- Notes on gradient descent, Toussaint 2012. pdf. Check Algorithm 2 and 3
- Review of expectation, 2009. pdf
- Deep Q-Learning. Keon
- Policy method for Cartpole, Kvfrans.
repo - Fundamentals of Policy Gradients, Seita, 2017-03.
blog - Deep Deterministic Policy Gradients in TensorFlow, Emami, 2016-08.
blog
- Deep Reinforcement Learning: Pong from Pixels. Karpathy.
blog - Beginner's guide.
- Riemannian manifolds lecture, slides
- Information Geometry lecture, slides
- Adversarial Attacks, Robustness, theory and practice
- Deep RL for checkers
- Variational Inference tutorial
- Blog on theory and code. Covers q-learning with frozen lake, deep q-learning on doom/space invaders, policy gradients on Doom, A2C/A3C with Sonic, PPO with Sonic. link
- Minimal clean examples. Iteration methods, policy gradient, Grid world, CartPole, Atari, etc.
repo - Many PyTorch tutorials, all levels, for image and text.
repo - OpenAi Universe starter code, A3C algo.
repo - Minimalist REINFORCE for discrete and continuous actions.
repo - RLCode. Minimal example of DQN, DDQN, PG, A2C, A3C
- ikostrikov: a2c, ppo, acktr
- Beam search implementation in PyTorch
- Deep RL Bootcamp,
site
- The Stanford Natural Language Inference (SNLI) Corpus. 570k human-written English sentences. Text entailment
site
- Gibson Environments: Real-World Perception for Embodied Agents. A virtual environment for agents which is quite realistic.
- VizDoom. Doom environment using only visual information. Visuals include: FPV game pixels, object labelling visual, depth map, 2D map. Should probably use with a gym wrapper, like this one. To understand how to setup the engine, checkout this minimalist example. Also, checkout this pytorch example.
- MAME tookit, wrapper around the popular MAME arcade emulator
- MiniWorl, 2D and 3D environments, minimial dependencies, gym friendly
- Unreasonable effectiveness of one neuron,
blog
- PyTorch tutorial
- Stat Trek, stats and prob course