Quantum Convolutional Neural Networks for the detection of Gamma-Ray Bursts in the AGILE space mission data

Quantum computing represents a cutting-edge frontier in artificial intelligence, proposing to enhance machine learning and deep learning techniques. Quantum Machine Learning (QML) algorithms try to leverage quantum mechanics principles, such as superposition and entanglement, to solve typical machine learning problems while outperforming classical approaches. This work falls within the context of the AGILE space mission, launched in 2007 by the Italian Space Agency to study X-ray and gamma-ray phenomena. This thesis aims to investigate the potential advantage of QML methods in analyzing data from AGILE. We implemented different Quantum Convolutional Neural Networks (QCNN) that process the data collected by the satellite to detect Gamma-Ray Bursts from sky maps or light curves. We used various frameworks: TensorFlow-Quantum, Qiskit and PennyLane. We also explored several embedding techniques to represent input data as quantum states, such as angle embedding, amplitude encoding and data re-uploading. We developed also a classical approach, implemented via a Convolutional Neural Network (CNN), to compare the quantum networks with classical ones. The CNN achieved an accuracy of 98.8% on sky maps and 95.5% on light curves, employing over a million of parameters for the former and over two thousand for the latter. The quantum models reached an accuracy of 95.1% on sky maps using PennyLane and only 51 parameters, while 81% on light curves with Qiskit and 16 parameters. QCNNs can yield results comparable to those of classical machine learning, but with sensitively fewer trainable parameters. It is important to emphasize that the training times are much longer with the quantum approach. Despite this drawback, we demonstrated that the adoption of quantum models for GRBs detection in the data acquired by the AGILE instruments allows us to deal with a possible simplification of the optimization process while achieving close to state-of-the-art performance results.