Highlights From a Recent Interview on Applying Machine Learning to Long-Term Investing
This month, John Alberg – Euclidean Partner and Co-Founder – was interviewed by Corey Hoffstein as part of the Flirting With Models Podcast Series. Corey and John’s conversation covers everything from the basics of machine learning and the evolution of Euclidean's approach over the last decade, to the implications of new developments and technologies relating to neural networks.
Below, we point out several parts of the interview that you may find particularly interesting. And, in the meantime, if you are interested in the perspectives and questions that guide the machine learning research John describes here, you can tap into that information in our most recent letter to investors.
3:00 - John postulates that if value investing works, then machine learning should be able to find evidence of that fact in historical data.
7:15 – Discusses the landscape of machine learning techniques and addresses the question, “where do advanced statistical techniques end and machine learning begin?”
14:30 – Explains how the process at Euclidean has changed over time as the landscape of machine learning techniques has evolved.
16:30 – Introduces the three reasons deep learning has been interesting to Euclidean, including the opportunities to perform less factor engineering, better model sequences of data through time, and include unstructured data in Euclidean’s analysis.
25:00 – Digs into this question, “How do you gain confidence in a machine learning approach and ensure that you are not over-fitting a model?”
29:15 – Engages around the question of whether machine learning is applicable to investing given that companies, accounting and markets change, thus causing data to be non-stationary.
32:15 - John explains what ties Euclidean to the legacy of value investing.
38:15 - Corey asks John to discuss the current "p-hacking" debate and whether there are techniques that traditional factor research can learn from machine learning.
47:00 - John offers some suggestions for people who are interested in learning more about machine learning.