Statements from FPRD Lab papers

Theorems

Search theorem, lemma, proposition, and corollary statements from FPRD Lab publications. Each entry links to the statement in its complete paper.

61 theorems · 18 lemmas · 19 propositions · 9 corollaries

37 of 107 statements

State equality preserves every future

If δ∗(s0,u)=δ∗(s0,u′)\delta^*(s_0,u)=\delta^*(s_0,u') , then for all v∈Σ∗v\in\Sigma^* , uv∈L(A)⟺u′v∈L(A).uv\in L(\mathcal A)\quad\Longleftrightarrow\quad u'v\in L(\mathcal A).

propositionProposition 3.5

Regular languages are closed under Boolean operations

If L1L_1 and L2L_2 are recognized by finite DFAs, then so are L1∩L2L_1\cap L_2 , L1∪L2L_1\cup L_2 , and Σ∗∖L1\Sigma^*\setminus L_1 .

Myhill–Nerode theorem, residual form

For a language L⊆Σ∗L\subseteq\Sigma^* , the following are equivalent: LL is recognized by a finite DFA; LL has finitely many left residuals; the relation ≡L\equiv_L has finite index. When these conditions hold, the residual automaton is the unique minimal reachable DFA up to isomorphism.

Finite intersection characterization

A space XX is compact if and only if every family of closed sets {Ci:i∈I}\{C_i:i\in I\} with the finite intersection property has nonempty total intersection. The finite intersection property means Ci1∩⋯∩Cik≠∅C_{i_1}\cap\cdots\cap C_{i_k}\ne\varnothing for every finite subfamily.

Let XnX_n be a nonempty finite discrete space for every n∈Nn\in\N . Then ∏n∈NXn\prod_{n\in\N}X_n is compact.

propositionProposition 8.2

Universal property of the inverse limit

Let YY be a set and suppose maps fn:Y→Xnf_n:Y\to X_n form a compatible cone: ρn+1,n∘fn+1=fnfor every n.\rho_{n+1,n}\circ f_{n+1}=f_n \qquad\text{for every }n. There is a unique map f:Y⟶lim←⁡nXnf:Y\longrightarrow\varprojlim_nX_n such that πn∘f=fn\pi_n\circ f=f_n for every coordinate projection πn\pi_n . If the spaces are topological and every fnf_n…

propositionProposition 8.5

Closedness of compatibility

For continuous maps between Hausdorff spaces, the set of compatible threads is closed in ∏nXn\prod_n X_n .

Finite discrete inverse-limit existence

Let every XnX_n be a nonempty finite discrete space, with arbitrary bonding maps ρn+1,n:Xn+1→Xn\rho_{n+1,n}:X_{n+1}\to X_n . Then lim←⁡nXn\varprojlim_n X_n is nonempty.

Contraction fixed point

If B\mathcal B is complete and FF is a contraction, then FF has a unique fixed point. Starting from any x0x_0 , the iterates xn+1=F(xn)x_{n+1}=F(x_n) converge to it.

Dyck residual classification

The residuals of DkD_k are precisely {Rs:s∈{1,…,k}∗}∪{∅}.\{R_s:s\in\{1,\ldots,k\}^*\}\cup\{\varnothing\}. Distinct stacks give distinct residuals.

Fixed-cut state minimum

The minimum number of reachable cut states in a deterministic strict layered recognizer is the number of distinct rows of Cm,nC_{m,n} .

Unique language from compatible cuts

Every compatible Boolean cut field determines a unique language LC={w:C∣w∣,0(w,ε)=1}L_C=\{w:C_{|w|,0}(w,\eps)=1\} such that Cm,n(u,v)=1[uv∈LC]C_{m,n}(u,v)=\mathbf 1[uv\in L_C] for every cut. Conversely, every language produces a compatible field.

Global adjacent plateau

If qL(N)=qL(N+1)q_L(N)=q_L(N+1) for some NN , then LL is regular and QN(L)Q_N(L) is the state set of its minimal residual DFA.

propositionProposition 19.1

Basic access laws

For every language over a finite alphabet: aL(N)a_L(N) exists; aL(N)a_L(N) is nondecreasing in NN ; and if LL has a reachable kk -state DFA, then aL(N)≤k−1a_L(N)\le k-1 .

Observation-growth gap

If LL is nonregular, then qL(N)q_L(N) is strictly increasing and qL(N)≥N+2q_L(N)\ge N+2 for every N≥0N\ge0 . Equivalently, if qLq_L ever fails to increase at an adjacent pair of horizons, then it remains bounded.

Bounded access characterizes regularity

For a language LL , the following are equivalent: L is regular⟺sup⁡NaL(N)<∞⟺aL(N) is eventually constant.L\text{ is regular} \quad\Longleftrightarrow\quad \sup_N a_L(N)<\infty \quad\Longleftrightarrow\quad a_L(N)\text{ is eventually constant}.

Exact access calculus for controlled Dyck presentations

For every finite controlled-Dyck presentation and every horizon NN , aLM(N)=max⁡K∈SNαM,N(K).a_{L_M}(N)=\max_{K\in S_N}\alpha_{M,N}(K). The quantity is effectively computable by finite balanced-distance relaxation, the stack recurrence above, and the finite controlled-Dyck shadow quotient. Moreover, aLM(N)≤max⁡{1,(N+1)BM+N}.\boxed{ a_{L_M}(N)\le\max\{1,(N+1)B_M+N\}. }…

The linear access certificate is pointwise sharp

For every horizon N≥0N\ge0 , some finite controlled-Dyck presentation attains the displayed upper bound. For every N≥1N\ge1 , one can arrange BM=2B_M=2 and aLM(N)=3N+2=(N+1)BM+N.a_{L_M}(N)=3N+2=(N+1)B_M+N. The presentation is allowed to depend on NN .

