Introduction-to-Reinforcement-Learning

  • Course Number: EE 4660
  • Year: 2020-Spring

This course introduces some basic Reinforcement Learning knowledge without complicated mathematical derivation, and focuses on implementing classic algorithms from scratch.

HW1

  • Multi-Armed Bandit

    Get familiar with basic action-value based methods in multi-armed bandit problems.

HW2

  • Grid World

    By Bellman Equation method to find optimal value function

HW3

  • Grid World

    Use dynamic programming to find an optimal policy

HW4

  • Swamp

    Monte Carlo control algorithm implementation

HW5

  • Swamp

    One step TD method, including Sarsa and Q-learning

HW6

  • Swamp

    n step TD method, mainly on 5-step Sarsa implementation

HW7

  • Swamp

    Planning and Learning with Tabular Methods, mainly on Dyna-Q algorithm

HW8

  • Mountain Car

    Control with Function Approximation method