FRT welcomes Dr. Bill Kahn, leading data scientist and financial modeling executive, formerly at Bank of America where he led the IIF Machine Learning Working Group throughout the last year. Bill joins us to discuss Machine Learning, demystifying key misconceptions in terms of its risks, as well as discussing the special complexity of interpretability. Bill also spoke on this topic at our recent IIF Digital Finance Symposium, the top takeaways of which were summarized on FRT Episode 37.
Bill relates that while at times some of the technical details, like overfitting, heterogeneity, and independence can matter, generally those are not the big mistakes. Rather he equates the big mistakes in machine learning as fundamentally the same as the big mistakes of all multiple regression. Bill discusses overfitting and heterogeneity, before delving into the topic of interpretability, where he differentiates between “what a model is doing” versus “how is the model doing it”. The discussion also includes a view of the special concerns related to the use of machine learning.
Previous FRT episodes can be found here, You can also now find FRT on iTunes or Apple Podcast, Google Play, Spotify, as well as continuing to be available on SoundCloud and the IIF website.
Looking ahead, we will debrief the IIF Roundtable on Machine Learning in the Financial Industry, as well as discussing the emergence of Sweden’s ‘Cashless economy’ with SEB CEO Johan Torgeby.