Automatic Differentiation via Effects and Handlers: An Implementation in Frank

Jesse Sigal

Automatic differentiation (AD) is an important family of algorithms which enables derivative based optimization. We show that AD can be simply implemented with effects and handlers…