Reinforcement Learning in Python — Markov Decision Processes

Led by Alan Turing Simulacrum

Formalising the reinforcement learning problem — states, actions, rewards, transitions, the Markov property, value functions, and the Bellman equation.

1. Module 1

   States, Actions, Rewards and the Markov Property

   Led by Alan Turing Simulacrum

   The question

   From bandits to full reinforcement learning · what is reinforcement learning (agent, environment, state, action, reward) · Gridworld as a canonical environment · choosing rewards (reward shaping, sparse vs dense) · the Markov property (memoryless, su...

   Outcome

   Demonstrates understanding and implementation of states, actions, rewards and the markov property.

   Sub-units

   1. ○ 1.1 States, Actions, Rewards and the Markov Property Open this unit → ✓ ↻
   Begin the course → Print certificate of completion
2. Module 2

   Value Functions and the Bellman Equation

   Led by Alan Turing Simulacrum

   The question

   Value functions (state value V(s), action value Q(s,a)) · the Bellman equation for V (expectation form) · the Bellman equation for Q · deriving the Bellman equation from first principles · worked Bellman examples · the optimal value function V* and Q...

   Outcome

   Demonstrates understanding and implementation of value functions and the bellman equation.

   Sub-units

   1. ○ 2.2 Value Functions and the Bellman Equation Open this unit → ✓ ↻
   Open this module → Print certificate of completion