Methods › Reinforcement Learning › Policy Gradient Methods › A3C

A3C

57 papers tagged archive 2025-07-28

Introduced by Volodymyr Mnih et al. in Asynchronous Methods for Deep Reinforcement Learning

archive 2025-07-28 Description, source and code snippet are the archive's method entry.

A3C, Asynchronous Advantage Actor Critic, is a policy gradient algorithm in reinforcement learning that maintains a policy π(aₜ|sₜ; θ) and an estimate of the value function V(sₜ; θᵥ). It operates in the forward view and uses a mix of n-step returns to update both the policy and the value-function. The policy and the value function are updated after every tₘₐₓ actions or when a terminal state is reached. The update performed by the algorithm can be seen as ∇_(θ′)logπ(aₜ|sₜ; θ′)A(sₜ, aₜ; θ, θᵥ) where A(sₜ, aₜ; θ, θᵥ) is an estimate of the advantage function given by:

∑ᵏ⁻¹ᵢ₌₀γⁱrₜ₊ᵢ + γᵏV(sₜ₊ₖ; θᵥ) - V(sₜ; θᵥ)

where k can vary from state to state and is upper-bounded by tₘₐₓ.

The critics in A3C learn the value function while multiple actors are trained in parallel and get synced with global parameters every so often. The gradients are accumulated as part of training for stability - this is like parallelized stochastic gradient descent.

Note that while the parameters θ of the policy and θᵥ of the value function are shown as being separate for generality, we always share some of the parameters in practice. We typically use a convolutional neural network that has one softmax output for the policy π(aₜ|sₜ; θ) and one linear output for the value function V(sₜ; θᵥ), with all non-output layers shared.

PaperSource

Papers archive 2025-07-28

30 shown of 57, newest first. Repository counts are the archive's code-links table. A Syntology line states what Syntology ran from that paper's harvested code; it is per sample and not a correctness claim.

Tasks archive 2025-07-28

20 shown of 61 tasks the archive attaches to papers tagged with this method, by distinct papers. A task without a page in the catalog is plain text.

TaskPapers
Reinforcement Learning40
Reinforcement Learning (RL)40
reinforcement-learning40
Deep Reinforcement Learning26
Atari Games12
Q-Learning8
Decision Making5
Scheduling5
Autonomous Driving3
Multi-agent Reinforcement Learning3
Representation Learning3
CPU2
Continuous Control2
GPU2
MuJoCo2
OpenAI Gym2
Partially Observable Reinforcement Learning2
Problem Decomposition2
Segmentation2
continuous-control2

Usage over time archive 2025-07-28

Papers per year tagged with A3C: 2016 to 2025, peak 11 11 0 2016: 2 papers 2016 2017: 11 papers 2017 2018: 5 papers 2018 2019: 10 papers 2019 2020: 8 papers 2020 2021: 3 papers 2021 2022: 4 papers 2022 2023: 4 papers 2023 2024: 7 papers 2024 2025: 3 papers 2025
Papers per year the archive tags with this method, by the paper's archive date (57 dated). Bars are counts, not a trend claim.

Components: the archive holds no method-to-method composition, so PwC's Components table cannot be rebuilt; the Papers list carries no Results column for the same reason (the archive does not join its leaderboard rows to method tags).

Categories archive 2025-07-28

Policy Gradient Methods

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections