In the last couple of years, applications of Quantum Computing have expanded across multiple industries, in particular finance, healthcare, telecommunications, cybersecurity and logistics among many others.
In fields such as Machine Learning, Artificial Intelligence, Combinatorial Optimization and Molecular Simulation, problems are often very complex and require a great amount of computational power to be solved. Very often, this complexity prevents organisations from tackling these problems altogether and allows only for heuristic approaches. In other cases, the results need to be improved. The reasons might be manifold, namely the task must be tackled quickly, in a limited amount of time and thus only suboptimal approaches can be used, complexity is so high that state-of-the-art algorithms are not able to yield the best solutions possible or the computational requirements are so heavy that no modern technology is able to handle it.
In recent years, machine learning (ML) has become a fundamental tool for extracting value out of data, enabling businesses to benefit greatly from data-driven products. However, classical ML models often suffer from generalization issues and being driven by the need for ever more precise predictions, they become more and more complex, data-hungry, and computationally expensive. This is where Quantum Computing sets itself as a potential game-changer, promising improved performance and better generalization when compared to existing classical ML techniques.
This new field, called quantum Machine Learning (qML), has a lot of potential as quantum computers are currently undergoing major improvements both in terms of computational power and robustness. General applications to Machine Learning are projected to be within reach in a couple of years.
With the quantum hardware and quantum development kits available today, it is already possible to run the first quantum Machine Learning algorithms. In Reply we are interested in exploring the business applications of these algorithms, therefore we applied them to some datasets that are similar to real-world data from our use-cases.