Achieving quantum utility and quantum advantage through AI-based approaches and efficient data driven models

With the advent of Deep Learning, Artificial Intelligence models have achieved state-of-the-art results across various applications, including natural language processing and image recognition. However, the vast number of parameters in these models makes them challenging to train. This difficulty is further exacerbated by the immense volume of data required for training and the significant computational resources needed to execute such algorithms. Developing foundational Deep Learning models, for example, can take months of training time and cost millions of dollars, restricting their development to large, specialized organizations and limiting their applicability to real-world problems. The aim of this thesis is to investigate Quantum Computation and Information Processing within the Data-Driven paradigm. The primary objective is to develop more efficient and effective quantum learning algorithms that can address and overcome the limitations of classical techniques. Central to this endeavor are Variational Quantum Algorithms and Quantum Neural Networks, which are anticipated to generalize faster and converge with fewer training samples or iterations compared to their classical counterparts. These quantum approaches offer significant advantages in managing high-dimensional datasets, where classical Deep Learning models often become computationally prohibitive. Furthermore, they provide quantum utility by being immediately applicable to current quantum devices while laying the groundwork for achieving quantum advantage as quantum technology continues to advance. The contributions of this thesis also include the development of novel methodologies for optimizing quantum algorithms such as the Quantum Approximate Optimization Algorithm, which embodies quantum utility and is a promising candidate for achieving quantum advantage. Finally, this work establishes frameworks for the efficient training of Quantum Neural Networks and creates hybrid models that bridge quantum and classical computing paradigms, thereby enhancing quantum utility. These advancements have profound implications for the quantum computing community, which has only recently begun to synergize the potential of quantum technologies with classical algorithms in a hybrid fashion. Moreover, this thesis ensures that its findings are accessible to a broader Quantum Machine Learning audience while complementing related results from the classical Machine Learning field. By achieving quantum utility and striving for quantum advantage, this research paves the way for more resource-efficient and effective quantum-based algorithms, significantly advancing the field of Artificial Intelligence.