@ibaibabaibai: “model = approximate how a phenomenon works” http://heartruptcy. blog.fc2.com/blog-entry-150.html 
 is an idea similar to a sense of physics. I guess it is the “Model = Portrait” theory. Akaike’s view of models is more like “Model = Glasses”. Statistical models are positioned somewhere between these two extremes.

  • The intention of “portrait” and “glasses” in Dr. Iba’s statement, “I am a portrait artist,” was discussed within the company, so Nishio explained the sentence.

I thought it was about philosophy epistemology.

  • The “model = portrait” theory
    • Assumption that “human reason can recognize phenomena.”
    • The idea that a model is “an approximate representation of the phenomenon.
    • The basic premise of a portrait is that the true face is recognizable and an approximation of it is created.
    • Philosophy: man is able to recognize truth through the use of reason (corroborative explanation)
  • The “model = glasses” theory
    • Assumption that “human reason is incapable of recognizing phenomena.”
    • The idea that a model is “a tool to better recognize a phenomenon.
    • Glasses exist as a tool to help us see faces better without them, as we cannot recognize faces without them
      • I didn’t say, “Put on your glasses and you’ll see their true faces.”
    • Philosophy: man cannot recognize truth and can only make imperfect copies (copy theory)

Dr. Akaike: “There is no one and only true structure. We approach the ultimate truth or true structure, which is always in an unknown state, through the search for a better model. The truth we pursue is only relative in the sense that it depends on current knowledge, such that it can be represented by a model that gives one approximation of the subject.”

In the latter half of the 20th century, as real-world problems became increasingly complex and diverse, the traditional framework of statistical inference, which assumes the existence of a “true model,” gradually became inappropriate for reality. In 1973, Akaike proposed the Akaike Information Criterion (AIC), a criterion for obtaining a model that gives the best predictive value for a given situation of predicting future data, leaving a great mark in the history of statistics.

In the history of philosophy, Dr. Akaike may be positioned as advancing statistics from the standpoint of the copycat theory that humans cannot know the “True Model”.

Related: All models are wrong
.

2017-06-09 2022-05-20 Slightly improved text.

relevance - Graphs are tools - Tools to see the truth and tools to get the message across.


This page is auto-translated from /nishio/ăƒąăƒ‡ăƒ«ăšçœŒéĄăšäŒŒéĄ”ç”” 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.