Machine Learning in Quantitative Finance History, Theory, and Applications

Written by a senior and well-known member of the Quantitative Finance community who currently runs a research group at a major investment bank, the book will demonstrate the use of machine learning techniques to tackle traditional data science type problems – time-series analysis and the prediction of realised volatility but will also look at novel applications. For example, the Universal Approximation Theorem of Neural Networks shows that a neural network can be used to approximate any function (subject to a number of weak conditions), although how the network is trained is not given. This will be explored within the book. Specific applications will include using a trained neural network to represent market-standard volatility smile models (such as SABR) as well as complex derivative pricing. The book will also potentially look at training a network via reinforcement learning to risk manage a derivatives portfolio. Readers will be attracted by a comprehensive presentation of the techniques available, with the historical perspective providing intuitive understanding of their development, combined with a range of practical examples from the trading floor.

Key features:

• Describes modern machine learning techniques including deep neural networks, reinforcement learning, long-short term memory networks, etc.
• Provides applications of these techniques to problems within Quantitative Finance (including applications to derivatives modelling)
• Presents the historical development of the subject from MENACE to Alpha Go Zero and AlphaZero