[KongchangAI]
ConceptUniversal Approximation Theorem / UAT

通用近似定理

1989年由George Cybenko证明的数学定理，指出只要层数和神经元足够多，神经网络可以以任意精度逼近任何连续函数，是深度学习理论基础之一。

Core Facts

The universal approximation theorem was rigorously proven by mathematician Cybenko in 1989 for the Sigmoid activation function.

Verified70% confidence4 sources

Timeline (last 90 days)

Jul 23

The universal approximation theorem was rigorously proven by mathematician Cybenko in 1989 for the Sigmoid activation function.

Verified70%

Source Articles

From Functions to Neural Networks: The Core Logic Behind AI, Explained SimplyJul 23What Is a Large Language Model, Really? The Core Principles Explained Through the Lens of FunctionsJul 21