Explore the innovative SciPhy RL method for solving stochastic optimal control problems in continuous time during this 48-minute talk from the Toronto Machine Learning Series. Delve into the application of neural networks to solve the 'soft HJB equation', a generalization of the classical Hamilton-Jacobi-Bellman (HJB) equation. Gain insights from Igor Halperin, AI Research Associate at Fidelity Investments, as he presents numerical examples demonstrating SciPhy RL's performance in high-dimensional optimal control tasks and discusses its potential applications. Learn from Halperin's extensive experience in statistical and financial modeling, including his work in option pricing, credit portfolio risk modeling, and portfolio optimization.
Introduction to SciPhy Reinforcement Learning - Part 1
Toronto Machine Learning Series (TMLS) via YouTube
Overview
Syllabus
Introduction to SciPhy Reinforcement Learning 1
Taught by
Toronto Machine Learning Series (TMLS)
Related Courses
-
Nonconvex Homogenization for 1D Controlled Brownian Motion in a Random Potential
-
An Optimal Control Perspective on Diffusion-Based Generative Modeling
-
Towards a Theory of Exploration in Continuous-Time Reinforcement Learning
-
Finite Dimensional Approximations of Hamilton-Jacobi-Bellman Equations in Probability Measure Spaces - SIAM PDE Seminar
-
Combining Reinforcement Learning and Inverse Reinforcement Learning for Optimal Asset Allocation
-
Reinforcement Learning
