concept
["CNN has spatial concepts a priori, but it's able to learn concepts a posteriori/experientially without any categories at all, right?"] 「 concept を経験的に学習」できているのだろうか??/villagepump/nishio.icon It depends on the definition of the word "concept"...
Even if a neural net can determine that cat image A and cat image B are similar, whether the neural net has the concept of "cat"...
We need to see a definition of the word "concept."
Translation of Latin conceptio, English and French concept, German Begriff, etc. It is the fundamental form and product of man's grasp of objects, and is sometimes used synonymously with conception, image, etc.
Simply put, concepts are nominals.
Naming by language is nothing but the production of concepts.
The concept of "individual" or "single concept" and the concept of "single" or "single concept" are the same as the concept of "individual" or "single concept.
General, universal, or class concepts.
The concept of [ connotation ], which is the semantic content, and the scope of application [ extension ], which is the scope of the concept. Furthermore, the concept of "superordinate" and "subordinate," and
The distinction between "like concepts" and "species concepts" and so on is also made.
It depends on the characteristics of the concept as a language.
Origin of the concept and its role in cognition
There have been various arguments and positions on such matters since ancient times.
>→Related Items Extension| category |Determination 言語による命名が概念の産出にほかならないだから/villagepump/nishio.icon
CNN doesn't produce concepts because it doesn't name them.
AI can acquire concepts named by humans by learning a set of language and images in a CLIP-like fashion.
---
This page is auto-translated from /nishio/概念 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.