</> Code with Papers by Syntology

Methods › Reinforcement Learning › Policy Gradient Methods › PPO

Proximal Policy Optimization

PPO

949 papers tagged archive 2025-07-28

Introduced by John Schulman et al. in Proximal Policy Optimization Algorithms

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

Proximal Policy Optimization, or PPO, is a policy gradient method for reinforcement learning. The motivation was to have an algorithm with the data efficiency and reliable performance of TRPO, while using only first-order optimization.

Let rₜ(θ) denote the probability ratio rₜ(θ) = (π_θ(aₜ|sₜ))/(π_(θ_(old))(aₜ|sₜ)), so r(θ_(old)) = 1. TRPO maximizes a “surrogate” objective:

L^(CPI)(θ) = 𝔼̂ₜ[(π_θ(aₜ|sₜ))/(π_(θ_(old))(aₜ|sₜ)))Âₜ] = 𝔼̂ₜ[rₜ(θ)Âₜ]

Where CPI refers to a conservative policy iteration. Without a constraint, maximization of L^(CPI) would lead to an excessively large policy update; hence, we PPO modifies the objective, to penalize changes to the policy that move rₜ(θ) away from 1:

J^(CLIP)(θ) = 𝔼̂ₜ[min(rₜ(θ)Âₜ, clip(rₜ(θ), 1-ϵ, 1+ϵ)Âₜ)]

where ϵ is a hyperparameter, say, ϵ= 0.2. The motivation for this objective is as follows. The first term inside the min is L^(CPI). The second term, clip(rₜ(θ), 1-ϵ, 1+ϵ)Âₜ modifies the surrogate objective by clipping the probability ratio, which removes the incentive for moving rₜ outside of the interval [1 − ϵ, 1 + ϵ]. Finally, we take the minimum of the clipped and unclipped objective, so the final objective is a lower bound (i.e., a pessimistic bound) on the unclipped objective. With this scheme, we only ignore the change in probability ratio when it would make the objective improve, and we include it when it makes the objective worse.

One detail to note is that when we apply PPO for a network where we have shared parameters for actor and critic functions, we typically add to the objective function an error term on value estimation and an entropy term to encourage exploration.

PaperSource

Papers archive 2025-07-28

30 shown of 949, 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 357 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 Learning (RL)314
reinforcement-learning274
Autonomous Driving249
Reinforcement Learning242
Deep Reinforcement Learning157
Autonomous Vehicles83
Decision Making65
Imitation Learning60
Object Detection55
object-detection51
Continuous Control41
MuJoCo41
continuous-control37
Semantic Segmentation34
Language Modelling33
Multi-agent Reinforcement Learning28
Q-Learning26
Language Modeling24
OpenAI Gym22
Data Augmentation21

Usage over time archive 2025-07-28

Papers per year tagged with PPO: 2017 to 2025, peak 232 232 0 2017: 4 papers 2017 2018: 21 papers 2018 2019: 50 papers 2019 2020: 95 papers 2020 2021: 113 papers 2021 2022: 140 papers 2022 2023: 152 papers 2023 2024: 232 papers 2024 2025: 142 papers 2025
Papers per year the archive tags with this method, by the paper's archive date (949 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