Baike.dev
All toolsAI codingTrendingOpen sourceNewsSubmit
Log in
< Back to tools
R

RL-Taxonomy

> 编程语言
Open source

Loose taxonomy of reinforcement learning algorithms

197 stars0 likes0 views
WebsiteGitHub

About

Loose taxonomy of reinforcement learning algorithms

RL Taxonomy

This is a loose taxonomy of reinforcement learning algorithms. I'm by no means expert in this area, I'm making this as part of my learning process. Note that there are a lot more algorithms than listed here, and often I don't even know how to categorize them. In any case, please PR to correct things or suggest new stuff.

Note that this file is generated by taxonomy.py.

Table of Contents:

Taxonomy

Reinforcement Learning

  • Model Free
    • Value Gradient
      • SARSA
      • Q-learning
      • TD-Gammon
      • DQN
      • DRQN
      • DDQN
      • PER
      • Duelling-DQN
      • QR-DQN
      • C51
      • RAINBOW
      • DQN+HER
      • IQN
      • APE-X DQN
      • R2D2
      • NGU
      • Agent57
    • Policy Gradient/Actor-Critic
      • REINFORCE
      • DPG
      • DDPG
      • TRPO
      • GAE
      • A3C
      • DDPG+HER
      • MADDPG
      • A2C
      • ACER
      • ACKTR
      • PPO
      • SVPG
      • Reactor
      • D4PG
      • APE-X DDPG
      • SAC
      • TD3
      • MPO
      • IMPALA
  • Model Based
    • Dyna-Q
    • Prioritized Sweeping
    • MCTS
    • PILCO
    • I2A
    • MBMF
    • Exit
    • AlphaZero
    • MVE
    • STEVE
    • ME-TRPO
    • MB-MPO
    • World Models
    • PETS
    • PlaNet
    • SimPLe
    • MuZero
  • Meta-RL
    • DMRL
    • RL^2
    • MAML
    • SNAIL
    • ProMP

Taxonomy

Below is the taxonomy of reinforcement learning algorithms. Solid line indicates some progression from one idea to another. Dashed line indicates a loose connection. On the bottom you can see the timeline of the publication year of the algorithms.

It's recommended to open the .SVG file in a new window, as hovering the mouse over the algorithm will show tooltip containing the description of the algorithm and clicking the node will open the link to its description.

Reinforcement Learning

Reinforcement learning (RL) is an area of machine learning concerned with how software agents ought to take actions in an environment in order to maximize the notion of cumulative reward [from Wikipedia]

  • Related to subsequent ideas:
    • Model Free
    • Model Based
    • Meta-RL
  • Useful links:
    • A (Long) Peek into Reinforcement Learning
    • (book) Reinforcement Learning: An Introduction - 2nd Edition - Richard S. Sutton and Andrew G. Barto
  • Videos:
    • (playlist) Introduction to Reinforcement learning with David Silver
    • (playlist) Reinforcement Learning Course | DeepMind & UCL
    • (playlist) Reinforcement Learning Tutorials
    • (playlist) Deep RL Bootcamp 2017
    • (playlist) CS885 Reinforcement Learning - Spring 2018 - University of Waterloo
    • (playlist) CS234: Reinforcement Learning | Winter 2019

Model Free

(Path: Reinforcement Learning --> Model Free)

In model free reinforcement learning, the agent directly tries to predict the value/policy without having or trying to model the environment

Value Gradient

(Path: Reinforcement Learning --> Model Free --> Value Gradient)

The algorithm is learning the value function of each state or state-action. The policy is implicit, usually by just selecting the best value

SARSA

(Path: Reinforcement Learning --> Model Free --> Value Gradient --> SARSA)

SARSA (State-Action-Reward-State-Action) is an on-policy TD control method

  • Paper: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.17.2539&rep=rep1&type=pdf
  • Authors: G. A. Rummery, M. Niranjan
  • Year: 1994
  • Flags:
    • On-Policy (ONP)
    • Discrete action space (DA)
  • Related to prior idea:
    • Value Gradient
Q-learning

(Path: Reinforcement Learning --> Model Free --> Value Gradient --> Q-learning)

Q-learning an off-policy TD control method. Unlike SARSA, it doesn't follow the policy to find the next action but rather chooses most optimal action in a greedy fashion

  • Paper: http://www.cs.rhul.ac.uk/~chrisw/new_thesis.pdf
  • Authors: Chris Watkins
  • Year: 1989
  • Flags:
    • Off-Policy (OFP)
    • Discrete action space (DA)
  • Related to prior idea:
    • Value Gradient
  • Related to subsequent idea:
    • DQN
  • Useful links:
    • Diving deeper into Reinforcement Learning with Q-Learning
    • Simple Reinforcement Learning with Tensorflow Part 0: Q-Learning with Tables and Neural Networks
TD-Gammon

(Path: Reinforcement Learning --> Model Free --> Value Gradient --> TD-Gammon)

TD-Gammon is a model-free reinforcement learning algorithm similar to Q-learning, and uses a multi-layer perceptron with one hidden layer as the value function approximator. It learns the game entirely by playing against itself and achieves superhuman level of play.

