ABMs in financial markets + ML/AI
Robert Hillman recently gave a talk on Financial Market Simulations, which can be found here.This is a great overview and intro of where things are at when it comes to financial market simulations using agent-based modeling (ABM) approaches.
Modeling financial markets started in the late 1980s (Hillman), but have been modeled successfully since the end of the 21st century. However, models were resource heavy and lacked detail. They were also difficult to create. Hillman highlights there is a more recent interest in ABMs for a couple of reasons: 1) increased computing capacity and computing approaches and 2) increased awareness of the ABM approach.
Financial markets have been modeled since the end of the 21st century, but models were heavy and lacked detail. They were also difficult to create. Hillman highlights there is a more recent interest in ABMs for a couple of reasons: 1) increased computing capacity and computing approaches and 2) increased awareness of the ABM approach.
ABMs and AI
The most exciting thing that Hillman highlights is the intersections of ABMs and machine learning / artificial intelligence (ML/AI)! ABM creators will often program learning in a crude a way, but with ML/AI, we advance the way agents learn in model — “the models will become more dynamic and contain elements of ML, shifting away in time from the currently predominantly mechanistic models we see policy makers using (Hillman 2018).”
Hillman highlights a paper by Lampert et. al that I am looking forward to digging into: Lamperti, F, Roventini, A, Sani, A (2017), Agent-Based Model Calibration using Machine Learning Surrogates
