The Setting
A finite prediction is selected before the retained children of a tree of nonexpansive maps have all been fixed. Its coefficients are indexed by candidate order, which includes candidates that will be discarded.
Let be a real normed space and a convex set. An ordered rooted tree is a tree with levels of positive finite sizes , with , given by (1). Its parent maps are order-preserving and onto. (1)✦ A weighted ordered tree is an ordered rooted tree with a positive weight for each child , satisfying the normalization in (2) at every parent .✦ (2)✦
A tree evaluation on is a family of selfmaps , one for each node , that are nonexpansive in the original ambient norm: for .✦ The cutoff evaluation of an assignment is the point of obtained by combining children with their weights and applying the node maps strictly above the cutoff, toward the root.✦ At level zero it returns .
Candidates and Prediction Parameters
Fix a level . A complete candidate ledger at level consists of a finite candidate ordinal , where is the number of candidates, an order-preserving parent assignment , points for , and an order-preserving injection satisfying (3).✦ The image of consists of the retained candidates. (3)
For a retained node , its earlier candidate set , the inclusion , and the restricted parent and point families and are defined in (4). The retained point at is denoted by there. (4)
The extended final coefficient family assigns the retained edge weights to their candidate positions and zero elsewhere.✦ The actual parameter at a retained node is the coefficient family on obtained by restricting the extended final coefficient family , as in (5).✦ (5)
For a finite candidate set , a parent assignment , and , the parent-restricted coefficient and the parent coefficient mass are given by (6). The admissible coefficient set is the subset of consisting of nonnegative coefficient families with mass at most one at every parent .✦ (6)
For points , a common point , and real coefficients , the affine residual parent point is the point of given by (7).✦ With and , the cutoff residual input is the affine residual parent point in (7).✦ (7)✦ The evaluated prediction for a parent assignment , points , and coefficients is the root output in (8).✦ (8)✦
The ideal cutoff output for and is the assignment in (9).✦ (9)✦
The Claim
Theorem 1. Let be a complete candidate ledger at level of a weighted ordered tree, and let be a tree evaluation on the convex set . For every retained node , the actual parameter satisfies (10): (10)✦ For every , the two root outputs agree as in (11): (11)✦
What Is Checked
The assertions are stated over the complete finite candidate ledger, with its order-preserving retention map, the positive normalized edge weights, and the nonexpansive node maps on a convex subset of the original normed space. Every cutoff residual input with admissible coefficients belongs to .