Advanced
Open
Pro
Bandit or Reinforcement Learning for Surge Pricing?
Part of the ML System Design Interview path →
Part of the Reinforcement Learning & Long-term Optimization path →
Part of the Advanced Data Science & Experimentation path →
A staff engineer argues that surge pricing is obviously a sequential decision problem (prices now change where drivers are in ten minutes) and that you should build a full reinforcement-learning system. A data scientist argues for a contextual bandit.
- Argue for the bandit as the first production system, naming the specific costs of RL here.
- Say what evidence would make you move to RL.
- Describe the exploration budget you would put on the bandit, and what its cost looked like in the lesson's simulation.
Share this question