手动更新
This commit is contained in:
28
wiki/concepts/Fixed-Point-Semantics.md
Normal file
28
wiki/concepts/Fixed-Point-Semantics.md
Normal file
@@ -0,0 +1,28 @@
|
||||
---
|
||||
title: "Fixed-Point Semantics"
|
||||
type: concept
|
||||
tags: []
|
||||
---
|
||||
|
||||
## Definition
|
||||
不动点语义是递归自我优化系统的收敛理论基础。稳定生成能力被定义为自映射 $\Phi: \mathcal{G} \to \mathcal{G}$ 的不动点 $G^* \in \mathcal{G}$,满足:
|
||||
$$\Phi(G^*) = G^*$$
|
||||
|
||||
即:在经历一次完整的"生成-优化-更新"循环后,生成器保持不变——它已经与自身的能力上限达成一致。
|
||||
|
||||
## Existence & Convergence
|
||||
不动点的存在性由 Banach 不动点定理保证:当 $\Phi$ 是收缩映射(contraction mapping)时,不动点存在且唯一,并且可以通过迭代获得:
|
||||
$$G^* = \lim_{n \to \infty} \Phi^n(G_0)$$
|
||||
|
||||
这意味着:**即使初始生成器 $G_0$ 很简单,通过足够多的迭代,生成器序列将收敛到稳定状态**。
|
||||
|
||||
## Significance
|
||||
不动点代表了一个生成器,其输出已经包含改进自身所需的全部信息——它不再需要外部优化器的干预。同时,不动点的存在证明了系统不会陷入无限循环或发散。
|
||||
|
||||
## Sources
|
||||
- [[a-formalization-of-recursive-self-optimizing-generative-systems]]
|
||||
|
||||
## Connections
|
||||
- [[Recursive Self-Optimization]] ← converges_to ← [[Fixed-Point Semantics]]
|
||||
- [[Self-Referential Computation]] ← is_grounded_in ← [[Fixed-Point Semantics]]
|
||||
- [[Y-Combinator]] ← computes ← [[Fixed-Point Semantics]]
|
||||
Reference in New Issue
Block a user