AI engineering glossary · Reinforcement learning
What is deadly triad?
Deadly triad: The combination of three ingredients that can make value learning diverge: function approximation, bootstrapping and off-policy training. Any two are usually safe, but all three together can make estimates blow up. Target networks and replay tricks reduce the risk.
Deadly triad: an example
A neural network trained with Q-learning-style targets on old replay data can see its Q values grow without bound.
Related terms
More Reinforcement learning terms
- Deep deterministic policy gradient
- Deep Q-network
- Deterministic policy
- Deterministic policy gradient
- Discount factor
- Distributional reinforcement learning
- Double DQN
- Double Q-learning
Where you learn it
Deadly triad is part of the AI Engineering Program, which teaches it with notes, a hands-on lab and an exam. The first lesson is free.