In this thesis, a study is conducted on the performance of different types of neural networks, which include classical neural networks, quantum neural networks, and hybrid quantum-classical neural networks (hybrid quantum) in the framework of physics-informed neural networks. Several algorithms are introduced for different types of neural networks, which have taken on the task of supervised learning and unsupervised learning to solve nonlinear partial differential equations in the architecture of deep learning models. This research eva‎luates the performance of each type of network for solving the equation of motion of the damped oscillator, solving the Einstein field equations, and solving the time-independent Schrodinger equation through rigorous testing on physics problems. In this way, quantum bits and gates combine to use the power and ability of emerging quantum technologies such as quantum superposition and entanglement, to investigate the potential of quantum computers in physics-informed neural network problems. On the other hand, combining quantum neural networks with classical neural networks is tested to optimize performance and accuracy against classical and quantum neural networks. This research helps to understand whether advanced computational models can smooth the way to more accurate and efficient solutions to problems in the field of physics.

بهروز ميرزا

حامد بخشيان

عبيده جعفري , مسلم زارعي