Honest finite assembly for the fixed-class inner sum #
This file supplies the finite-sum layer between the local copy estimate in
SingleScaleBridge and the parity bookkeeping in paper (5.87)--(5.92).
The carrier is the paper's labeled-copy carrier
Σ l : HeppLeaf t, Fin (leafMultiplicity mu l).
Conditioning on copies already used by a partial permutation is represented
by deleting a finite set from each (N,X) fiber. The resulting one- and
two-variable sums are proved to be sub-sums of the unrestricted copy sums;
no independence or product-separability of the two edge kernels is assumed.
The last section isolates the purely numerical parity ledger: one parity can
lose at most twenty gains, the two parity exceptional sets have union of
cardinality at most forty, and geometric-mean interpolation changes the gain
exponent from 1/8 to 1/16.
Labeled copies and conditioned local choices #
The distinct copies of all leaves. This is the codomain of a labeled arrangement in the factorial ledger.
Equations
- Anderson4D.XYCluster.HeppLabeledCopy mu = ((l : Anderson4D.HeppLeaf t) × Fin (Anderson4D.leafMultiplicity mu l))
Instances For
The lattice point carried by a labeled copy.
Equations
Instances For
All labeled copies whose leaf lies in the fixed (N,X) fiber a.
Equations
- Anderson4D.XYCluster.labeledCopiesAtNX Nm mu a = {c : Anderson4D.XYCluster.HeppLabeledCopy mu | Anderson4D.singleScaleSigma1 Nm mu c.fst = a}
Instances For
Copies in a fixed fiber which have not occurred in the already exposed part of a labeled permutation.
Equations
- Anderson4D.XYCluster.conditionedCopiesAtNX Nm mu used a = Anderson4D.XYCluster.labeledCopiesAtNX Nm mu a \ used
Instances For
Link to the factorial-ledger arrangement carrier #
Copies occupying a finite set of positions in a labeled arrangement.
Equations
Instances For
A position not yet exposed by the backward summation is one of the
conditioned local choices in its fixed (N,X) class.
The induced leaf word of every labeled arrangement is an admissible word in the factorial ledger. This records that the local carrier above does not introduce a different notion of "valid word".
Two distinct, not-yet-exposed positions give a legal conditioned pair: the second choice sees both the previously used positions and the first new copy. This is the exact local selection made in a labeled permutation.
A sum over all labeled copies in a fiber is exactly the multiplicity
weighted leaf sum used by nxCopyWeightedSum.
Dropping already-used copies only decreases a nonnegative local sum.
A conditioned two-variable block #
The honest conditional local sum for two consecutive positions of a
labeled permutation. The second copy is selected after inserting the first
one into used, so the two positions can never reuse the same labeled copy.
The first copy may nevertheless have the same underlying leaf as the second
when the incoming edge is skipped, exactly as in Proposition 5.10.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Forgetting both the used-copy condition and the inequality between the
two labeled copies enlarges the conditional block to the unrestricted
copy-level sum from SingleScaleBridge.
A conditioned one-variable block #
The honest conditional local sum for a position that cannot be paired in the chosen parity.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The exact dyadic right-hand factor in the one-variable estimate (5.92), before a universal comparison constant.
Equations
Instances For
Copy-level, conditioned version of the one-variable rough estimate
(5.92). The explicit constant 16 audits the two copies-to-points loss,
the dyadic cardinality loss, and the use of the separation scale N/2.
The rough two-variable block (5.90) #
The dyadic target in (5.90) is the product of its two one-variable (5.92) factors.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Conditioned copy-level form of the rough pair estimate (5.90). It is
obtained by two honest one-variable eliminations, hence the audited constant
16² = 256.
Backward elimination of a whole parity #
Two consecutive positions eliminated together. skipLeft refers to
the edge from the already fixed point into left; skipRight refers to the
edge between the two positions.
- left : ActiveNXClass Nm mu
- right : ActiveNXClass Nm mu
- skipLeft : Bool
- skipRight : Bool
Instances For
The exact paper dyadic target attached to a paired elimination block.
Equations
- Anderson4D.XYCluster.nxPairBlockTarget Nm mu p = Anderson4D.XYCluster.paperDyadicPairTarget Nm mu (↑p.left) (↑p.right) p.skipLeft p.skipRight
Instances For
The actual nested finite sum obtained by eliminating consecutive pairs
from left to right. used is threaded through the recursion, so choices in
different blocks are also distinct. The endpoint of a block becomes the
fixed preceding point for the next block; this is the finite Fubini step
behind the paper's instruction to sum in decreasing position order.
Equations
- One or more equations did not get rendered due to their size.
- Anderson4D.XYCluster.conditionedNXPairChainSum ht hroot Nm mu z hz R [] x✝¹ x✝ = 1
Instances For
One-parity fixed-class assembly. The sum is over honest distinct labeled-copy choices and is nested in the elimination order. Every local bound is the uniform copy-level (5.91) theorem; induction, rather than an assumed factorization, yields the product of dyadic targets.
