Lecture 4b - Multi-Arm Bandits | Reasoning LLMs from Scratch
In this lecture, we look at the category of multi-arm bandit problems. Bandit problems are the first step towards understanding how agents learn from interaction.
They offer a simplified setting to understand the meaning of exploration and exploitation.
We understand the core intuition behind bandit problems and learn about ways to balance exploration and exploitation.
Link to the Google Colab Notebook for implementation of Multi-Arm Bandit Problems: https://colab.research.google.com/dri...