← Archive
lm-005070 · 2026-11

空間資訊佔據 V:跨解析度 Generator 系統與連續極限

下載 MD 檔 ⬇
📎 附件 · Companion files — 隨文交付的程式 / 證明 / 資料,可獨立下載重驗

空間資訊佔據 V:跨解析度 Generator 系統與連續極限

Spatial Information Occupancy V: Cross-Resolution Generator Systems and Continuum Limits

系列: Spatial Information Occupancy and Generator Geometry
篇次: Paper 05 / Resolution and Continuum Limit
版本: v1.0
日期: 2026-10-02
作者: Neo.K
AI 協力研究者: Aletheia / ChatGPT, GPT-5.6 Sol
機構: EveMissLab/一言諾科技有限公司
狀態: Formal Research Draft / Exact Refinement and Limit Theory

直接前置:

  • Paper 01《空間資訊佔據 I:超越幾何測度的容器相對自由結構》;
  • Paper 02《空間資訊佔據 II:自由度傳輸、邊界飽和與結構逃逸》;
  • Paper 03《空間資訊佔據 III:協同構成元代數、潛在自由與纖維資訊場》;
  • Paper 04《空間資訊佔據 IV:Generator Quotient 與跨表示資訊不變量》;
  • E6 Cross-Resolution Generator Limit;
  • 主客邊三域代數(SOBTA)。

摘要

Paper 04 的 Generator Quotient:

Qq(X)=[Hq(X)]≅\mathfrak Q_q(X) = [\mathcal H_q(X)]_{\cong}

可以 quotient 掉 vertex labels、radix、symbol encoding 與其他 lossless representation redundancy,但它仍假設被比較的 finite hypergraphs 具有相同 vertex cardinality。

解析度改變時:

N→2N,N \rightarrow 2N,

constituent 數、minimal-generator 數與 generator cardinality 都會改變,普通 hypergraph isomorphism 因而不再適用。

本文建立第一個 cross-resolution exact model。

令連續 substrate 為:

[0,1],[0,1],

固定 physical generator thickness:

0<α≤12,\boxed{ 0<\alpha\le\frac12, }

並取 admissible resolutions:

Nα={N∈N:hN=αN∈N}.\mathcal N_\alpha = \{ N\in\mathbb N: h_N=\alpha N\in\mathbb N \}.

在 resolution NN 下:

VN={0,…,N−1},V_N = \{0,\ldots,N-1\},

minimal generators 定義為所有長度:

hNh_N

的 contiguous blocks:

GN,j={j,j+1,…,j+hN−1},\boxed{ G_{N,j} = \{j,j+1,\ldots,j+h_N-1\}, }

其中:

0≤j≤N−hN.0\le j\le N-h_N.

將 cell ii 嵌入 physical interval:

[iN,i+1N],\left[ \frac{i}{N}, \frac{i+1}{N} \right],

則每個 discrete generator 的 physical realization 精確為:

[jN,jN+α].\boxed{ \left[ \frac{j}{N}, \frac{j}{N}+\alpha \right]. }

所以 generator physical size 不隨解析度改變;改變的是可取 start positions 的密度。

本文證明 generator family:

GN={[jN,jN+α]:0≤j≤N−hN}\mathscr G_N = \left\{ \left[ \frac jN, \frac jN+\alpha \right] : 0\le j\le N-h_N \right\}

在 interval Hausdorff metric 所誘導的 hyperspace metric 下,以精確速率:

dH(GN,Gα∞)=12N\boxed{ d_H( \mathscr G_N, \mathscr G_\alpha^\infty ) = \frac{1}{2N} }

收斂到 continuum generator family:

Gα∞={[s,s+α]:0≤s≤1−α}.\boxed{ \mathscr G_\alpha^\infty = \{ [s,s+\alpha]: 0\le s\le1-\alpha \}. }

本文亦證明 normalized generator-overlap kernel 在所有 resolution 上精確取樣同一函數:

Kα(δ)=max⁡(0,α−δ).\boxed{ K_\alpha(\delta) = \max(0,\alpha-\delta). }

對 dyadic refinement:

2N→N,2N\rightarrow N,

定義 cell map:

c(i)=⌊i2⌋.c(i) = \left\lfloor\frac i2\right\rfloor.

