A brief summary for :
Computational rationality: A converging paradigm for intelligence in brains, minds, and machines
Computational rationality based on three core ideas:
- We aim to plan for actions for maximizing expected utility.
- For real-world problems, we can only use effectively approximated rational algorithms.
- Algortihms can be adpted for specific needs.(offline, online).
Computational tradeoffs in sequential decision-making:
- Researchers have shown that the human brain might use an algorithm like MCTS to sovle spatial navigation problmes. 【interesting！！！】
- Hybird model-free and model-based decision-making systems may be a promising route to explain how human decision-making in complex sequential taks can be so accurate and fast.
Rational decisions under bounded computational resources
- The complexity of probabilistic inference in Bayesian networks has been shown to be in the nondeterministic polynomial-time (NP)–hard complexity class.
- A tapestry of approximate inferential methods have been developed, include Monte Carlo simulation, modulating the complexity of models(changing the size or level of abstraction)
