ASEN 6519

Advanced Survey of Sequential Decision Making

Is Q-learning guaranteed to converge?

Under what conditions?

Which is more sample-efficient? Model-based or Model-free Reinforcement Learning?

How much computational power did AlphaGo use?

Do you need less for chess?

How much more do you need to learn StarCraft?

Learning Objectives

  1. Broad familiarity with range of algorithms
  2. Theoretical tools
  3. Communication (hard!)
  4. Open research topics (know and investigate)

How?

(Assignments)

  • Presentations and in-class discussions (3 each)
  • Short-Answer Quizzes
  • Final Project

Schedule

Logistics: Survey

Will send out tomorrow; due by Midnight Thursday

Logistics: Paper Access

Advice: How to read a paper

Optional Presentation Template

Presentation Tips

  • Tell the story of the paper - why does it matter? People remember stories, and we will all be better off if you can make it engaging.
  • Use precision in your language (this takes practice)
  • Practice
  • Understand the important parts so that you can answer questions, and be honest if you don't
  • Information density is crucial - use appear animations to guide attention

Auditing

Since the course relies on student participation, I prefer for students to enroll rather than audit.

If you intend to audit:

  • Please email me.
  • I request that you contribute a paper presentation.

Quick Intro

  • Name
  • Station (e.g. third year grad student in AERO)
  • What you want to learn about
  • Background
  • Research focus

DMU++ Intro

By Zachary Sunberg

DMU++ Intro

  • 393