本文證明:

  • even-start fine generator 精確映到 coarse minimal generator;
  • odd-start fine generator 映到一個長度 hN+1h_N+1 的 contiguous winning super-set;
  • 每個 coarse minimal generator 都有 aligned fine minimal preimage。

因此:

cc

不是 minimal-hyperedge isomorphism,而是一個 winning-preserving refinement morphism。

更一般地,對任意 fine released state:

A⊆V2N,A\subseteq V_{2N},

定義 inner coarse-graining:

C−(A)C^-(A)

只在兩個 fine children 都 released 時才釋放 coarse cell;定義 outer coarse-graining:

C+(A)C^+(A)

只要任一 child released 就釋放 coarse cell。本文解析證明:

VN(C−A)≤V2N(A)≤VN(C+A).\boxed{ V_N(C^-A) \le V_{2N}(A) \le V_N(C^+A). }

所以 lossy coarse projection 不需要被錯當成 fine ontological identity;它可以保留合法上下界。

本文最後證明 Paper 03 的 Latent Freedom Potential 與 constituent capacity field 都存在 normalized continuum limit。對一段 physical released interval II,有:

∣dN(AN(I))N−max⁡(0,α−∣I∣)∣≤1N.\boxed{ \left| \frac{d_N(A_N(I))}{N} - \max(0,\alpha-|I|) \right| \le \frac1N. }

對 capacity Shapley field:

ΦN(xi),xi=i+1/2N,\Phi_N(x_i), \qquad x_i=\frac{i+1/2}{N},

則:

Φα(x)=min⁡(1,xα,1−xα)\boxed{ \Phi_\alpha(x) = \min \left( 1, \frac{x}{\alpha}, \frac{1-x}{\alpha} \right) }

且:

∥ΦN−Φα∥∞≤12αN.\boxed{ \|\Phi_N-\Phi_\alpha\|_\infty \le \frac{1}{2\alpha N}. }

因此在此 interval-generator family 中,不同 constituent counts 可以共同逼近同一個 continuum generator object。


Abstract

Paper 04 introduced the Generator Quotient for finite freedom-generator hypergraphs with fixed vertex cardinality. Resolution changes are harder: under N↦2NN\mapsto2N, the number of constituents, the number of generators, and generator cardinalities all change, so ordinary hypergraph isomorphism no longer applies.

This paper develops an exact cross-resolution model on the continuous substrate [0,1][0,1]. Fix a physical generator thickness 0<α≤1/20<\alpha\le1/2 and resolutions NN such that hN=αNh_N=\alpha N is integral. The resolution- NN generators are all contiguous hNh_N -cell blocks.

Each discrete generator embeds exactly as a physical interval of length α\alpha. The resulting discrete generator family converges to the continuum interval family

Gα∞={[s,s+α]:0≤s≤1−α}\mathscr G_\alpha^\infty = \{ [s,s+\alpha]: 0\le s\le1-\alpha \}

with exact Hausdorff rate

dH(GN,Gα∞)=12N.d_H( \mathscr G_N, \mathscr G_\alpha^\infty ) = \frac{1}{2N}.

The normalized overlap of two generators separated by physical displacement δ\delta exactly samples the resolution-independent kernel

Kα(δ)=max⁡(0,α−δ).K_\alpha(\delta) = \max(0,\alpha-\delta).

Under dyadic coarse-graining, even-start fine generators map exactly to coarse minimal generators, while odd-start fine generators map to winning coarse supersets. More generally, for every fine released state AA,

VN(C−A)≤V2N(A)≤VN(C+A),V_N(C^-A) \le V_{2N}(A) \le V_N(C^+A),

where C−C^- and C+C^+ are inner and outer coarse-graining operators. This inequality is proved analytically and audited exhaustively for all 2162^{16} states of the 16→816\to8 case.

The normalized latent-freedom potential converges to the continuous deficit max⁡(0,α−ℓ)\max(0,\alpha-\ell) with error at most 1/N1/N, and the constituent capacity field converges uniformly to a continuum ramp-plateau-ramp field with error at most 1/(2αN)1/(2\alpha N).

The paper therefore replaces same-cardinality quotienting with a refinement-system view: constituents may change with resolution while physical generator relations converge to a common continuum object.


1. 從 Representation Invariance 到 Resolution Invariance

Paper 04 處理:

∣W1∣=∣W2∣|W_1| = |W_2|

時的 representation redundancy。

若:

H1≅H2,\mathcal H_1 \cong \mathcal H_2,

則:

Qq(H1)=Qq(H2).\mathfrak Q_q(\mathcal H_1) = \mathfrak Q_q(\mathcal H_2).

但解析度改變:

N→2NN\rightarrow2N

時:

∣VN∣=N,|V_N|=N, ∣V2N∣=2N.|V_{2N}|=2N.

所以:

HN≇H2N\mathcal H_N \not\cong \mathcal H_{2N}

甚至在它們明顯描述同一 physical freedom relation 時也必然如此。

因此 cross-resolution 問題不能只靠 finite isomorphism。

需要新的結構:

refinement+physical embedding+limit.\boxed{ \text{refinement} + \text{physical embedding} + \text{limit}. }

2. Continuous Target

固定:

0<α≤12.\boxed{ 0<\alpha\le\frac12. }

定義 continuum substrate:

X∞=[0,1].\boxed{ X^\infty=[0,1]. }

continuum generator family:

Gα∞={[s,s+α]:0≤s≤1−α}.\boxed{ \mathscr G_\alpha^\infty = \{ [s,s+\alpha]: 0\le s\le1-\alpha \}. }

本文稱:

Hα∞=([0,1],Gα∞)\boxed{ \mathcal H_\alpha^\infty = ( [0,1], \mathscr G_\alpha^\infty ) }

