The Setting
Switching a consecutive block of leaves changes the root output of a tree of nonexpansive maps. The question is how a continuous linear functional detects that change when every switched leaf is replaced by the same resolvent point.
Let be a real Banach space. A minimal invariant set for a selfmap is a nonempty, norm-closed, bounded, convex, norm-separable, weakly compact subset of such that is nonexpansive, has no fixed point, and every nonempty norm-closed convex subset with equals .✦ All norms are the original ambient norm. Normalized minimal invariant data consist of a minimal invariant set for a selfmap , with , and a sequence dense in for the norm topology. Write for the weakly topologized copy of and for its subset corresponding to ; weak compactness means compactness in the weak topology. Assume is metrizable and the whole closed unit ball of the closed difference span of is weakly compact. (1) A subset of is invariant under if sends every point of into . The closed difference span of is the norm closure in of the real linear span of all differences of two points of .
For an anchor and , the anchored averaging map is the selfmap in (2); for , its unique fixed point is the weighted resolvent . The discrete resolvent uses the positive integer index , whereas the terminal sequence uses an anchor and starts at zero. (2) The generated maps are the least family of selfmaps of containing the identity and every , closed under composition and convex mixing: for maps in the family and with , their mixture sends to .✦
An ordered rooted tree has finite levels with and , and nondecreasing surjective parent maps between successive levels.✦ Its root is at level zero.✦ A node at a positive level retains its depth together with its index✦; the truncation through depth consists of those with .✦ A weighted ordered tree is an ordered rooted tree with a positive edge weight for each child , such that the weights sum to one over the children of every parent.✦ The root path weight of a node of the weighted ordered tree is the product of the edge weights along the path from to that node, as in (3). (3)✦
A tree evaluation on an ordered weighted tree is a family of nonexpansive selfmaps , indexed by the nodes of the tree.✦ For an assignment on level , its child average and its one-level collapse are defined in (4). For an assignment on level and a node , the subtree output is the point of obtained by the collapses in (4) from level to that node✦; is the root output.✦ The maps are applied strictly above the cutoff level, whose values are not mapped again. (4)
The ancestor of at level is denoted by . The descendants of at level are the nodes with .✦ For a node , the inclusive lower endpoint of its descendant block at level is the first descendant index, and the exclusive upper endpoint is one greater than the last descendant index. For leaf values , a point , and a switching index , the switched assignment , its root , and the roots before and after switching the descendant block of are those in (5). (5) For , the ideal cutoff is the assignment on level that uses the node images before and the common residual point thereafter; its evaluated prediction is in (6). (6)
Errors and Detectors
For positive budgets on the positive level nodes, the cumulative minimum is the minimum of one and all budgets through depth , with .✦ The dyadic tolerance is defined in (7). The injective candidate ordinals determine the retained budgets in the same display. These ordinals start at zero and count discarded candidates as well as retained ones. (7)
For a node and a point , the ancestor displacement is the sum of the displacements of under every proper ancestor map, as in (8). The total terminal displacement sums the displacements of all maps at depths less than .✦ For and , the descendant level displacement is zero when ; otherwise it is the sum over the level descendants of of , each multiplied by its root path weight . For a node and a point , the internal displacement sum is the sum of the descendant level displacements over , as in (8). (8)
Switching tree data on a normalized minimal invariant set consist of a weighted ordered tree, a tree evaluation by generated maps with , an anchor , and the candidate ordinals specified in (7), together with the functionals, prediction limits and ultrafilter satisfying (9). Put and abbreviate and for the retained budgets. Each positive level node has a functional with and a prediction limit . There is an ultrafilter on refining the cofinite filter, with the conditions in (9); the weak limit there is taken after the nonlinear root output has been evaluated. Extend and by zero at the root. (9)
The Claim
Theorem 1. For an ordered weighted tree and a tree evaluation on a convex set , the following estimates hold.
For positive node budgets, , and , the tolerance tail satisfies (10). (10)✦
Let and real tolerances satisfy the child-average bounds (11) through depth . (11) For , , and any assignment with on the descendants of , the untouched subtree estimate is (12). (12)✦ For every , , and , the constant subtree estimate is (13). (13)✦
Suppose , , , , and has norm at most one. Assume (11) and, for a real number , the bounds in (14). For the leaf assignment , the actual block estimate is (15). (14) (15)✦
For switching tree data and every , there is an index satisfying (16). (16)✦ For any satisfying (16), the roots for the level assignment satisfy (17) at every node of the nonroot truncation. (17)✦
In the paper, the result reads: a single can be chosen so that every nonroot node through depth satisfies and the sum of all map displacements through depth is at most the minimum budget through depth ; untouched subtrees have error at most , fully switched subtrees have error at most that common displacement bound, and these give .
What Is Checked
The common terminal and finite block assertions are stated over switching tree data on normalized minimal invariant sets; the subtree and actual block estimates are stated over arbitrary weighted ordered trees of nonexpansive selfmaps with the displayed hypotheses.
For , the anchored averaging map is a strict contraction with factor ✦, and positive integer indices satisfy .
Every node of the ordered rooted tree has a nonempty descendant set at every later level.✦