On finite {s-2, s}-semiaffine linear spaces

In this paper finite {s-2, s}-semiaffine linear spaces of order n are studied. It is proved that if s= 6 or then there is only a finite number of such linear spaces. Received 28 May 1999; revised 28 December 1999.     

Universal extrapolation spaces for \hbox {C}_{0}-semigroups

The classical theory of Sobolev towers allows for the construction of an infinite ascending chain of extrapolation spaces and an infinite descending chain of interpolation spaces associated with a given \(C_0\) -semigroup on a Banach space. In this note we first generalize the latter to the case of a strongly continuous and exponentially equicontinuous semigroup on a complete locally convex space. As a new concept—even for \(C_0\) -semigroups on Banach spaces—we then define a universal extrapolation space as the completion of the inductive limit of the ascending chain. Under mild assumptions we show that the semigroup extends to this space and that it is generated by an automorphism of the latter. Dually, we define a universal interpolation space as the projective limit of the descending chain. We show that the restriction of the initial semigroup to this space is again a semigroup and always has an automorphism as generator.     

Finite Partially {0, 1}-Semiaffine Linear Spaces

In this paper finite, partially proper {0,1}-semiaffine, planes of order n are studied and completely characterized. Finite, partially {0}-semiaffine, planes are completely classified and finite, partially {1}-semiaffine, planes are classified for bn ²+n+1.     

Comments on some recent results concerning {2, 3} and {2, 4}-generalized inverses

We give a comment on some recent results concerning the representations of generalized {2, 3} and {2, 4}-inverses. Shorter proofs of some previous results are presented.     

{0, 1/2}-Chvátal-Gomory cuts

Given the integer polyhedronP _t := conv{x ∈ℤ ⁿ :Ax⩽b}, whereA ∈ℤ ^{m × n} andb ∈ℤ ^m , aChvátal-Gomory (CG)cut is a valid inequality forP ₁ of the type λτAx⩽⌊λτb⌋ for some λ∈ℝ ₊ ^m such that λτA∈ℤ ⁿ . In this paper we study {0, 1/2}-CG cuts, arising for λ∈{0, 1/2} ^m . We show that the associated separation problem, {0, 1/2}-SEP, is equivalent to finding a minimum-weight member of a binary clutter. This implies that {0, 1/2}-SEP is NP-complete in the general case, but polynomially solvable whenA is related to the edge-path incidence matrix of a tree. We show that {0, 1/2}-SEP can be solved in polynomial time for a convenient relaxation of the systemAx<-b. This leads to an efficient separation algorithm for a subclass of {0, 1/2}-CG cuts, which often contains wide families of strong inequalities forP ₁. Applications to the clique partitioning, asymmetric traveling salesman, plant location, acyclic subgraph and linear ordering polytopes are briefly discussed.     

Maximal three-independent subsets of {0, 1, 2} n

We consider a variant of the classical problem of finding the size of the largest cap in ther-dimensional projective geometry PG(r, 3) over the field IF₃ with 3 elements. We study the maximum sizef(n) of a subsetS of IF ₃ ⁿ with the property that the only solution to the equationx ₁+x₂+x₃=0 isx ₁=x₂=x₃. Letc _n=f(n)^1/n andc=sup{c ₁, c₂, ...}. We prove thatc>2.21, improving the previous lower bound of 2.1955 ...     

On the {H^1}-stability of the {L_2}-projection onto finite element spaces

We study the stability in the $H^1$ -seminorm of the $L_2$ -projection onto finite element spaces in the case of nonuniform but shape regular meshes in two and three dimensions and prove, in particular, stability for conforming triangular elements up to order twelve and conforming tetrahedral elements up to order seven.     

Interpolation spaces between(L_1^{W_0 } ,L_1^{W_1 } ) and (L1, L∞)

Let A₀, A₁ be a pair of normed spaces, having the property that the difference K(x, t; A₀, A₁) ?K(x, s; A₀, A₁) regarded as a function of xε A₀+ A₁ is a seminorm for t>s (here K is the Oklander-Peetre functional). All the pairs A, L of normed spaces, such that, if a linear operator is bounded from A₀ into L₁ and from A₁ into L_∞, then it is bounded from A into L, are characterized in the following article.     

