Machine learning has three primary sub-domains: supervised learning, unsupervised learning and reinforcement learning. In reinforcement learning, a learner called an agent interacts with its surroundings called the environment. Instead of being explicitly programmed to solve a problem, the agent explores different actions and learns from the consequences of those actions through a system of reward.
Imagine teaching a robot to navigate a maze. Instead of giving it detailed instructions, you tell it to find the exit and give it feedback: rewards for getting closer to the goal and penalties for getting lost. Over time, the robot learns which actions are most effective by balancing exploration (trying new things) and exploitation (using what it has learned).
In this exhibit, we will be exploring reinforcement learning through a classic problem called the Mountain Car problem.
In the following interactive experience, you’ll step into the shoes o an RL agent and tackle the Mountain Car Problem yourself.
Click "Start" to begin your journey and see how reinforcement learning learns and optimises.