观点

Lemma 3 ([Interpolation]). If A B , then λ(A)≥λ(B) for every tensor character λ .

证明 1证否 0作为观点被引用 2链条图发起推演

发起推演1

证明
This is the interpolation passage from approximate to exact bilinear algorithms~, applied to characters. Let F()=B+ F_1+⋯+^L F_L be the transformed tensor. For each nonzero , it is a restriction of A . The polynomial F()^ ⊗j has degree at most jL and constant coefficient B^ ⊗j . Interpolation at jL+1 distinct nonzero points writes this coefficient as a linear combination of specializations. A linear combination of tensors on common spaces is a restriction of their full direct sum: identify the corresponding variables and absorb each scalar coefficient into one leg. Therefore λ(B)^j≤(jL+1)λ(A)^j. Taking j th roots and letting j→∞ proves the result. The degeneration, and hence L , is fixed during this limit.