為 continuum generator system。

這不是 finite hypergraph。

它只是一個被 finite resolution systems 逼近的 continuous relational object。


3. Discrete Resolution System

定義 admissible resolution set:

Nα={N∈N:αN∈N}.\boxed{ \mathcal N_\alpha = \{ N\in\mathbb N: \alpha N\in\mathbb N \}. }

對:

N∈Nα,N\in\mathcal N_\alpha,

令:

hN=αN.\boxed{ h_N = \alpha N. }

constituent set:

VN={0,…,N−1}.\boxed{ V_N = \{0,\ldots,N-1\}. }

minimal generators:

GN,j={j,j+1,…,j+hN−1},\boxed{ G_{N,j} = \{j,j+1,\ldots,j+h_N-1\}, }

其中:

0≤j≤N−hN.0\le j\le N-h_N.

因此:

GN={GN,j:0≤j≤N−hN}.\boxed{ \mathcal G_N = \{ G_{N,j} : 0\le j\le N-h_N \}. }

binary freedom:

VN(A)=1[∃G∈GN:G⊆A].\boxed{ V_N(A) = \mathbf 1[ \exists G\in\mathcal G_N: G\subseteq A ]. }

4. Normalized Generator Order

Proposition 4.1

對所有:

N∈Nα,N\in\mathcal N_\alpha,

有:

∣GN,j∣∣VN∣=hNN=α.\boxed{ \frac{|G_{N,j}|}{|V_N|} = \frac{h_N}{N} = \alpha. }

所以 generator cardinality 雖隨 NN 改變,但 normalized order 精確固定。

Proof

由:

hN=αN.h_N=\alpha N.

直接得到。證畢。


5. Physical Embedding

把 constituent ii 嵌入:

ιN(i)=[iN,i+1N].\boxed{ \iota_N(i) = \left[ \frac{i}{N}, \frac{i+1}{N} \right]. }

對 subset:

A⊆VN,A\subseteq V_N,

令:

ιN(A)=⋃i∈AιN(i).\iota_N(A) = \bigcup_{i\in A} \iota_N(i).

Theorem 5.1 — Exact Physical Generator Realization

對任意 minimal generator:

GN,j,G_{N,j},

有:

ιN(GN,j)=[jN,jN+α].\boxed{ \iota_N(G_{N,j}) = \left[ \frac{j}{N}, \frac{j}{N}+\alpha \right]. }

特別地:

μ(ιN(GN,j))=α.\boxed{ \mu( \iota_N(G_{N,j}) ) = \alpha. }

Proof

generator 由連續 cells:

j,j+1,…,j+hN−1j,j+1,\ldots,j+h_N-1

組成。

其 physical union:

[jN,j+hNN].\left[ \frac jN, \frac{j+h_N}{N} \right].

因:

hN/N=α,h_N/N=\alpha,

所以:

ιN(GN,j)=[jN,jN+α].\iota_N(G_{N,j}) = \left[ \frac jN, \frac jN+\alpha \right].

證畢。


6. Generator Start Set

定義:

SN={jN:0≤j≤N−hN}.\boxed{ S_N = \left\{ \frac jN: 0\le j\le N-h_N \right\}. }

因:

hN/N=α,h_N/N=\alpha,

最大 start:

N−hNN=1−α.\frac{N-h_N}{N} = 1-\alpha.

所以:

SN⊆[0,1−α].S_N \subseteq [0,1-\alpha].

7. Hausdorff Convergence

對 compact subsets:

A,B⊆R,A,B\subseteq\mathbb R,

使用標準 Hausdorff distance:

dH(A,B).d_H(A,B).

Theorem 7.1 — Exact Start-Set Hausdorff Distance

若:

0<α<10<\alpha<1

且:

N−hN≥1,N-h_N\ge1,

則:

dH(SN,[0,1−α])=12N.\boxed{ d_H( S_N, [0,1-\alpha] ) = \frac{1}{2N}. }

Proof

因:

SN⊆[0,1−α],S_N \subseteq [0,1-\alpha],

故:

sup⁡s∈SNd(s,[0,1−α])=0.\sup_{s\in S_N} d( s, [0,1-\alpha] ) = 0.

相鄰 start points 距離:

1N.\frac1N.

且兩端:

0,1−α0, \qquad 1-\alpha

都屬於 SNS_N。

所以連續 interval 中距最近 start point 最遠的位置是任一相鄰 grid starts 的 midpoint,其距離:

12N.\frac{1}{2N}.

因此 Hausdorff distance 即:

12N.\frac{1}{2N}.

