JPMorgan: a guide to quantum machine learning for Finance

"The impact of quantum computing on financial services will be faster than you think," said Marco Pistoia, head of quantum technology and research at JPMorgan Chase, and his team members. JPMorgan said the financial industry would benefit from quantum computing "even in the short term". This review article introduces the application status of quantum algorithms in the financial field, with special attention to the use cases that can be solved by machine learning (ML).

After continuous development, efficient ml algorithms can support different data types and expand to larger data sets. The applicability of ML in the financial field becomes more and more important.

At present, ML operations applicable to the financial industry include regression of asset pricing, classification of portfolio optimization, clustering of portfolio risk analysis and stock selection, generation modeling of market system identification, feature extraction of fraud detection, reinforcement learning of algorithmic transactions, and natural language processing (NLP) of risk assessment, financial prediction, accounting and audit. In addition, deep learning is usually used for image recognition and text classification, as well as any use case characterized by large unstructured data sets.

Considering the complexity of the algorithms involved and the size of the data analyzed, ML has been identified as one of the most important fields of quantum computing applications. Moreover, with the discovery of new linear algebraic quantum algorithms, this becomes more obvious. These algorithms provide the potential to perform linear algebraic calculations more effectively and accurately than the corresponding classical algorithms on quantum computers.

Under some conditions, quantum acceleration can even be exponential. Although there is no end-to-end application of exponentially accelerated quantum ml, several promising directions have been proposed. At the same time, a lot of research and engineering work has been successfully committed to the implementation of quantum algorithm, which has significant polynomial acceleration in its data processing subroutine (except data loading and output extraction).

Link：https://arxiv.org/pdf/2109.04298.pdf