【HRL】option-critic & FeUdal

Overview

This is the first day when I start to scribe some ideas when I read papers. Why do so? This marks the start of my making a change to the target for my career. So in this pages, I will think about two papers which fall into the field I am recently focuing on. Let’s start it.

The Optic-Critic Architecture

This paper provides the fundation of how to derive the hierarchical structure of the high-level option and termination function.

Abstract

1. Temporal abstraction is key to scaling up learning and planning in reinforcement learning.
   Directly introduce the RL.

2. While planning with temporally extended actions is well understood, creating such abstractions autonomously from data has remained challenging.
   Propose the problem

3. We tackle this problem in the framework of opti