證畢。


8. Generator-Family Hausdorff Limit

對固定長度 α\alpha 的 closed intervals:

Is=[s,s+α],I_s=[s,s+\alpha], It=[t,t+α],I_t=[t,t+\alpha],

有:

dH(Is,It)=∣s−t∣.\boxed{ d_H(I_s,I_t) = |s-t|. }

因此在 generator hyperspace 中,離散 family:

GN={[jN,jN+α]:0≤j≤N−hN}\mathscr G_N = \left\{ \left[ \frac jN, \frac jN+\alpha \right] : 0\le j\le N-h_N \right\}

與 continuum family:

Gα∞\mathscr G_\alpha^\infty

之間的 Hausdorff distance等於其 start sets 的 Hausdorff distance。

Corollary 8.1 — Exact Generator-Family Convergence Rate

dH(GN,Gα∞)=12N.\boxed{ d_H( \mathscr G_N, \mathscr G_\alpha^\infty ) = \frac{1}{2N}. }

所以:

GN→N→∞HGα∞.\boxed{ \mathscr G_N \xrightarrow[N\to\infty]{H} \mathscr G_\alpha^\infty. }

9. Cross-Resolution Overlap Kernel

對兩個 discrete generators:

GN,j,GN,k,G_{N,j}, \qquad G_{N,k},

令:

r=∣j−k∣.r=|j-k|.

Theorem 9.1 — Exact Discrete Overlap

∣GN,j∩GN,k∣=max⁡(0,hN−r).\boxed{ |G_{N,j}\cap G_{N,k}| = \max(0,h_N-r). }

Proof

兩個長度皆為:

hNh_N

的 contiguous integer blocks,相隔:

rr

個 start indices。

若:

r≥hN,r\ge h_N,

不重疊。

若:

r<hN,r<h_N,

重疊長度:

hN−r.h_N-r.

證畢。


令 normalized displacement:

δ=rN.\boxed{ \delta=\frac rN. }

定義 overlap normalized by substrate size:

KN(δ)=∣GN,j∩GN,k∣N.\boxed{ K_N(\delta) = \frac{ |G_{N,j}\cap G_{N,k}| }{N}. }

Corollary 9.2 — Resolution-Independent Overlap Kernel

對所有 grid-representable:

δ=r/N,\delta=r/N,

有:

KN(δ)=Kα(δ)=max⁡(0,α−δ).\boxed{ K_N(\delta) = K_\alpha(\delta) = \max(0,\alpha-\delta). }

所以不同 resolutions 不是只「趨近」同一 kernel;它們在共同可表示 displacement 上精確取樣同一函數。


10. Dyadic Coarse Map

現在令:

N∈NαN\in\mathcal N_\alpha

且:

2N∈Nα.2N\in\mathcal N_\alpha.

則:

h2N=2hN.h_{2N} = 2h_N.

定義 fine-to-coarse cell map:

c2N→N(i)=⌊i2⌋.\boxed{ c_{2N\rightarrow N}(i) = \left\lfloor \frac i2 \right\rfloor. }

對 subset:

A⊆V2N,A\subseteq V_{2N},

定義其 direct coarse image:

c(A)={c(i):i∈A}.c(A) = \{ c(i): i\in A \}.

11. Fine Generator Image Theorem

Theorem 11.1 — Dyadic Generator Image

對 fine minimal generator:

G2N,j,G_{2N,j},

有:

Even start

若:

j=2k,j=2k,

則:

c(G2N,2k)=GN,k.\boxed{ c(G_{2N,2k}) = G_{N,k}. }

Odd start

若:

j=2k+1,j=2k+1,

則:

c(G2N,2k+1)={k,k+1,…,k+hN}.\boxed{ c(G_{2N,2k+1}) = \{k,k+1,\ldots,k+h_N\}. }

因此其 cardinality:

hN+1,h_N+1,

並包含兩個相鄰 coarse minimal generators:

GN,k,G_{N,k},

以及:

GN,k+1,G_{N,k+1},

只要二者都位於合法 start range。

Proof

fine generator 長度:

h2N=2hN.h_{2N}=2h_N.

若:

j=2k,j=2k,

fine indices:

2k,…,2k+2hN−12k,\ldots,2k+2h_N-1

在 floor-divide-by-two 後恰成:

k,…,k+hN−1.k,\ldots,k+h_N-1.

若:

j=2k+1,j=2k+1,

fine indices:

2k+1,…,2k+2hN2k+1,\ldots,2k+2h_N

映成:

k,…,k+hN.k,\ldots,k+h_N.

證畢。


Corollary 11.2 — Winning Preservation

對所有 fine minimal generators:

G∈G2N,G\in\mathcal G_{2N},

其 coarse image:

c(G)c(G)

滿足:

VN(c(G))=1.\boxed{ V_N(c(G))=1. }

所以:

cc

是 minimal-generator 到 coarse winning-state 的 preserving map,但不是一般的 minimal-to-minimal homomorphism。


Corollary 11.3 — Coarse Generator Preimage

對每個 coarse minimal generator:

GN,k,G_{N,k},

aligned fine generator:

G2N,2kG_{2N,2k}

滿足:

c(G2N,2k)=GN,k.\boxed{ c(G_{2N,2k}) = G_{N,k}. }

因此 coarse minimal generators 都有 exact fine representatives。


12. Inner and Outer Coarse-Graining

對:

A⊆V2N,A\subseteq V_{2N},

定義 inner coarse-graining:

C−(A)={i∈VN:{2i,2i+1}⊆A}.\boxed{ C^-(A) = \left\{ i\in V_N: \{2i,2i+1\}\subseteq A \right\}. }

定義 outer coarse-graining:

C+(A)={i∈VN:{2i,2i+1}∩A≠∅}.\boxed{ C^+(A) = \left\{ i\in V_N: \{2i,2i+1\}\cap A\neq\varnothing \right\}. }

顯然:

C−(A)⊆C+(A).C^-(A) \subseteq C^+(A).

