- Concepts used in [reinforcement learning](/en/reinforcement%20learning).
- exploitation-exploration trade-offs
- There are variations on whether to say "use" or "exploration" first and how to translate "exploitation.
        - [Trade-off between exploration and exploitation](/en/Trade-off%20between%20exploration%20and%20exploitation) (I had a notation quirk myself.)
    - Trade-off between use and exploration 62200 98
    - Trade-off between search and use 50900 294
    - The Dilemma of Search and Knowledge Use

• 

  • https://www.slideshare.net/nishio/1-70974083#14

• If one chooses an option that seems useful, the opportunity to discover that other options are more useful is lost.

  • Selecting the optimal solution based on past experience prevents new searches and gets stuck in [local minimum - Ypsilon Greedy : Introducing Randomness for the Search

• On the other hand, if they repeatedly search for more useful options, they will not benefit from the useful options they have found.

  • To find another minimum from local minimum, you have to go up once.
  • see You have to climb up to get down.
  • leader of a monkey's mountain (e.g. a gazelle) 。 parochial 。

• see In times of uncertainty, be optimistic.

• Significant development in the field of reinforcement learning, but first published much older. - George Edward Pelham Box

  • Box, G. E., 1954. The exploration and exploitation of response surfaces: some general considerations and examples. Biometrics, 10(1), pp.16-60.

• Also used in the area of organizational learning

  • March, J.G., 1991. Exploration and exploitation in organizational learning. Organization science, 2(1), pp.71-87.

• relevance - filter bubble [filter bubble - Wikipedia https://ja.wikipedia.org/wiki/%E3%83%95%E3%82%A3%E3%83%AB%E3%82%BF%E3%83%BC%E3%83%90%E3%83%96%E3% 83%AB] - By getting information through a filter that matches your ideology, you get stuck in the local minimum of the ideology. - selection and concentration - Often used as an excuse to reduce search costs

#No pictured blind spot card yet 1031

BELOW_IS_AI_GENERATED

利用と探索のトレードオフ

2023-09-05 01:15

Summary of notes

. This section describes the "trade-off between search and use" in reinforcement learning. By choosing a useful alternative, one may miss other possibilities, and by searching for new alternatives, one may not benefit from known useful alternatives. This concept has been applied in other areas such as organizational learning and information filtering.

Relation to Fragment

. The fragment "Revised Differences for the Fourth Printing" describes the trade-off between exploration and exploitation and is directly related to the note. Specifically, it states, "If you only choose the option that you think is best based on past experience, you will never find a better option. That is not enough exploration." This part of the note is consistent with the main theme of the note.

deep thinking

The trade-off between exploration and exploitation represents a balance between pursuing new possibilities and maximizing the benefits from known useful alternatives. This is an important consideration when choosing how to obtain information and learn.

summary of thoughts and title

. The trade-off between exploration and exploitation represents a balance between new possibilities and known benefits."

extra info

titles: ["Revision differences for the fourth printing", "Proofreading for the English version (chapters 2 and 3)", "Utilization-exploitation tradeoff", "(2.2.3.1) Exploration-exploitation tradeoff", "Disabling 🌀nominalization", "Flow and utilization and exploration", " Intellectual production techniques for engineers All hierarchical table of contents", "(6.2.2.2) Advantages and disadvantages of framework"] generated: 2023-09-05 01:15

This page is auto-translated from /nishio/利用と探索のトレードオフ using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I'm very happy to spread my thought to non-Japanese readers. Trade-offs between use and exploration see Trade-offs between use and exploration

This page is auto-translated from /nishio/利用と探索のトレード・オフ using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I'm very happy to spread my thought to non-Japanese readers.