Deep deterministic policy gradient
Algorithms AIMech:DeepDeterministicPolicyGradient DRAFT
Aliases: DDPG
An off-policy actor–critic method for continuous control using neural function approximators and deterministic policy gradients.
Identity Notes & Concept Scope
DDPG combines deterministic policy gradients with deep actor/critic approximators, replay and target networks. It is distinct from stochastic maximum-entropy SAC and discrete-action DQN.
Claims & Grounded Evidence
1 Claim
Mechanism
#learning-method
Replay samples train a critic toward bootstrapped targets and update a deterministic actor through the critic's action gradient; slowly updated target actor and critic networks supply the bootstrap targets.
⚠️ Curated Limitations & Negative Results
- Exploration in the original algorithm adds noise to the actor's action; a deterministic actor does not remove the need for exploration.
- The source's simulated-control results do not establish universal stability or performance in every continuous-action environment.
Artificial Intelligence Ontology (AIO) Grounding
Mapped AIO Classes
| AIO Target Class | Mapping Relation | Mapping Rationale |
|---|---|---|
AIO:ReinforcementLearning
Reinforcement Learning
|
related_match | This algorithm implements reinforcement-learning updates. The AIO target is a learning task, not the same method or an asserted algorithm superclass. |
Evidence Sources & Primary Literature
1 Source| Source ID | Publication / Document | Type / Access | Reference & Checksum |
|---|---|---|---|
original-paper
|
Timothy P. Lillicrap, Jonathan J. Hunt, Alexander Pritzel, Nicolas Heess, Tom Erez, Yuval Tassa, David Silver, Daan Wierstra
|
primary paper
full_text
|
ID: arXiv:1509.02971v6
sha256:34ef677ecdee…
|
Curation Provenance
Curator: Codex
Method: ai_assisted
Model: GPT-6
Created: 2026-10-11T03:44:41Z
Updated: 2026-10-11T03:44:41Z
Source File:
data/curated/DeepDeterministicPolicyGradient.yaml