Entry diameter is monotone under DFA quotient

For every such morphism, BM′ent≤BMent.B_{M'}^{\mathrm{ent}}\le B_M^{\mathrm{ent}}. Consequently the reachable minimal DFA of one fixed regular language has no larger entry diameter than any DFA recognizing that regular language.

Dyck-trace minimax certificate

For every controlled-Dyck language inside the fixed envelope DD , aL(N)≤max⁡{1,(N+1)βD(L)+N}.a_L(N)\le\max\{1,(N+1)\beta_D(L)+N\}. The number βD(L)\beta_D(L) depends only on LL and the selected Dyck envelope, not on one chosen regular controller.

Balanced residual access lower bound

Every regular controller presenting LL satisfies BMent≥γD(L),B_M^{\mathrm{ent}}\ge\gamma_D(L), and therefore βD(L)≥γD(L).\beta_D(L)\ge\gamma_D(L).

Entry-context residual lower bound

Every regular controller presenting LL satisfies BMent≥ηD(L).B_M^{\mathrm{ent}}\ge\eta_D(L). Consequently βD(L)≥ηD(L)≥γD(L).\boxed{ \beta_D(L)\ge\eta_D(L)\ge\gamma_D(L). }

Finite completion-summary saturation

For every controlled-Dyck language, ηD(L)\eta_D(L) is exactly computable from any finite controller presentation by a terminating finite saturation.

Minimax sandwich stopping rule

If bM=ηD(L)b_M=\eta_D(L) , then the supplied presentation attains the global Dyck-trace minimum and βD(L)=ηD(L)=bM.\beta_D(L)=\eta_D(L)=b_M. Both endpoints are computed by terminating finite procedures.

First-flip lower bound

Every regular controller presenting LL satisfies BMent≥τD(L),B_M^{\mathrm{ent}}\ge\tau_D(L), and therefore βD(L)≥τD(L).\beta_D(L)\ge\tau_D(L). Moreover, βD(L)=0\beta_D(L)=0 exactly for L=∅L=\varnothing and L=DL=D .

Punctured-threshold separation

For every r≥2r\ge2 , τD(Lr∘)=2<γD(Lr∘)=βD(Lr∘)=2r.\boxed{ \tau_D(L_r^\circ)=2 <\gamma_D(L_r^\circ) =\beta_D(L_r^\circ)=2r. }

First-component contextual separation

For every r≥3r\ge3 , τD(Hr)=2,γD(Hr)=4,ηD(Hr)=βD(Hr)=2(r−1).\boxed{ \tau_D(H_r)=2, \qquad \gamma_D(H_r)=4, \qquad \eta_D(H_r)=\beta_D(H_r)=2(r-1). } In particular, βD(Hr)>γD(Hr)\beta_D(H_r)>\gamma_D(H_r) for r≥4r\ge4 .

Terminal-descent sewing collapse

For the trace K3K_3 , ηD(K3)=βD(K3)=4.\boxed{\eta_D(K_3)=\beta_D(K_3)=4.} The three-state presentation is state-minimal but not access-optimal.

Exact state/access frontier of K3K_3

(Exact state/access frontier of K3K_3 ). The minimum size of a controller presenting K3K_3 is three, and νD(K3;4)=6.\nu_D(K_3;4)=6. The nondominated state/access pairs are exactly (3,6)and(6,4).\boxed{(3,6)\quad\text{and}\quad(6,4)}.

Terminal-descent family law

For every m≥2m\ge2 , ηD(K2)=βD(K2)=4,ηD(Km)=βD(Km)=2(m−1)(m≥3).\eta_D(K_2)=\beta_D(K_2)=4, \qquad \eta_D(K_m)=\beta_D(K_m)=2(m-1)\quad(m\ge3).

Even positional construction

For every even mm , an (m+2)(m+2) -state controller presents KmK_m at the optimal entry cost βD(Km)\beta_D(K_m) . At m=4m=4 it is isomorphic to the six-state controller displayed above.

Exact access radius for square lengths

For every m≥0m\ge0 , aL□(2m)=(m+1)2+1,aL□(2m+1)=(m+1)2+2.a_{L_{\square}}(2m)=(m+1)^2+1, \qquad a_{L_{\square}}(2m+1)=(m+1)^2+2.

Exact depth-threshold profiles

Let Dk≥rD_{k}^{\ge r} be the balanced words over kk bracket types whose stack height reaches at least r≥1r\ge1 . Then qk,r≥(N)=1+∑h=0Nkh+∑h=max⁡(0,2r−N)r−1kh,q_{k,r}^{\ge}(N) =1+\sum_{h=0}^{N}k^h +\sum_{h=\max(0,2r-N)}^{r-1}k^h, ak,r≥(N)=max⁡(2r,N).a_{k,r}^{\ge}(N)=\max(2r,N). For the complementary balanced language Dk<rD_k^{<r} that stays below rr ,…