One- and two-variable blocks in the same parity #
The elimination schedule for one parity. Interior positions occur in
pair blocks; parity endpoints and the split around the minimal-scale
position occur in single blocks.
- single {t : PlaneTree} {Nm : t.HeppMarking} {mu : t.Multiplicities} (a : ActiveNXClass Nm mu) (skipped : Bool) : NXParityBlock Nm mu
- pair {t : PlaneTree} {Nm : t.HeppMarking} {mu : t.Multiplicities} (p : NXPairBlock Nm mu) : NXParityBlock Nm mu
- roughPair {t : PlaneTree} {Nm : t.HeppMarking} {mu : t.Multiplicities} (p : NXPairBlock Nm mu) : NXParityBlock Nm mu
Instances For
Local target for a mixed one-parity schedule.
Equations
- One or more equations did not get rendered due to their size.
- Anderson4D.XYCluster.nxParityBlockTarget Nm mu R (Anderson4D.XYCluster.NXParityBlock.single a skipped) = Anderson4D.XYCluster.paperDyadicSingleRoughTarget Nm mu R (↑a) skipped
- Anderson4D.XYCluster.nxParityBlockTarget Nm mu R (Anderson4D.XYCluster.NXParityBlock.pair p) = Anderson4D.XYCluster.nxPairBlockTarget Nm mu p
Instances For
Honest nested sum for an arbitrary schedule of single and paired eliminations. As above, the used-copy set is threaded through every choice.
Equations
- One or more equations did not get rendered due to their size.
- Anderson4D.XYCluster.conditionedNXParityChainSum ht hroot Nm mu z hz R [] x✝¹ x✝ = 1
Instances For
Mixed one-parity assembly. This is the paper-facing finite elimination theorem for a fixed class sequence: it allows the at most four one-variable boundary blocks (5.92) and all two-variable blocks (5.91) in one nested sum. The universal constant is outside every schedule and every conditioning set.
The twenty/forty-exception parity ledger #
A conservative explicit certificate for the gains spoiled by the rough
and one-variable local estimates in one parity. Six interior operations may
touch three neighboring gains each and one boundary operation may touch two,
which is the paper's 6 · 3 + 1 · 2 = 20 audit.
- interiorOps : Finset κ
- boundaryOps : Finset κ
- affected : κ → Finset ι
- affected_interior (k : κ) : k ∈ self.interiorOps → (self.affected k).card ≤ 3
- affected_boundary (k : κ) : k ∈ self.boundaryOps → (self.affected k).card ≤ 2
Instances For
All gains discarded by the certified exceptional local operations.
Equations
- d.exceptions = d.interiorOps.biUnion d.affected ∪ d.boundaryOps.biUnion d.affected
Instances For
One parity discards at most twenty gains, with the constant obtained from the explicit local-operation ledger rather than assumed as a field.
Combining the even and odd exceptional ledgers loses at most forty gain indices.
Interpolation changes 1/8 to 1/16 #
A generic ratio gain, used only for the interpolation identity.
Equations
- Anderson4D.XYCluster.ratioGain θ r = min 1 (r ^ θ)
Instances For
Square root commutes with the retained finite product of 1/8 gains,
turning every retained factor into exponent 1/16.
The retained even and odd index sets join to the complement of the union of their exceptions.
Exact product form of even/odd interpolation after deleting their two
exceptional sets. Together with the following geometric-mean lemma this is
the formal content of the paper's transition from at most twenty exceptions
per parity to at most forty exceptions and exponent 1/16.
If a nonnegative quantity satisfies the even and odd estimates, it is bounded by their geometric mean. This lemma performs the interpolation rather than assuming it as an opaque global summation step.
Remaining statement-boundary obligations for the literal (5.87) #
The analytic and finite-Fubini parts are closed above. The following position-level bookkeeping is intentionally not hidden in a hypothesis here:
- construct, from
Fin m, the split at the minimal-scale position and the even/oddNXParityBlockschedules, proving at most foursingleblocks; - select at most three skipped-edge blocks as
roughPairand prove that their explicit(N/R)²factors give(N_root/R)^(min (2s) 3); - map those single/rough blocks to a concrete
ParityExceptionLedger, and discharge its6interior plus1boundary operation bounds; - compare the heterogeneous
single/roughPairtargets with the common refined target in (5.87), paying the explicitY^(1/2)andΞ⁻¹losses, proving that at most twenty such factors occur and absorbing the resultingm^20into a universal exponential; - supply the mirrored elimination theorem on the left of the minimal-scale
anchor (or an equivalent reversal argument), including the reversed
ratio gain, the shared
usedset, and the anchor/cross-side distinct-copy bookkeeping; - identify the arrangement sum after fixing the
(N,X)word with the corresponding conditioned chain, then reinsertR^(2s)and the factorial ledger at the statement boundary.
Thus no theorem in this file is named as Proposition 5.10 or as (5.87) itself; the declarations above are the strict, reusable middle layer needed to close that final position-level assembly.