UmbrellaRL ☂️

Generic solution for reinforcement learning problems ☂️

Version 1.0.0 will be a full tabular solution method release. See below for currently implemented & planned features.

UmbrellaRL is intended to ease some of the requirements for working with reinforcement learning problems. By providing logic & mathematics, the UmbrellaRL agent only requires an environment to interact with. This allows for faster solution methods to reinforcement learning problems.

Implementations

• Project Specific

  • Graphing functions
    • plot state space rewards
    • plot learning rate

• Tabular Solution Methods

  • Dynamic programming
    • Policy Evaluation
    • Policy Improvement
    • Policy Iteration
      • Jack's Car Rental
    • Value Iteration
    • Generalised Policy Iteration
  • Monte Carlo Methods
  • Temporal Difference Learning
  • n-step Bootstrapping

• Approximate Solution Methods

Usage

UmbrellaRL comes with abstract Agent, Environment & Policy classes. These abstract classes are intended to be inherited from to define the solution to your problem. UmbrellaRL also comes with various types to make implementation easier. A 'Solutions' directory is included in the package. The modules within the 'Solutions' directory are commonly found reinforcement learning problems that implement UmbrellaRL's classes & types. The modules within the 'Solutions' directory could also function as tutorials.