13. Coarse-Bracketing Theorem

Theorem 13.1 — Exact Inner/Outer Freedom Bracket

對所有:

A⊆V2N,A\subseteq V_{2N},

有:

VN(C−A)≤V2N(A)≤VN(C+A).\boxed{ V_N(C^-A) \le V_{2N}(A) \le V_N(C^+A). }

Proof

因三個量皆為 binary,只需證明兩個 implication。

Lower bound

若:

VN(C−A)=1,V_N(C^-A)=1,

則存在 coarse generator:

GN,k={k,…,k+hN−1}G_{N,k} = \{k,\ldots,k+h_N-1\}

使:

GN,k⊆C−A.G_{N,k} \subseteq C^-A.

由 C−C^- 定義,對每個:

i∈GN,k,i\in G_{N,k},

都有:

2i,2i+12i, \quad 2i+1

屬於 AA。

所以 AA 包含完整 fine block:

{2k,…,2k+2hN−1}=G2N,2k.\{2k,\ldots,2k+2h_N-1\} = G_{2N,2k}.

因此:

V2N(A)=1.V_{2N}(A)=1.

故:

VN(C−A)≤V2N(A).V_N(C^-A) \le V_{2N}(A).

Upper bound

若:

V2N(A)=1,V_{2N}(A)=1,

則 AA 包含某個 fine generator:

G2N,jG_{2N,j}

長度為:

2hN.2h_N.

其 coarse parents:

c(G2N,j)c(G_{2N,j})

由 Theorem 11.1 是:

  • 一個長度 hNh_N coarse block;或
  • 一個長度 hN+1h_N+1 coarse block。

因每個 parent coarse cell 至少有一個 child 屬於:

G2N,j⊆A,G_{2N,j}\subseteq A,

所以:

c(G2N,j)⊆C+(A).c(G_{2N,j}) \subseteq C^+(A).

因此:

C+(A)C^+(A)

至少包含一個長度 hNh_N coarse generator。

所以:

VN(C+A)=1.V_N(C^+A)=1.

故:

V2N(A)≤VN(C+A).V_{2N}(A) \le V_N(C^+A).

證畢。


14. Projection without Annexation

Theorem 13.1 的意義不是 coarse layer 等於 fine layer。

相反地:

C−(A)C^-(A)

與:

C+(A)C^+(A)

通常不同。

因此 fine state:

AA

不必存在唯一 coarse state可以完整代表。

但 coarse layer可以給出:

lower freedom certificate\boxed{ \text{lower freedom certificate} }

與:

upper freedom certificate.\boxed{ \text{upper freedom certificate}. }

因此:

coarse projection≠fine ontological identity,\boxed{ \text{coarse projection} \neq \text{fine ontological identity}, }

同時:

coarse projection can still preserve valid bounds.\boxed{ \text{coarse projection can still preserve valid bounds}. }

這是 SOBTA:

Projection without Annexation\text{Projection without Annexation}

在本模型中的精確數學實例。


15. Latent Freedom Potential Across Resolution

Paper 03 定義:

dN(A)=min⁡G∈GN∣G∖A∣.d_N(A) = \min_{G\in\mathcal G_N} |G\setminus A|.

先處理 contiguous released block。

Lemma 15.1 — Contiguous-Block Potential

若:

AA

是長度:

kk

的 contiguous released block,則:

dN(A)=max⁡(0,hN−k).\boxed{ d_N(A) = \max(0,h_N-k). }

Proof

若:

k≥hN,k\ge h_N,

則 AA 內包含某個 length- hNh_N generator,所以:

dN(A)=0.d_N(A)=0.

若:

k<hN,k<h_N,

任何 generator 有 hNh_N 個 cells,而 AA 至多能提供其中 kk 個,因此:

dN(A)≥hN−k.d_N(A)\ge h_N-k.

另一方面,因 AA 本身為 contiguous block 且位於:

VN,V_N,

可將它向左或右延伸成某個 length- hNh_N legal generator,所以存在 generator:

GG

使:

A⊆G,A\subseteq G,

且:

∣G∖A∣=hN−k.|G\setminus A| = h_N-k.

因此 equality 成立。證畢。


16. Center-Sampled Physical Interval

令 physical released interval:

I=[s,s+ℓ]⊆[0,1].I=[s,s+\ell] \subseteq [0,1].

