@TechReport{Gagola90,
  authorkey =    "GagolaJ MS",
  author =       "Jr. Gagola and Stephen M.",
  title =        "Lifting Properties of Some Universal Orthogonal
                 Representations",
  institution =  "Kent State University",
  number =       "MA-9000-04",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "If $G\rightarrow O(V)$ is an orthogonal representation
                 of the group $G$, then a double cover of $G$ is
                 determined from the natural cover of the orthogonal
                 group: $\{\pm 1\}\rightarrow{\rm pin}(V)\rightarrow
                 O(V)$. If $\chi$ is the character of $G$ afforded by
                 $V$, let $h(\chi )$ denote the element in the Schur
                 multiplier of $G$ determined by the double cover of
                 $G$. Relations satisfied by this invariant are
                 determined by making use of some properties of the
                 standard embeddings of $GL(V)$ and $PGL(V)$ in
                 $O(V\oplus\hat{V})$ and $O(V\otimes \hat{V})$
                 respectively. In particular, $h(\chi
                 +\overline{\chi})=h(\chi\overline{\chi})=1$ for any
                 character $\chi$.",
  note =         "(Journal of Pure and Applied Algebra 61 (1989)
                 147-164.)",
}

@TechReport{Fridy90,
  authorkey =    "FridyJA MillerHI",
  author =       "J. A. Fridy and H. I. Miller",
  title =        "A Matrix Characterization of Statistical Convergence",
  institution =  "Kent State University",
  number =       "MA-9001-06",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "The sequence $x$ is {\em statistically convergent} to
                 $L$ if for each $\epsilon >0$, \[ \lim_{n} n^{-1}\{{\em
                 the~number~of}~k\leq n: |x_{k}-L| \leq \epsilon\}=0. \]
                 It is known that this method of summability cannot be
                 included by any matrix method, but for bounded
                 sequences it is included by the Ces\'{a}ro matrix
                 method $C_{1}$. In this paper these results are
                 extended by comparing statistical convergence with the
                 intersection of a collection $\cal T$ of summability
                 matrices, each of which is somewhat like $C_{1}$. It is
                 shown that a bounded sequence is statistically
                 convergent if and only if it is summable by every
                 matrix in $\cal T$. On the other hand, no countable
                 collection of matrices can include statistical
                 convergence for unbounded sequences. Also, the class
                 $\cal T$ is studies to determine which classical
                 summability matrices belong to $\cal T$.",
}

@TechReport{Shih90,
  authorkey =    "ShihCM",
  author =       "Chih-Ming Shih",
  title =        "A Parallel Routing Algorithm for Solving Conflict in a
                 Multistage Cube Network Architecture",
  institution =  "Kent State University",
  number =       "CS-9001-08",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "Multistage cube networks are blocking networks.
                 Multiprocessor systems based on this type of network
                 may have routing conflicts. We solve conflicts by
                 treating the multistage cube network as a cascade
                 consisting of the multistage cube network and an
                 inverse multistage cube network, and by inventing a
                 parallel routing algorithm based on the cascading
                 networks. Since the multibank shared memory is between
                 the cascading networks two passes are required to
                 communicate between two ends of the cascading networks
                 using the shared memory. Destination tags computed by
                 the algorithm will apply to both ends, source
                 processors and destination processors. In the first
                 pass, source processors route their data through one
                 side of the cascading networks and write to the shared
                 memory. In the second pass, destination processors
                 route themselves through the other side of the networks
                 and retrieve the data from the shared memory. The
                 complexity of computing destination tags ends up with
                 $.72(\log^{2}N+\log N)$ times the complexity of the
                 sort.",
  note =         "(International Conference on Parallel Processing
                 1990.)",
}

