The Setting
Earlier candidate points enter the predictions at a finite frontier of the tree. Their coefficients vary before the next map is chosen, while the predictions are root outputs of the maps already present. Let be a real normed space and a convex subset of . All norms below are the ambient norm of . Write for the same vector space with its weak topology. Weak compactness of means compactness of its image in .✦
An ordered rooted tree has nonempty finite levels , with , and order-preserving surjections assigning each child its parent.✦ A weighted ordered tree assigns positive real weights to the children, with the normalization in (1) at each parent: (1)✦
A tree of nonexpansive maps on assigns a nonexpansive selfmap of to each node .✦ The cutoff evaluation of a tree of nonexpansive maps on is the map sending a frontier assignment to the root value obtained by successive weighted collapses of the levels.✦ This evaluation applies node maps at levels and does not reapply the maps at level ; at level zero it takes the value at the root.✦
Earlier Candidate Coefficients
Let be a finite set indexing every earlier candidate, including discarded candidates, and let be a finite frontier. The map assigns a parent to each candidate. For , the parent-restricted coefficients and parent coefficient masses are the real numbers defined in (2): (2) The admissible coefficient set is the subset of whose coefficients are nonnegative and whose parent masses are at most one, as in (3): (3)✦ Discarded candidates have coefficient zero in the final coefficient choice.
For points and , the predicted parent point is the affine combination in (4), defined for every : (4)✦ When , and , the residual input at is regarded as a point of . This definition applies both to a tree cutoff, where , and to an arbitrary finite frontier.
Evaluated Predictions and Weak Limits
Fix . At a tree cutoff , the evaluated prediction is the root output in (5): (5)✦
Assume is weakly compact. For an indexed family and an ultrafilter on , let be a chosen limit of along , with convergence in .✦ Fix sequences and in , and an ultrafilter on . For , the prediction weak limit is the point defined in (6), with the weak limit taken after evaluating the nonlinear root outputs: (6)✦ The cutoff detector set is the union in (7): (7)✦
For an arbitrary finite frontier , take selfmaps and an evaluator . For , and , the evaluated prediction at the finite frontier , its prediction weak limit , and the detector set are respectively the points and set defined in (8): (8)
The Claim
Theorem 1. Suppose is weakly compact and for all . Fix the weighted tree of nonexpansive maps, the earlier candidates and their parent assignment at cutoff . For and , the evaluated prediction bound is (9): (9)✦ The prediction weak limits associated with any sequence in and any ultrafilter on satisfy (10): (10)✦ For any sequence in , the cutoff detector set has the compactness property (11): (11)✦
For the finite frontier construction, suppose every is nonexpansive and preserves every common pointwise norm bound: for and , the inequalities for all imply . For every , the corresponding compactness property is (12): (12)✦
What Is Checked
The statements concern convex weakly compact subsets of diameter at most one, finite earlier candidate families, and either a weighted ordered tree of nonexpansive maps or a finite frontier evaluator preserving common pointwise norm bounds.