CS 188 notes III

CS 188 notes III

crawler.py

Markov Decision Processes MDP

“Markov” generally means that given the present state, the future and the past are independent

know world model

-interactions with the world

(Deep) Reinforcement Learning

Trust Region Policy Optimization https://arxiv.org/abs/1502.05477

Nuclear fusion plasma control

Magnetic control of tokamak plasmas through deep reinforcement learning https://pubmed.ncbi.nlm.nih.gov/35173339/

Training language models to follow instructions with human feedback https://arxiv.org/abs/2203.02155

Time-Limited Values similar to HTML sitemap.xml

Dangerous Optimism: Assuming chance when the world is adversarial

Dangerous Pessimism: Assuming the worst case when it’s not likely

...

Human Machine Reward System: Heaven, Reincarnation, Nothingness

work harder if Nothingness?

...

Monty Hall Problem

Reinforcement Learning Problems

maximize expected rewards

DayDreamer: World Models for Physical Robot Learning https://autolab.berkeley.edu/assets/publications/media/2022-12-DayDreamer-CoRL.pdf

Temporal difference learning

Active Reinforcement Learning

Regret is a measure of your total mistake cost

CS 188 notes https://blogbarley.blogspot.com/2025/09/cs-188-notes.html

CS 188 notes II https://blogbarley.blogspot.com/2025/10/cs-188-notes-ii.html