Reinforcement Learning for Quantitative Finance


In this session, we will provide insight in on how Wall Street is eyeing reinforcement learning as an opportunity to tackle some of the most difficult machine learning problems.

Reinforcement learning (RL) attempts to influence a bot’s actions based on the future rewards earned in the complex dynamics of “approximated” reality. In the early days of Wall Street, modelers would attempt to estimate parameters of a model based on past values and popular probability frameworks. Reinforcement learning brings a lot of the mathematics together to enable quants to leverage an extremely powerful framework for working in the real world of finance, yet it comes with a set of challenges that can easily overwhelm the faint of heart.

The common challenge with reinforcement learning today is not just to implement, but how to do so with compliance in mind? Marshall Alphonso will take you on a journey of our current understanding of reinforcement learning and its potential game changing impact on finance. After Marshall, Andrew Mann will walk through specific examples where he has seen real capital at risk. His examples will include an agent based execution algorithm (implementation shortfall), a passive directional trading strategy and a multi-agent approach to portfolio management. 

About the Presenters

Marshall Alphonso

Senior – Global 5 Lead Engineer, MathWorks

Marshall Alphonso specializes in quantitative finance and is currently the global lead engineer for the top five banks. He has over 10+ years’ experience training clients at over 250 companies, including top hedge funds, banks and other financial institutions around the world.

As advisor to the CRO of McKinsey & Co. Investment Office, Marshall was responsible for the design and implementation of the fund liquidity framework, stress testing framework and a multitude of quantitative risk and investment tools, enabling evaluation of exposures for risk and attribution. Prior experience included use of artificial intelligence and advanced statistical signal processing in communication and geostationary satellite systems.

He holds a B.S. in electrical engineering and mathematics from Purdue University and an M.S. in electrical engineering from George Mason University.

Andrew Mann

Andrew Mann is the co-founder of Coinstrats, a systematic trading firm who specialize in propriety trading and market making at the high-to-mid-frequency timescale. His main focus is in the development of novel distributed machine learning algorithms applicable to noisy environments. Coinstrats runs a fully autonomous multi-agent based trading strategy across the cryptocurrency and commodity markets. Andrew previously held positions at Virtu Financial in proprietary trading, Matrix8J as a research scientist and as a PhD researcher at University College London to name a few. 

Product Focus

Registration closed