@TechReport{Cavaretta90,
  authorkey =    "CavarettaAS DahmenW MicchelliCA",
  author =       "Alfred S. Cavaretta and Wolfgang Dahmen and Charles A.
                 Micchelli",
  title =        "Stationary Subdivision",
  institution =  "Kent State University",
  number =       "MA-9000-15",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "Subdivision methods in computer graphics constitute a
                 large class of recursive schemes for computing curves
                 and surfaces. They seem to have their origin in the
                 geometric problem of smoothing corners of a given
                 polyhedral surface; in fact these methods are sometimes
                 referred to as ``wood carver'' algorithms because the
                 repeated smoothing operations are analogous to
                 sculpting wood. The algorithms begin with some initial
                 set of discrete data, called the control points, which
                 one can visualize as the vertices of a given polyhedral
                 surface. A particular algorithm is determined by a few
                 simple linear rules which, used repeatedly,
                 successively generate 'denser' data sets from the
                 initial set of control points. If the rules are well
                 chosen, these ever more dense data sets will approach
                 some continuous curve or surface in the limit. At any
                 particular stage, the limiting surface can be
                 approximated by a polyhedral surface which interpolates
                 the newly generated data. The whole process is
                 recursively elementary and computational stable;
                 therein lies the appeal of these algorithms. They offer
                 efficient ways of displaying curves and surfaces as
                 well as methods to design particular shapes, to
                 determine intersections of surfaces, to obtain level
                 curves and surfaces, and so on. Despite the simplicity
                 of the algorithms themselves, the analysis of the
                 limiting curve or surface associated with any given
                 subdivision algorithm seems to be formidable. There
                 are, however, remarkable exceptions to this. For
                 instance, by cutting the consecutive edges of a planar
                 polygon in the ratio 1/4: 1/2: 1/4 and then repeating
                 this surgery on the resulting polygon, we obtain a
                 sequence of polygons which converges in the limit to a
                 continuously differentiable curve consisting of
                 quadratic arcs. This is a special case of methods first
                 proposed by de Rham. It was later popularized by
                 Chaikin because of its rapid convergence, and the
                 piecewise quadrated nature of the limiting curve was
                 rediscovered by Riesenfeld. The monograph represents a
                 detailed study of what we call stationary subdivision
                 schemes. We will consider essentially two fundamental
                 questions: \begin{enumerate} \item What are necessary
                 and/or sufficient conditions on the operator $S$ for
                 convergence of the associated subdivision algorithm?
                 \item What can be said about the structure of the
                 limiting functions when the algorithm is convergent?
                 \end{enumerate} Each chapter concerns one of the many
                 different aspects and perspectives we have pursued
                 while addressing these two questions. Throughout our
                 study, we emphasize the role of the (unique) refinable
                 function associated with a given convergent subdivision
                 scheme. \end{tabular}",
}

@TechReport{Fellah90,
  authorkey =    "FellahA JurgensenH YuS",
  author =       "A. Fellah and H. J{\"u}rgensen and S. Yu",
  title =        "Alternating Finite Automata",
  institution =  "Kent State University",
  number =       "CS-9005-16",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "Alternation is a natural generalization of
                 nondeterminism. The model of alternating finite
                 automata was first introduced and studied by Chandra,
                 Kozen, and Stockmeyer. Although alternating finite
                 automata are no more powerful than deterministic finite
                 automata with respect to language recognition, special
                 features of alternating finite automata may provide new
                 approaches and techniques for solving theoretical and
                 practical problems concerning regular languages. In
                 this paper we present direct constructions for the
                 usual language theoretic operations in terms of
                 alternating finite automata. Moreover, we discuss
                 minimization and direct transformations between
                 alternating, non-deterministic, and deterministic
                 finite automata.",
  note =         "(International Conference on Computing, and
                 Information (ICCI) 1990.)",
}

