PRODUCTION SCENARIO
A logistics firm is training a picking robot in a simulated warehouse. There is no dataset of correct moves: the robot earns points for fast, undamaged picks, loses points for collisions, and improves over thousands of attempts.
Which machine learning approach describes this training?
Answering here is anonymous. Nothing is saved unless you sign in.
Show answer and explanation
Answer: Reinforcement learning, where an agent improves through trial and error
Reinforcement learning is learning by doing: the agent attempts a task, receives positive reinforcement when it performs well and negative reinforcement when it performs poorly, and improves over many attempts. No labeled examples of correct moves exist here, so this is not supervised learning.
Official Source: https://cloud.google.com/learn/what-is-machine-learning ↗