The Setting
Take the stationary Gaussian autoregressive chain in , with coefficient in coordinate , and a convex body around the origin. The probability that the chain stays inside a convex body around the origin for all steps is its survival probability. Lemma 2.3 asks how much of that probability is lost when the body is cut by slabs, sets for a real matrix whose Hilbert-Schmidt norm (the square root of the sum of the squares of its entries) is at most one and a fixed radius , and how the loss per step depends on weights attached to the coordinates.
The Chain and Its Slabs
Write for a point of and for the standard Gaussian measure on that space. The autoregressive chain with coefficients up to step is the sequence of linear maps from to given coordinatewise by (1)✦
The Hilbert-Schmidt norm of a real matrix with columns is the square root of the sum of the squares of its entries. The weighted square of by an matrix is the number ; for with it equals .✦ The slab radius is the constant .✦
Filters and the Application Domain
A filter step on matrices is a triple of a real matrix , a vector and a coordinate of ; it acts on a matrix by (2)✦ where is the -th coordinate vector of . A good step is a filter step with invertible and .✦ The run of a list of filter steps from is the matrix obtained from by applying in this order.✦ A list of filter steps is fresh for a matrix when the column of at the coordinate of each step of is zero.✦
Two constants enter the survival estimate in the application domain defined below. The body rate is (3) with state rate , epoch factor , gain constant and exponent . The survival radius is (4)✦ with moment constant and increment constant , and the state radius is ✦.
For a family of matrices and a finite family of rows in , the application domain of the two families is the set of points that satisfy the open norm constraints (5)✦
The Claim
Theorem 1 (Adding slabs). Let and .
(a) Let satisfy for every coordinate , let , and let be the autoregressive chain with coefficients . If is a real matrix with columns and , then (6)✦ If is a finite family of real matrices with columns and for every , then (7)✦
(b) If is a finite family of rows in with for every coordinate , then (8)✦
(c) Suppose for every . Let be a matrix with orthonormal rows, let be a list of good filter steps that is fresh for and whose steps have distinct coordinates , and let each matrix of a family be the run from of an initial segment of . Let be a finite family of rows in with for every and for every coordinate , and let be the application domain of the two families. Then there is a real number such that, for every , the autoregressive chain with coefficients satisfies (9)✦
In the paper, the result reads: if is bounded, nonempty, open, convex and centrally symmetric, , , are real rows and , then for every positive definite diagonal , where is the paper’s diagonal Dirichlet energy. The library states the estimate for survival of the autoregressive chain, with the rate in (7) standing where the paper has its slab term.
What Is Checked
The theorem is stated over the autoregressive chain (1) under the standard Gaussian measure, the Hilbert-Schmidt norm and the weighted square, the slab radius , good filter steps with their runs and freshness, the application domain (5), and the constants and of (3) and (4).