Methods › General › Stochastic Optimization › RMSProp

RMSProp

introduced 2013 519 papers tagged archive 2025-07-28

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

RMSProp is an unpublished adaptive learning rate optimizer proposed by Geoff Hinton. The motivation is that the magnitude of gradients can differ for different weights, and can change during learning, making it hard to choose a single global learning rate. RMSProp tackles this by keeping a moving average of the squared gradient and adjusting the weight updates by this magnitude. The gradient updates are performed as:

E[g²]ₜ = γE[g²]ₜ₋₁ + (1 - γ) g²ₜ

θₜ₊₁ = θₜ - η/(√(E[g²]ₜ + ϵ))gₜ

Hinton suggests γ=0.9, with a good default for η as $0.001$.

Image: Alec Radford

See Code · pytorch/pytorch

Papers archive 2025-07-28

30 shown of 519, 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 353 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
Image Classification97
image-classification62
Transfer Learning60
Classification39
Deep Learning37
General Classification35
Object Detection31
Semantic Segmentation31
Neural Architecture Search26
Data Augmentation25
Stochastic Optimization20
object-detection20
Diagnostic18
Reinforcement Learning16
Segmentation14
GPU12
Prediction12
Reinforcement Learning (RL)12
Quantization10
reinforcement-learning10

Usage over time archive 2025-07-28

Papers per year tagged with RMSProp: 2015 to 2025, peak 86 86 0 2015: 3 papers 2015 2016: 9 papers 2016 2017: 20 papers 2017 2018: 33 papers 2018 2019: 49 papers 2019 2020: 83 papers 2020 2021: 86 papers 2021 2022: 65 papers 2022 2023: 66 papers 2023 2024: 79 papers 2024 2025: 26 papers 2025
Papers per year the archive tags with this method, by the paper's archive date (519 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

Stochastic Optimization

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