Helpful Resources

Resources and links for acitivities.

---

Sports

Surfing

• MagicSeaWeed - the best 7 day wave forecast across the world.

Skiing

• OnTheSnow - snow forecast and history (Winter Park)
• OpenSnow - 10 day snow forecast (Colorado Rockies)
• Evo - discounted gear, not just skiing

Hiking

• AllTrails

Yoga

• Yoga with Adriene - awesome guided yoga lessons with insightful perceptions of movement.

Mountain Biking

• TrailForks - routes with their difficulty for mountain biking.

---

Courses:

Game theory/(Multi-Agent) Reinforcement Learning (MARL)

• (Foundations):
  • Game Theory (Standord, UBC), part 1 - lacks intuition, but a good structure. Introduces Nash’s equilibrium, strategies, convergence intuition for strategies, (non)zero-sum games, (im)perfect information games.
  • Game Theory (Standord, UBC), part 2 - extends to marketplaces, voting. Feel free to skip, but look up Walras Equilibrium Theorem, because it’s cool.
    • (If you speak Russian) Game theory, MIPT - combines the structure of 2 courses above, adds more examples for intuition, overall explained better.
  • To make this theory practical, understand alpha-beta pruning, min-max in detail. Simple AI agents use this exact framework, but use heuristics to evaluate non-final states. Check out checkers-ai, chess-ai.
• Moving into “deep” Game Theory:
  • Fictitious play convergence towards Nash Equilibrium is one of the most helpful concepts from pure game theory. I see GANs as a fictitious play convergence framework. Explicitely, there is fictitious GAN.
  • Neural Fictitious Self-Play is one of the most effective and practical algorithms used in neural agents. Used by Deepmind’s AlphaZero to achieve superhuman performance from scratch.
  • Overview of RL by Ilya: Ilya Sutskever: OpenAI Meta-Learning and Self-Play
  • Advanced RL by Deepmind - further material
• Multi-Agent RL:
  • Deepmind’s tutorial on MARL - a great summary of everything above.
  • AlphaStar - Deepmind’s Starcraft 2 paper. They used human games to pretrain agents via imatation learning, but later used self-play to advance. Notably, human games were necessary here, in comparision to AlphaZero.

Network Analysis and Modeling

The most informative and helpful study of networks-graphs I’ve ever taken/seen. Aaron Clauset is a world expert on the topic and this is his view on the field.

Machine Learning - math perspective

Here, you will derive from scratch the standard ML toolset (Regressions, Support Vector Machines, Regularization, Boosting etc.). If you want to understand the ins and outs of ML, take this class.

Algorithmic Economics and Machine Learning

Information elicitation, game theoretic lens on Economics. Really elegant.

Learning How To Learn

Neuroscience backed facts about how your brain acquires and solidifies knowledge. A class everyone should take as early as possible in any career they choose. Applicable to sports, science, self-improvement and most things really.