@TechReport{Shah90,
  authorkey =    "ShahMJ",
  author =       "Mihr J. Shah",
  title =        "Hankel Transform Type Integrals and Applications",
  institution =  "Kent State University",
  number =       "MA-9000-17",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "Let $J_{\mu}(x)$ and $Y_{\nu}(x)$ be Bessel functions
                 of the first and second kind, respectively. The Hankel
                 transform of order $\nu$ of function $f(x)$, if it
                 exists, is defined by: \[ H_{\nu}\{f(x);
                 \zeta\}=\int^{\infty}_{0}J_{\nu}(x\zeta)f(x)dx. \] It
                 arises axis-symmetric boundary value problems in
                 elasticity and potential theory. Sonine showed that for
                 \[ f(x)=x^{-\nu}J_{\nu}(ax)J_{\nu}(bx), \]
                 $H_{\nu}\{f(x): x\}=0$ whenever, the triangular
                 inequality \[ |a-b|<c<a+b \] is violated. Recently;
                 Askey, Koornwinder and Rahman evaluated the Hankel
                 transform $H_{\nu}\{g(x);x\}$ of a function \[
                 g(x)=x^{\nu}Y_{\nu}Y_{\nu}(ax)J_{\nu}(bx)~~{\rm and} \]
                 showed that $H_{\nu}\{g(x);c\}$ vanishes whenever the
                 triangular inequality is satisfied. In this manuscript
                 we generalize the two ideas and compute the Hankel
                 transforms of following functions: \[
                 h(x)=x^{-\rho}J_{\mu}(ax)J_{\lambda}(bx) \] and \[
                 k(x)=x^{\rho} Y_{\mu}(ax)J_{\lambda}(bx). \] We discuss
                 some special cases and applications of the results.",
}

@TechReport{Shah90a,
  authorkey =    "ShahMJ",
  author =       "Mihr J. Shah",
  title =        "Certain Integrals of Products of Ultraspherical
                 Functions.",
  institution =  "Kent State University",
  number =       "MA-9000-18",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "Let $C^{\lambda}_{n}(x)$ and $D^{\lambda}_{n}(x)$ be
                 ultraspherical polynomial and ultraspherical function
                 of the second kind, respectively. Recently, Rahman and
                 Shah showed that the series \[
                 \sum^{\infty}_{n=0}(n+\lambda)\left[
                 \frac{\Gamma(n+1)}{\Gamma(n+2\lambda)}\right]^{2}D^{\lambda}_{n}
                 (\cos\alpha)C^{\lambda}_{n} (\cos
                 \beta)C^{\lambda}_{n}(\cos \gamma)C^{\lambda}_{n} (\cos
                 \delta) \] vanishes if \[ \begin{array}{l} \cos(\beta
                 +\gamma)<\cos(\alpha\pm \delta)<\cos(\beta-\gamma)
                 ~~{\rm and~if}~~\cos(\alpha +\delta)<\cos(\beta
                 +\gamma), \\ \cos(\alpha-\delta) < \cos (\beta-\gamma).
                 \end{array} \] We find a dual to this result. We prove
                 that the integral \[
                 \int^{1}_{-1}D^{\lambda}_{k}(x)C^{\lambda}_{\ell}(x)
                 C^{\lambda}_{m}(x)C^{\lambda}_{n}(x)(1-x^{2})^{2\lambda
                 -1}dx \] vanishes if~~i)~~$k+\ell+m+n$ is even and
                 if~~ii)~~$|\ell+k |>m+n; \,\,|\ell -k|<|m-n|$. Also,
                 the integral is a multiple of terminating, 2-balanced
                 and very well-poised $~_{9}F_{8}$ generalized
                 hypergeometric series in the regions where it is
                 non-zero.",
}

@TechReport{Khan90,
  authorkey =    "KhanMK",
  author =       "Mohammad K. Khan",
  title =        "Some Tauberian Theorems for Absolute Summability",
  institution =  "Kent State University",
  number =       "MA-9004-19",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "A convolution summability method is presented which
                 generalizes the classical Euler, Borel and Taylor type
                 matrix methods. Its regularity in the ordinary and
                 $\ell-\ell$ setting is proved. It is also shown that
                 $\dis{\sum^{\infty}_{n=0}}|x_{n+1}-x_{n}|\sqrt{n}<\infty$
                 is a Tauberian condition in the $\ell-\ell$ setting.
                 This generalizes a theorem of Fridy and Roberts, ``Some
                 Tauberian Theorems for Euler and Borel Summability'',",
  note =         "(Internat. J. Math. Math. Sci. Vol. 3, no. 4(1990),
                 pp. 731-738.)",
}