定義 resolution- NN sampled set:

AN(I)={i:i+1/2N∈I}.\boxed{ A_N(I) = \left\{ i: \frac{i+1/2}{N} \in I \right\}. }

因 cell centers 等距排列, AN(I)A_N(I) 是 contiguous block。

令:

kN=∣AN(I)∣.k_N = |A_N(I)|.

Lemma 16.1 — Sampling Count Error

∣kNN−ℓ∣≤1N.\boxed{ \left| \frac{k_N}{N} - \ell \right| \le \frac1N. }

Proof

kNk_N 等於 spacing 1/N1/N 的 arithmetic progression 落在長度 ℓ\ell interval 中的點數。

將座標乘以 NN,等價於計數 unit-spaced points 落在長度:

NℓN\ell

的 interval 中。

任意長度 LL 的 closed interval 所含 unit-spaced lattice points 數與 LL 的差至多 11。

因此:

∣kN−Nℓ∣≤1.|k_N-N\ell| \le1.

除以 NN 得證。證畢。


17. Latent-Potential Continuum Limit

定義 continuum deficit:

d∞(I)=max⁡(0,α−ℓ).\boxed{ d_\infty(I) = \max(0,\alpha-\ell). }

以及 normalized discrete potential:

δN(I)=dN(AN(I))N.\boxed{ \delta_N(I) = \frac{ d_N(A_N(I)) }{ N }. }

Theorem 17.1 — Uniform O(1/N)O(1/N) Potential Error

∣δN(I)−d∞(I)∣≤1N.\boxed{ \left| \delta_N(I) - d_\infty(I) \right| \le \frac1N. }

Proof

由 Lemma 15.1:

δN(I)=max⁡(0,α−kNN).\delta_N(I) = \max \left( 0, \alpha-\frac{k_N}{N} \right).

函數:

f(x)=max⁡(0,α−x)f(x)=\max(0,\alpha-x)

為 11 -Lipschitz。

因此:

∣f(kN/N)−f(ℓ)∣≤∣kN/N−ℓ∣.\left| f(k_N/N)-f(\ell) \right| \le \left| k_N/N-\ell \right|.

再由 Lemma 16.1:

≤1N.\le \frac1N.

證畢。

因此:

δN(I)→d∞(I).\boxed{ \delta_N(I) \rightarrow d_\infty(I). }

18. Constituent Capacity Field

Paper 03 對 aperture capacity game 得到:

ϕi=ni(hN)hN,\phi_i = \frac{n_i(h_N)}{h_N},

其中:

ni(hN)n_i(h_N)

是包含 constituent ii 的 generators 數。

本文定義:

ΦN(xi)=ni(hN)hN,\boxed{ \Phi_N(x_i) = \frac{n_i(h_N)}{h_N}, }

其中 physical cell center:

xi=i+1/2N.\boxed{ x_i = \frac{i+1/2}{N}. }

19. Exact Discrete Field Formula

因:

0<α≤12,0<\alpha\le\frac12,

有:

hN≤N2.h_N\le\frac N2.

所以:

ni(hN)=min⁡(i+1,hN,N−i).n_i(h_N) = \min \left( i+1, h_N, N-i \right).

因此:

ΦN(xi)=min⁡(1,i+1hN,N−ihN).\boxed{ \Phi_N(x_i) = \min \left( 1, \frac{i+1}{h_N}, \frac{N-i}{h_N} \right). }

20. Continuum Capacity Field

定義:

Φα(x)=min⁡(1,xα,1−xα).\boxed{ \Phi_\alpha(x) = \min \left( 1, \frac{x}{\alpha}, \frac{1-x}{\alpha} \right). }

其形狀為:

  • 左 boundary ramp;
  • central plateau;
  • 右 boundary ramp。

21. Uniform Capacity-Field Limit

Theorem 21.1

對所有 discrete centers:

xi=i+1/2N,x_i = \frac{i+1/2}{N},

有:

∣ΦN(xi)−Φα(xi)∣≤12αN.\boxed{ \left| \Phi_N(x_i) - \Phi_\alpha(x_i) \right| \le \frac{ 1 }{ 2\alpha N }. }

而最大誤差在 boundary ramp 上可達此 bound。

Proof

左 boundary ramp 上:

i+1≤hN,i+1\le h_N,

且該點尚未進入 plateau。

則:

ΦN(xi)=i+1hN.\Phi_N(x_i) = \frac{i+1}{h_N}.

而:

Φα(xi)=xiα=i+1/2αN=i+1/2hN.\Phi_\alpha(x_i) = \frac{x_i}{\alpha} = \frac{i+1/2}{\alpha N} = \frac{i+1/2}{h_N}.

兩者差:

12hN=12αN.\frac{1}{2h_N} = \frac{1}{2\alpha N}.

右 boundary ramp 對稱。

central plateau 中:

ΦN=Φα=1.\Phi_N = \Phi_\alpha = 1.

