Peter Carr Brooklyn Quant Experience (BQE) Lecture Series: Cristian Homescu
NYU Students are highly encouraged to attend in person.
All other non-NYU guests are invited to attend virtually.
Title
Machine Learning in Quantitative Wealth and Investment Management (ML in QWIM): Hype Versus Reality
Abstract
It is more and more clear that, in practice, Financial Machine Learning is a rather specialized area rather than simply an application of standard Machine Learning and Financial Data.
This observation brings up a very practical question: “What is hype and reality when applying machine learning ML to quantitative investment and wealth management (QWIM)?”
This presentation delves into what appears to work well (and, respectively, not so well) when ML is used within context of QWIM, while also discussing the practical challenges (including QWIM specific challenges).
The QWIM applications are categorized into areas:
- classification and pattern recognition, such as:
- classification and partition of the investment universe
- investing based on alternative data
- text analysis of company and regulatory documents
- sentiment analysis of news and social media
- ESG (Environmental, Social and Governance) investing
- fund decomposition, inference and replication
- network analysis and clustering, such as:
- clustering-based portfolio optimization
- network-based portfolio optimization
- analysis of interconnectedness risk
- network effects on investment portfolios.
- time series forecasting, such as:
- forecasting of financial time series
- empirical asset pricing
- reinforcement learning, such as:
- pricing and hedging of financial derivatives
- optimal dynamic trading strategies
- portfolio allocation
- goals-based investing
Other practical ML applications in QWIM include:
- synthetic financial data generation
- testing investment strategies and portfolios
- factor-based investment strategies
- nowcasting
- incorporating market states and regimes into investment portfolios.
A section is dedicated to the very recent developments on AI tools based on very large pre-trained models (such as ChatGPT, GPT-4, LLaMA, Bard, Alpaca, BloombergGPT).
The presentation also describes some of the challenges, including
- lack of sufficient data
- need to satisfy privacy, fairness and regulatory requirements
- model overfitting
- causality
- explainability and interpretability
- hyperparameter tuning
Bio
Cristian Homescu is a Director in the Portfolio Analytics team within the Chief Investment Office for Global Wealth and Investment Management division of Bank of America. Previously Cristian was on the sell side, as contributor and leader of front office quant teams for various trading desks (including interest rate, FX and Commodities). His current interests are in effective and practical quantitative investment and wealth management, delivered through a combination of advanced quantitative techniques and modeling (including Machine Learning) and, respectively, high performance computational methods.
Cristian has extensive experience in computational mathematics applied to cutting edge and large-scale practical problems in finance, engineering, physics, biology and chemistry, with research published in leading scientific journals and presented at practitioner conferences. Cristian has a PhD in Applied Mathematics from Florida State University and masters in Numerical Analysis and Applied Mathematics from University of Paris XI (France) and University of Craiova (Romania).