Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore specific relative entropy through this 54-minute conference talk that examines a refined notion of entropy emerging from the discretization of continuous-time martingales. Delve into how discrete-time settings enable meaningful comparisons via scaled entropy limits, despite continuous martingales typically having mutually singular laws with infinite relative entropy. Learn about the closed-form expression for specific relative entropy in terms of martingales' quadratic variation and discover its fascinating application to prediction markets. Understand how this framework addresses David Aldous's question about identifying the "most exciting" game—the market with highest entropy—through stochastic control theory. Examine the extension to multi-outcome games and uncover the novel connection to Monge-Ampère equations through collaborative research. Gain insights into the intersection of probability theory, information theory, and mathematical finance, with practical applications for stochastic processes and market modeling research.