Independence entropy of \mathbb{Z}^{d}-shift spaces

We introduce a concept of independence entropy for symbolic dynamical systems. This notion of entropy measures the extent to which one can freely insert symbols in positions without violating the constraint defined by the shift space. We show that for a certain class of one-dimensional shift spaces X, the independence entropy coincides with the limiting, as d tends to infinity, topological entropy of the dimensional shift defined by imposing the constraints of X in each of the d cardinal directions. This is of interest because for these shift spaces independence entropy is easy to compute. Thus, while in these cases, the topological entropy of the d-dimensional shift (d≥2) is difficult to compute, the limiting topological entropy is easy to compute. In some cases, we also compute the rate of convergence of the sequence of d-dimensional entropies. This work generalizes earlier work on constrained systems with unconstrained positions.     

Subspaces and quotient spaces of(\Sigma G_n )_{l_p } and(\Sigma G_n )_{c_0 }

Let (G _n) be a sequence which is dense (in the sense of the Banach-Mazur distance coefficient) in the class of all finite dimensional Banach spaces. Set \(C_p = (\Sigma G_n )_{l_p } (1< p< \infty ) = (\Sigma G_n )_{c_0 } \) . It is shown that a Banach spaceX is isomorphic to a subspace ofC _p (1<p≦∞) if and only ifX is isomorphic to a quotient space ofC _p.     

Generalized {G_\delta}-submaximal spaces

We study some properties of generalized submaximal spaces and introduce the notion of generalized door spaces. Furthermore we extend these notions to generalized \({G_\delta}\)-submaximal spaces by replacing \({\mu}\)-open sets by \({\mu-G_\delta}\)-sets and consider some of their properties.     

A threshold phenomenon for embeddings of $${H^m_0}$$ into Orlicz spaces

Given an open bounded domain \({\Omega\subset\mathbb {R}^{2m}}\) with smooth boundary, we consider a sequence \({(u_k)_{k\in\mathbb{N}}}\) of positive smooth solutions to

$\left\{\begin{array}{ll} (-\Delta)^m u_k=\lambda_k u_k e^{mu_k^2} \quad\quad\quad\quad\quad {\rm in}\,\Omega\\ u_k=\partial_\nu u_k=\cdots =\partial_\nu^{m-1} u_k=0 \quad {\rm on }\, \partial \Omega, \end{array}\right.$

where λ _k → 0⁺. Assuming that the sequence is bounded in \({H^m_0(\Omega)}\) , we study its blow-up behavior. We show that if the sequence is not precompact, then

$\liminf_{k\to\infty}\|u_k\|^2_{H^m_0}:=\liminf_{k\to\infty}\int\limits_\Omega u_k(-\Delta)^m u_k dx\geq \Lambda_1,$

where Λ₁ = (2m ? 1)!vol(S ^2m ) is the total Q-curvature of S ^2m .

     

Full-rank representations of {2, 4}, {2, 3}-inverses and successive matrix squaring algorithm

We present the full-rank representations of {2, 4} and {2, 3}-inverses (with given rank as well as with prescribed range and null space) as particular cases of the full-rank representation of outer inverses. As a consequence, two applications of the successive matrix squaring (SMS) algorithm from [P.S. Stanimirovi?, D.S. Cvetkovi?-Ili?, Successive matrix squaring algorithm for computing outer inverses, Appl. Math. Comput. 203 (2008) 19-29] are defined using the full-rank representations of {2, 4} and {2, 3}-inverses. The first application is used to approximate {2, 4}-inverses. The second application, after appropriate modifications of the SMS iterative procedure, computes {2, 3}-inverses of a given matrix. Presented numerical examples clarify the purpose of the introduced methods.     

Scaling transformations for {0, 1}-valued sequences

Summary Let r2 be an integer and let : {0, 1} ^r {0,1} be a function. Let T be the transformation on ={0, 1}zgiven by (T)(i)=((ri), (ri +1), ..., (ri+r–1)) for all iZ. For P in the class of strongly-mixing shiftinvariant measures on , we investigate when P is invariant with respect to T and when T ⁿP converges. For example if r is odd and ( ₀,..., r–1)=1 iff >1/2r, the invariant measures are the Bernoulli measures with means 0, 1/2 or 1 and T ⁿP must converge to one of these three measures. Other choices of can give more complicated behaviour.Research supported in part by the National and Engineering Research Council of Canada.