We give an efficient algorithm which can obtain a relative error approximation to the spectral norm of a matrix, combining the power iteration method with some techniques from matrix reconstruction which use random sampling.

We investigate the Lp-spectrum of linear operators defined consistently on Lp(Ω) for p0 ≤ p ≤ p1, where (Ω, μ) is an arbitrary σ-finite measure space and 1 ≤ p0 < p1 ≤ ∞. We prove p-independence of the Lp-spectrum assuming weighted norm estimates. The assumptions are formulated in terms of a measurable semi-metric d on (Ω, μ); the balls with respect to this semi-metric are required to satisfy a...

Let E be a real Banach space. We study the Ornstein-Uhlenbeck semigroup P = {P (t)}t≥0 associated with the Ornstein-Uhlenbeck operator Lf(x) = 1 2 TrQDf(x) + 〈Ax,Df(x)〉, x ∈ E. Here Q ∈ L (E, E) is a positive symmetric operator and A is the generator of a C0-semigroup S = {S(t)}t≥0 on E. Under the assumption that P admits an invariant measure μ∞ we prove that if S is eventually compact and the ...

We present a generalization bound for feedforward neural networks in terms of the product of the spectral norms of the layers and the Frobenius norm of the weights. The generalization bound is derived using a PAC-Bayes analysis.

Following Hartigan [1975], a cluster is defined as a connected component of the t-level set of the underlying density, i.e., the set of points for which the density is greater than t. A clustering algorithm which combines a density estimate with spectral clustering techniques is proposed. Our algorithm is composed of two steps. First, a nonparametric density estimate is used to extract the data...

Abstract Suppose that Tn is a Toeplitz matrix whose entries come from a sequence of independent but not necessarily identically distributed random variables with mean zero. Under some additional tail conditions, we show that the spectral norm of Tn is of the order √ n log n. The same result holds for random Hankel matrices as well as other variants of random Toeplitz matrices which have been st...

In this paper theoretical results regarding a generalized minimum rank matrix approximation problem in the spectral norm are presented. An alternative solution expression for the generalized matrix approximation problem is obtained. This alternative expression provides a simple characterization of the achievable minimum rank, which is shown to be the same as the optimal objective value of the c...

Let A be an n × n nonnegative matrix. In this paper we consider the problems of maximizing the spectral radii of (i) A + X and (ii) A + D, where X is a real n × n matrix whose Frobenius norm is restricted to be 1 and where D is as X but is further constrained to be a diagonal matrix. For both problems the maximums occur at nonnegative X and D, and we use tools of nonnegative matrices, most nota...