在 ramp / plateau transition 亦不超過同一 bound。

證畢。


Corollary 21.2 — Uniform Convergence

∥ΦN−Φα∥∞,centers→0.\boxed{ \|\Phi_N-\Phi_\alpha\|_{\infty,\mathrm{centers}} \rightarrow0. }

因此 constituent-level cooperative influence field 不必是 pixel-resolution artifact。


22. Resolution System

Paper 04 的 finite quotient:

[Hq]≅[\mathcal H_q]_{\cong}

在 fixed cardinality 下適用。

本文提出跨解析度 object:

Rα=({HN}N∈Nα,{c2N→N},{ιN},Hα∞).\boxed{ \mathfrak R_\alpha = \left( \{ \mathcal H_N \}_{N\in\mathcal N_\alpha}, \{ c_{2N\rightarrow N} \}, \{ \iota_N \}, \mathcal H_\alpha^\infty \right). }

其中:

  • HN\mathcal H_N:finite generator system;
  • c2N→Nc_{2N\rightarrow N}:winning-preserving coarse map;
  • ιN\iota_N:physical embedding;
  • Hα∞\mathcal H_\alpha^\infty:continuum generator limit。

所以 cross-resolution invariant 不再是一個單一 finite equivalence class,而更接近:

a refinement system with a common physical limit.\boxed{ \text{a refinement system with a common physical limit}. }

23. Constituent Is Not the Invariant

N=16N=16 時:

∣VN∣=16.|V_N|=16.

N=256N=256 時:

∣VN∣=256.|V_N|=256.

離散 constituent identity 已完全改變。

但仍有:

hNN=α,\frac{h_N}{N} = \alpha, KN(δ)=Kα(δ),K_N(\delta) = K_\alpha(\delta), GN→Gα∞,\mathscr G_N \rightarrow \mathscr G_\alpha^\infty, δN→d∞,\delta_N \rightarrow d_\infty, ΦN→Φα.\Phi_N \rightarrow \Phi_\alpha.

因此在本模型中:

constituent identity\boxed{ \text{constituent identity} }

不是最穩定的 information carrier。

更準確地:

constituent=resolution-relative carrier of a deeper admissibility relation.\boxed{ \text{constituent} = \text{resolution-relative carrier of a deeper admissibility relation}. }

24. ACTM:Resolution Transport

現在 refinement:

HN→H2N\mathcal H_N \rightarrow \mathcal H_{2N}

可做 structural audit。

Preserve Exactly

hNN=α.\boxed{ \frac{h_N}{N} = \alpha. }

Preserve as Exact Samples

KN(δ)=Kα(δ).\boxed{ K_N(\delta) = K_\alpha(\delta). }

Refine

SN⊂S2NS_N \subset S_{2N}

對 dyadic admissible resolutions成立。

Map Minimal to Winning

G2N,j↦c(G2N,j),G_{2N,j} \mapsto c(G_{2N,j}),

其中 image 一定 coarse winning。

Bracket Arbitrary States

VN(C−A)≤V2N(A)≤VN(C+A).V_N(C^-A) \le V_{2N}(A) \le V_N(C^+A).

Converge

GN→Gα∞,\mathscr G_N \rightarrow \mathscr G_\alpha^\infty, dN/N→d∞,d_N/N \rightarrow d_\infty, ΦN→Φα.\Phi_N \rightarrow \Phi_\alpha.

因此:

refinement transport\boxed{ \text{refinement transport} }

不是單純 representation rewrite,而是在 constituent system 變動中保存/收斂 relational structure。


25. SOBTA:Coarse Projection without Identity

Theorem 13.1 提供一個很清楚的三域投影例子。

fine state:

AA

是客體層的高解析狀態。

coarse query:

SNS_N

只能透過:

C−,C+C^-, \qquad C^+

讀取一個 lower / upper admissibility envelope。

所以:

coarse observation≠fine state itself.\boxed{ \text{coarse observation} \neq \text{fine state itself}. }

但又不是完全失去結構,因為:

VN(C−A)≤V2N(A)≤VN(C+A).\boxed{ V_N(C^-A) \le V_{2N}(A) \le V_N(C^+A). }

這是:

Projection without Annexation\boxed{ \text{Projection without Annexation} }

的 operational realization。


26. Exact Audit

本文附:

paper05_exact_checks.py

驗證:

  1. normalized generator order;
  2. exact physical generator length;
  3. Hausdorff rate;
  4. overlap kernel;
  5. dyadic generator image;
  6. all-state 16→816\rightarrow8 bracket;
  7. latent-potential error bound;
  8. capacity-field error bound。

其中 coarse-bracket theorem 已在正文解析證明;全 6553665536 states 的 exhaustive check 只作 audit。


27. 可證偽性與限制

F1 — Resolution Normalization Failure

若:

hN/Nh_N/N

不能穩定表示同一 physical scale,本文此模型失去 cross-resolution interpretation。

F2 — Refinement Non-Preservation

若 fine minimal generator 的 coarse image不再是 coarse winning set,Theorem 11.1 / Corollary 11.2 失效。

F3 — Coarse-Bracket Failure

若存在:

AA

使:

VN(C−A)>V2N(A)V_N(C^-A) > V_{2N}(A)

或:

V2N(A)>VN(C+A),V_{2N}(A) > V_N(C^+A),

Theorem 13.1 失效。

F4 — Limit Overreach

本文只證 interval-generator family 的 limit。

不宣稱所有:

  • irregular geometry;
  • curved boundary;
  • rotation group;
  • non-monotone freedom;
  • physical field theory;

都具有同樣 limit structure。

F5 — Constituent Ontology Overreach

本文不宣稱 continuum object 比所有 discrete layers「更真」。

它只表示:

多個 discrete representations 可由同一 continuous admissibility relation 一致生成與逼近。


28. 與下一篇的接口

本文的一維 continuum family 已有:

Gα∞,\mathscr G_\alpha^\infty, Kα,K_\alpha, Φα.\Phi_\alpha.

下一步 Paper 06 將回到真正二維:

  • disk-like probe;
  • square probe;
  • irregular probe;
  • translation;
  • 90∘90^\circ rotation;
  • generator covariogram;
  • container-induced capacity field;
  • dyadic refinement。

目標是把:

interval generator system\boxed{ \text{interval generator system} }

提升成:

2D Generator Complex.\boxed{ \text{2D Generator Complex}. }

29. 結論

Paper 04 說明:

labels can change while a finite Generator Quotient remains fixed.\boxed{ \text{labels can change while a finite Generator Quotient remains fixed}. }

本文進一步說明:

constituent count can change while a physical generator relation converges.\boxed{ \text{constituent count can change while a physical generator relation converges}. }

在 interval-generator model 中,我們得到:

dH(GN,Gα∞)=12N,\boxed{ d_H( \mathscr G_N, \mathscr G_\alpha^\infty ) = \frac{1}{2N}, } Kα(δ)=max⁡(0,α−δ),\boxed{ K_\alpha(\delta) = \max(0,\alpha-\delta), } VN(C−A)≤V2N(A)≤VN(C+A),\boxed{ V_N(C^-A) \le V_{2N}(A) \le V_N(C^+A), } ∣dNN−d∞∣≤1N,\boxed{ \left| \frac{d_N}{N} - d_\infty \right| \le \frac1N, }

以及:

∥ΦN−Φα∥∞≤12αN.\boxed{ \|\Phi_N-\Phi_\alpha\|_\infty \le \frac{1}{2\alpha N}. }

因此在這個明確模型中,不同 resolution hypergraphs 可以被理解成同一 continuum generator relation 的不同 finite carriers。

本文一句話總結:

The constituent grid may refine; the admissible generator relation can converge to the same physical object.\boxed{ \textbf{The constituent grid may refine; the admissible generator relation can converge to the same physical object.} }

中文:

構成格點可以持續細化;被它承載的合法生成關係仍可以收斂到同一個物理尺度結構。


Appendix A. 核心公式

Normalized generator order:

hNN=α.\frac{h_N}{N} = \alpha.

Physical generator:

ιN(GN,j)=[jN,jN+α].\iota_N(G_{N,j}) = \left[ \frac jN, \frac jN+\alpha \right].

Generator-family Hausdorff rate:

dH(GN,Gα∞)=12N.d_H( \mathscr G_N, \mathscr G_\alpha^\infty ) = \frac{1}{2N}.

Overlap kernel:

Kα(δ)=max⁡(0,α−δ).K_\alpha(\delta) = \max(0,\alpha-\delta).

Coarse bracket:

VN(C−A)≤V2N(A)≤VN(C+A).V_N(C^-A) \le V_{2N}(A) \le V_N(C^+A).

Latent-potential limit:

∣δN−d∞∣≤1N.\left| \delta_N-d_\infty \right| \le \frac1N.

Capacity-field limit:

∣ΦN(xi)−Φα(xi)∣≤12αN.\left| \Phi_N(x_i)-\Phi_\alpha(x_i) \right| \le \frac{1}{2\alpha N}.

Appendix B. Formal Status

解析定理:

  • Proposition 4.1;
  • Theorem 5.1;
  • Theorem 7.1;
  • Corollary 8.1;
  • Theorem 9.1;
  • Corollary 9.2;
  • Theorem 11.1;
  • Corollaries 11.2、11.3;
  • Theorem 13.1;
  • Lemmas 15.1、16.1;
  • Theorem 17.1;
  • Theorem 21.1;
  • Corollary 21.2。

Exact computational audit:

  • E6 resolutions 16,32,64,128,25616,32,64,128,256 ;
  • dyadic generator mapping;
  • all-state 16→816\rightarrow8 bracketing over 6553665536 states;
  • latent-potential samples;
  • capacity-field samples。

尚未宣稱:

  • universal continuum limit for arbitrary generator systems;
  • arbitrary-dimensional theorem;
  • arbitrary curved geometry refinement theorem;
  • physical spacetime discreteness or continuity;
  • category-theoretic universality of Rα\mathfrak R_\alpha。