  • Paper: https://dl.acm.org/doi/10.1145/203330.203343
  • Authors: Gerald Tesauro
  • Year: 1995
  • Related to prior idea:
    • Value Gradient
DQN

(Path: Reinforcement Learning --> Model Free --> Value Gradient --> DQN)

Deep Q Network (DQN) is Q-Learning with deep neural network as state-action value estimator and uses a replay buffer to sample experiences from previous trajectories to make learning more stable.

  • Paper: https://arxiv.org/abs/1312.5602
  • Authors: Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, Martin Riedmiller
  • Year: 2013
  • Flags:
    • Off-Policy (OFP)
    • Continuous state space (CS)
    • Discrete action space (DA)
    • Replay Buffer (RB)
  • Related to prior idea:
    • Q-learning
  • Related to subsequent ideas:
    • DRQN
    • DDQN
    • PER
    • QR-DQN
    • C51
    • DQN+HER
    • IQN
    • APE-X DQN
    • R2D2
    • DDPG (replay buffer)
    • ACER (replay buffer)
  • Useful links:
    • (tutorial) Deep Q Learning for the CartPole
    • An introduction to Deep Q-Learning: let’s play Doom
DRQN

(Path: Reinforcement Learning --> Model Free --> Value Gradient --> DRQN)

Deep Recurrent Q-Learning. Adding recurrency to a Deep Q-Network (DQN) by replacing the first post-convolutional fully-connected layer with a recurrent LSTM

  • Paper: https://arxiv.org/abs/1507.06527
  • Authors: Matthew Hausknecht, Peter Stone
  • Year: 2015
  • Flags:
    • Off-Policy (OFP)
    • Continuous state space (CS)
    • Discrete action space (DA)
    • Replay Buffer (RB)
    • Recurrent Neural Network (RNN)
  • Related to prior idea:
    • DQN
DDQN

(Path: Reinforcement Learning --> Model Free --> Value Gradient --> DDQN)

Double DQN adds another neural network, making separate network for policy and target. The target network is only updated after certain number of steps/episodes. This makes the learning more stable.

  • Paper: https://arxiv.org/abs/1509.06461
  • Authors: Hado van Hasselt, Arthur Guez, David Silver
  • Year: 2015
  • Flags:
    • Off-Policy (OFP)
    • Continuous state space (CS)
    • Discrete action space (DA)
  • Related to prior idea:
    • DQN
  • Related to subsequent ideas:
    • Duelling-DQN
    • RAINBOW
    • TD3 (double Q-learning)
  • Useful links:
    • (tutorial) Deep Q Learning for the CartPole
PER

(Path: Reinforcement Learning --> Model Free --> Value Gradient --> PER)

Prioritized Experience Replay (PER) improves data efficiency by replaying transitions from which there is more to learn more often

  • Paper: https://arxiv.org/abs/1511.05952
  • Authors: Tom Schaul, John Quan, Ioannis Antonoglou, David Silver
  • Year: 2015
  • Flags:
    • Off-Policy (OFP)
    • Continuous state space (CS)
    • Discrete action space (DA)
    • Replay Buffer (RB)
  • Related to prior idea:
    • DQN
  • Related to subsequent idea:
    • RAINBOW
Duelling-DQN

(Path: Reinforcement Learning --> Model Free --> Value Gradient --> Duelling-DQN)

Duelling DQN represents two separate estimators: one for the state value function and one for the state-dependent action advantage function. The main benefit of this factoring is to generalize learning across actions without imposing any change to the underlying reinforcement learning algorithm.

  • Paper: https://arxiv.org/abs/1511.06581
  • Authors: Ziyu Wang, Tom Schaul, Matteo Hessel, Hado van Hasselt, Marc Lanctot, Nando de Freitas
  • Year: 2016
  • Flags:
    • Off-Policy (OFP)
    • Continuous state space (CS)
    • Discrete action space (DA)
  • Related to prior idea:
    • DDQN
  • Related to subsequent idea:
    • RAINBOW
QR-DQN

(Path: Reinforcement Learning --> Model Free --> Value Gradient --> QR-DQN)

Distributional Reinforcement Learning with Quantile Regression (QR-DQN). In QR-DQN, distribution of values values are used for each state-action pair instead of a single mean value

  • Paper: https://arxiv.org/abs/1710.10044
  • Authors: Will Dabney, Mark Rowland, Marc G. Bellemare, Rémi Munos
  • Year: 2017
  • Flags:
    • Off-Policy (OFP)
    • Continuous state space (CS)
    • Discrete action space (DA)
    • Replay Buffer (RB)
    • Distributional (DI)
  • Related to prior idea:
    • DQN
  • Related to subsequent idea:
    • RAINBOW
  • Useful links:
    • (GitHub) Quantile Regression DQN
C51

(Path: Reinforcement Learning --> Model Free --> Value Gradient --> C51)

C51 Algorithm. The core idea of Distributional Bellman is to ask the following questions. If we can model the Distribution of the total future rewards, why restrict ourselves to the expected value (i.e. Q

Issues· 0 open

View all issuesOpen on GitHub

No open issues yet, or sync has not completed.

> Tags

Python

No comments yet. Be the first to share.

> Details

PublishedAug 1, 2026
UpdatedSep 18, 2026
Category编程语言
PricingOpen source

> Related tools

T
TypeScript
JavaScript 的超集,为前端与全栈提供静态类型
P
Python
通用编程语言,广泛用于 Web、数据与 AI
G
Go
Google 推出的简洁高效系统语言