Schedule: Spring 2020

Lectures are Tuesday and Thursday 2:30-4:00, in 2-105

Here is a tentative schedule of lectures, readings, assignments, and final project. Readings and assignments will be added as they become available. Before every lecture, you need to give feedback on the papers listed in the Readings section.

Event Date Lecture Readings Logistics
Lecture Feb 4 Questions in Intelligence None
Lecture Feb 6 Evaluating Intelligence
  1. Imitation Game
  2. Chapters 1 and 2 of this thesis
  3. Optional: Chapter 1 of this thesis
HW0: Familiarization with Infrastructure
Lecture Feb 11 Overview of RL (submit feedback)
  1. Comparing Policy-Gradient Algorithms
  2. Optional: A Tour of Reinforcement Learning: The View from Continuous Control
Release HW1
Discussion Feb 13 On-Policy RL Algorithms (submit feedback)
  1. PPO
  2. IMPALA
  3. A3C
  4. Optional: GAE
Holiday Feb 18 Monday Calendar
Discussion Feb 20 Off-Policy RL Algorithms (submit feedback)
  1. DQN
  2. DDPG
  3. Rainbow
  4. Optional: SAC
  5. Optional: TD3
 Submit Project Abstract
Discussion Feb 25 RL Applications (submit feedback)
  1. Learning Dexterity
  2. Alpha Go
  3. Alpha Zero
  4. Optional: Mu Zero
  5. Optional: Playing DOTA
Discussion Feb 27 Algorithms for Exploration (submit feedback)
  1. Curiosity-driven Exploration by self-supervised Prediction
  2. Unifying Count-Based Exploration and Intrinsic Motivation
  3. Diversity is All You Need
  4. Optional: Empowerment
  5. Optional: What is Intrinsic Motivation?
Discussion Mar 3 Transfer Learning in Context of Decision Making (submit feedback)
  1. RL2
  2. MAML
  3. Gotta Learn Fast
  4. Optional: Domain Randomization
  5. Optional: Policy Sketches
  6. Optional: Procgen
Discussion Mar 5 Curriculum Learning (submit feedback)
  1. Curriculum Learning
  2. POET
  3. Asymmetric Self-Play
  4. Optional: PowerPlay
  5. Optional: Goal GAN
  6. Optional: Teacher-Student Curriculum Learning
Lecture Mar 10 Learning Models (submit feedback)
  1. MOSAIC
  2. Supervised Learning with Distal Teacher
  3. DYNA
Discussion Mar 12 Papers on Learning Models (submit feedback)
  1. Hindsight Experience Replay
  2. Visual Foresight
  3. Optional: Learning to Poke by Poking
  4. Optional: Embed to Control
  5. Optional: World Models
Holiday Mar 17 COVID-19 Release HW2; HW1 due
Holiday Mar 19 COVID-19
Holiday Mar 24 Spring Break
Holiday Mar 26 Spring Break
Presentation Mar 31 Midterm Project Presentations
Discussion Apr 2 Neural Network Architectures (submit feedback)
  1. Relational Deep Reinforcement Learning
  2. Attention is All You Need (Transformer)
  3. Optional: Memory Networks
  4. Optional: PathNet
  5. Optional: Parameter Superposition
  6. Optional: Randomly Wired Networks
Discussion Apr 7 Representation Learning (submit feedback)
  1. Intelligence Without Representations
  2. Survey of Self-Supervised Learning
  3. Optional: Simple Framework for Contrastive Learning
  4. Optional: Tutorial on VAEs
  5. Optional: Unsupervised Learning of Object Keypoints for Perception and Control
  6. Optional: Navigation using Mid-Level Priors
Lecture Apr 9 Imitation Learning (submit feedback)
  1. Is Imitation Learning the Route to Humanoid Robots?
  2. DAGGER
  3. Optional: Mirror Neurons
Discussion Apr 14 Papers on Imitation Learning (submit feedback)
  1. Zero-Shot Visual Imitation
  2. Divergence Minimization Perspective
  3. Optional: GAIL
  4. Optional: One Shot Visual Imitation Learning
  5. Optional: Deep Mimic
  6. Optional: One Shot Imitation Learning
Release HW3; HW2 due
Discussion Apr 16 Papers on Inverse RL (submit feedback)
  1. Inverse Reinfocement Learning
  2. Maximum Entropy IRL
  3. Time Contrastive Networks
  4. Optional: Guided Cost Learning
Discussion April 21 Hierarchial Learning (RL + Imitation) (submit feedback)
  1. Option-Critic Architecture
  2. FeUdal Networks for HRL
  3. Optional: Feudal Reinforcement Learning.
Discussion Apr 23 Learning from Touch, Vision and Sound (submit feedback)
  1. GelSight
  2. Learning to Grasp and Regrasp using Vision and Touch
Discussion April 28 Multi Agent Systems (submit feedback)
  1. Survey of MARL
  2. Learning with Opponent-Learning Awareness (LOLA)
  3. Useful Link for MARL papers (not a reading)
Discussion April 30 Role of Language (submit feedback)
  1. A Survey of RL Informed by Natural Language
  2. Emergence of Grounded Compositional Language in Multi-Agent Populations
Discussion May 5 Miscellaneous (submit feedback)
  1. Successor Representations
  2. Deep RL That Matters
  3. Benchmarking Model Based RL
  4. Optional Learning Complex Dexterous Manipulation with DRL and Demonstrations
Presentation May 12 Final Project Presentations

For questions or comments, email csl-staff [AT] mit [DOT] edu.

Back to 6.884 home.