The Setting
Section 2 of the paper builds the signed walk, and its Proposition 2.1 is applied there to one convex body of coefficient vectors , where is a finite index set of columns. State constraints cut that body out together with a coefficient ball and predictor slabs: the state constraints bound the norms for matrices that a filter, a sequence of steps defined below, produces from an initial matrix one step at a time, the ball bounds , and the slabs bound for rows of norm at most one.
The question is how far this body has to be dilated to become a good convex body, a notion built on heat operators compressed to the body. The item fixes that dilation, the body scale , along with the chain of constants behind it.
Filter Steps and Their Runs
Fix finite index sets for rows and for columns. A filter step on real matrices is a triple of a real matrix , a vector and a column , acting on a matrix by (1)✦ with the -th standard basis vector of . A filter step is good if is invertible and (2)✦
The run of a list of filter steps on a matrix is the matrix ; the run of the empty list is .✦ A list of filter steps has fresh columns for a matrix if each step of writes into a zero column of : for every row .✦
The Application Body
The constants of the body are real numbers defined in a chain from the exponent . The gain constant , the epoch factor , the state rate and the body rate are (3) the increment constant , the moment constant , the survival radius and the slab radius are (4) the state radius is ✦, and the body scale is (5)✦
For matrices indexed by and rows indexed by , the application domain is the set of vectors in the coefficient space with (6)✦
The squared Hilbert-Schmidt norm of a real matrix is the sum of the squares of its entries.✦
Good Convex Bodies
For a measurable function on with almost everywhere, the multiplication operator is the bounded operator on the complex space .✦ For a measurable real function with , the Fourier multiplier operator with multiplier is , with the unitary Fourier transform.✦ For a measurable function , the heat operator is the Fourier multiplier operator with multiplier ✦, and for a measurable set the compressed heat operator is .
For a weight , the Gaussian kernel symbol is the function (7)✦ A measurable set satisfies the diagonal heat bound with constant if, for every weight , the operator norm on obeys (8)✦ A good convex body is a set in the coefficient space that is nonempty, open, bounded and convex, with for every , and that satisfies the diagonal heat bound with constant .
The Claim
Theorem 1. Let be a real matrix with , and let be a list of good filter steps that has fresh columns for and in which no two steps write into the same column. For each in an index set , let be the run on of a prefix of . Let , for in a finite set , be rows with and with for every column , and let be the application domain of the matrices and the rows . Then the dilation of by the body scale, (9)✦ is a good convex body, and the body scale satisfies (10)✦
In the paper, the result reads: the body , built from a domain , rows and a width , satisfies the all-diagonal energy bound for the diagonal Dirichlet energy and every positive definite diagonal , the constants and being fixed so that it does.
What Is Checked
The claim is stated over filter steps, their goodness (2), their runs and fresh columns; over the application domain (6), with its open norm constraints, the state radius , the slab radius and the Hilbert-Schmidt norm of its rows; over the body scale (5) and the constants (3) and (4) behind it; and over good convex bodies, defined through the diagonal heat bound (8) on compressed heat operators.
About these constructions the library also proves that the Gaussian kernel symbol is nonnegative at every frequency✦, and that for a measurable function with almost everywhere and every , the product lies in ✦ and .✦