in Physics from the Istanbul University TEBIP program.  He is working on machine learning applications on physical systems and pursuing his M.Sc in Physics at Özyeğin University.