Continuing Education

Some projects I'm using to learn AI since my MS

TODO

* Means priority

• Reinforcement Learning
  • Value Based Methods - I'm pretty much up to date with these methods, but might as well implement them. I may go into less explanation though.
    • Monte-Carlo Methods
    • $TD(\lambda)$
    • Deep Q Learning *
    • Prioritized Experience Replay
    • Double Q Learning
    • Dueling Q Learning
    • Multi Step Learning
    • Distributional DQN
    • Noisy Nets
    • RAINBOW
  • Policy Based Methods
    • REINFORCE *
    • Actor-Critic *
    • Proximal Policy Optimization (PPO) *
  • Model Based Reinforcement Learning
    • AlphaZero
  • Exploration in RL
  • Multi Agent RL
    • Emergent Communication through Negotiation
• Transformers
  • Tokenization *
  • Word Embeddings *
  • Transformers *
  • Fine Tuning *
  • RLHF *
  • Self Implemented LLM
  • Multimodality
• Diffusion Models
• Graph Neural Networks (GNN)
• Techniques
  • Distribution
    • pytorch.distributed
    • rllib
  • Profiling
  • Debugging Metrics

Other interests slightly related

• linguistic philosophy
  • Wittgenstein
  • Chompsky
• cognitive science
  • Predictive Coding Models
• economics and game theory
  • Von Neuman

Sources for further work

• Key Papers in Deep RL

Installation

First follow specific instructions for your operating system listed in subheaders below.

Then make sure you have cloned the submodules:

git submodule update --init --recursive

Then install the dependencies:

poetry install

After you run poetry install, pip install pytorch the way it describes here for your system. This ensures you get the best performance. Always use the version listed in .github/workflows/ci.yaml for the best compatibility.

MacOSX

brew install sdl sdl_ttf sdl_image sdl_mixer portmidi

Ubuntu

sudo apt-get install libsdl2-dev libsdl2-image-dev libsdl2-mixer-dev libsdl2-ttf-dev libfreetype6-dev libportmidi-dev libjpeg-dev python3-setuptools python3-dev python3-numpy