Reinforcement Learning Hub
A repo dedicated to all things reinforcement learning (RL). Here, you’ll find a collection of essential resources including papers, talks, lectures and code. (maintained by Zelal “Lain” Mustafaoglu)
Topics
- Value-based RL
- Markov Decision Processes (MDPs): The foundational framework for RL, including states, actions, rewards, and transitions.
- Q-learning: A model-free value-based method for learning action-value functions.
- Model-based RL: Techniques that assume a known model or learn a model of the environment.
- Value Iteration and Policy Iteration: Dynamic programming methods that use Bellman equations to compute optimal policies, typically under a known environment model.
- World models
- Model-free RL:
- Deep reinforcement learning
- Deep Q Networks (DQNs)
- Policy gradient methods
- On-policy RL:
- REINFORCE
- Advanced policy gradient methods
- REINFORCE with value function baseline
- Actor-Critic Methods
- Off-policy RL:
- On-policy RL:
- Deep reinforcement learning
- Inverse reinforcement learning: Methods that infer the reward function from observed behavior, allowing agents to learn by imitation
- Reinforcement learning with human feedback (RLHF)
- Evolutionary reinforcement learning
Papers
Deep Reinforcement Learning
More information on deep RL here
Policy Gradient Methods
More information on policy gradient methods here
- Proximal Policy Optimization Algorithms (Schulman et al. 2017) link
- Trust Region Policy Optimization (Schulman et al. 2015) link
- Algorithms for Inverse Reinforcement Learning (Ng and Russell 2000) link
- Asynchronous Methods for Deep Reinforcement Learning (A2C) (Mnih et al. 2016) link
- A3C (Mnih et al. 2016) link
- R^2: Fast Reinforcement Learning via Slow Reinforcement Learning (Duan et al. 2016) link
- Policy Gradient Methods for Reinforcement Learning with Function Approximation (Sutton et al. 1999) link
- High-Dimensional Continuous Control Using Generalized Advantage Estimation (Schulman et al. 2018) link
- Equivalence Between Policy Gradients and Soft Q-Learning (Schulman et al. 2016) link
- More to be added
Miscallenous Papers
- #Exploration: A Study of Count-Based Exploration for Deep Reinforcement Learning link
- Benchmarking Deep Reinforcement Learning for Continuous Control link / code
- MuJoCo: A physics engine for model-based control link
Evolutionary Reinforcement Learning
More information on evolution here
- Evolution Strategies as a Scalable Alternative to Reinforcement Learning by OpenAI (September 2017) link
- Uber Deep Neuroevolution code
- Deep Neuroevolution: Genetic Algorithms are a Competitive Alternative for Training Deep Neural Networks for Reinforcement Learning by Uber (first paper, April 2018) link
- Improving Exploration in Evolution Strategies for Deep Reinforcement Learning via a Population of Novelty-Seeking Agents by Uber (second paper, October 2018) link
- Evolutionary Policy Optimization (Mustafaoglu et al. 2025
Reinforcement Learning with Human Feedback (RLHF)
More information on RLHF here
- Training language models to follow instructions with human feedback (Ouyang et al. 2022) paper
- Learning to summarize from human feedback (Stiennon et al. 2020) paper
- Deep Reinforcement Learning from Human Preferences (Christiano et al. 2023) paper
Resources
Background Material
- David Silver’s Reinforcement Learning Course website
- DeepRL Bootcamp website
- Pieter Abbeel’s “Foundations of DeepRL” series playlist
- John Schulman’s “Deep Reinforcement Learning via Policy Optimization” lecture (2017)
- John Schulman’s RLHF lecture (2023)
- Abbeel’s Inverse RL lecture
- ICML “Reinforcement Learning at the Hyperscale” talk (2024)