1/4

UCL / DeepMind RL Lecture 6: Value Function Approximation & Deep Q-Networks (DQN)

David Silver UCL / Google DeepMind

Principal Research Scientist at DeepMind & Professor at UCL

1:36:45·2015·graduate level

Diagnoses the 'deadly triad' of function approximation + bootstrapping + off-policy training, and shows how experience replay and fixed target networks stabilized Atari DQN.

AI & ML Reinforcement Learning Deep Q-Networks Function Approximation Deadly Triad

#Reinforcement Learning Course by David Silver# Lecture 6: Value Function Approximation

#Slides and more info about the course: http://goo.gl/vUiyjq

Previous · Lecture 5

UCL / DeepMind RL Lecture 5: Model-Free Control — ε-Greedy SARSA & Off-Policy Q-Learning

Next · Lecture 7

RL Course by David Silver - Lecture 7: Policy Gradient Methods

Seminar notes & discussion

Sign in to comment

No public seminar notes on this lecture yet.

Part of course10 lectures

COMP0089: Introduction to Reinforcement Learning

Prof. David Silver

Foundational reinforcement learning course taught by AlphaGo lead researcher David Silver: Markov Decision Processes, Bellman optimality, dynamic programming, Monte Carlo, TD learning, and function approximation.

View full syllabus →

Related lectures

RL Course by David Silver - Lecture 7: Policy Gradient Methods

David Silver · UCL / Google DeepMind

Next in COMP0089 · Lecture 7

UCL / DeepMind RL Lecture 5: Model-Free Control — ε-Greedy SARSA & Off-Policy Q-Learning

David Silver · UCL / Google DeepMind

Shared subfield · Reinforcement Learning

UCL / DeepMind RL Lecture 1: Introduction to Reinforcement Learning

David Silver · UCL / Google DeepMind

Shared subfield · Reinforcement Learning

UCL / DeepMind RL Lecture 2: Markov Processes, Reward Processes & Bellman Equations

David Silver · UCL / Google DeepMind

Shared subfield · Reinforcement Learning

Lecture 9: Exploration and Exploitation (COMP0089)

David Silver · UCL / Google DeepMind

Shared subfield · Reinforcement Learning

UCL / DeepMind RL Lecture 4: Model-Free Prediction — Monte Carlo, TD(0) & TD(λ)

David Silver · UCL / Google DeepMind

Shared subfield · Reinforcement Learning