@TechReport{Asthariri90,
  authorkey =    "AsthaririCR PotterJL",
  author =       "Chandra R. Asthariri and Jerry L. Potter",
  title =        "Using Massive Parallelism for Context Sensitive Rule
                 Compilation,",
  institution =  "Kent State University",
  number =       "CS-9006-22",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "This paper describes a ``{\bf Context Sensitive
                 Grammar (CSG)}'' model which allows the use of an
                 associative processor to compile production rules into
                 equivalent procedural rules thus increasing the
                 executional efficiency of a production system. A
                 compiled procedural rule is a sequence of {\bf ANDIFs}
                 and {\bf ANDFORs} which are special parallel constructs
                 in the ASP ({\bf AS}sociative {\bf P}rogramming)
                 language that preserve the production rule concept. The
                 massive parallelism of associative processors using the
                 CSLG model and the ANDIF and ANDFOR constructs allow
                 direct non-RETE matching resulting in an estimated
                 throughput of about 6000 rules per second [18] as
                 opposed to 70 [12] and 900 [15] rules for SIMD
                 approaches adhering to the sequential RETE algorithm
                 [7].",
}

@TechReport{Ansemil90,
  authorkey =    "AnsemilJM AronRM PonteS",
  author =       "J. M. Ansemil and R. M. Aron and S. Ponte",
  title =        "Embeddings of Spaces of Holomorphic Functions of
                 Bounded Type",
  institution =  "Kent State University",
  number =       "MA-9011-23",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "Let $U$ be an open subset of a complex locally convex
                 space $E,$ let $F$ be a closed subspace of $E,$ and let
                 $\Pi:E \rightarrow E/F$ be the canonical quotient
                 mapping. In this paper we study the induced mapping
                 $\Pi^{*},$ taking $f \in {\cal H}_{b}(\Pi(U))
                 \rightarrow f \circ \Pi \in {\cal H}_{b}(U),$ where
                 ${\cal H}_{b}(V)$ denotes the space of holomorphic
                 functions of bounded type on $V$. We prove that this
                 mapping is an embedding when $E$ is a
                 Fr\'{e}chet-Schwartz space, and that it is not an
                 embedding for certain subspaces $F$ of every
                 Fr\'{e}chet-Montel, not Schwartz, space. We provide
                 several examples in the case where $E$ is a Banach
                 space to illustrate the sharpness of our results.",
}

@TechReport{Buoni90,
  authorkey =    "BuoniJ FarrellP RuttanA",
  author =       "John Buoni and Paul Farrell and Arden Ruttan",
  title =        "Algorithms for {LU} Decomposition on a Shared Memory
                 Multiprocessor",
  institution =  "Kent State University",
  number =       "CS-9010-28",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "In this paper we propose a number of algorithms for
                 the parallel LU decomposition of an upper Hessenberg
                 matrix on a shared memory multi-processor. We consider
                 the general case of $p$ processors, where $p$ is not
                 related to the size of the matrix problem. We show that
                 the LU decomposition of an $(m+1)$-banded Hessenberg
                 matrix can be achieved in $O(\frac{3nm^{2}}{p})$
                 operations, where $n$ is the dimension of the matrix
                 and $p$ is the number of processors. We show that for
                 the special case of tridiagonal matrices these
                 algorithms have a lower operation count than those in
                 the literature. In addition, they yield the best
                 existing algorithm for the solution of tridiagonal
                 systems of equations.",
}

@TechReport{Asthagiri90,
  authorkey =    "AsthagiriCR PotterJL",
  author =       "Chandra R. Asthagiri and Jerry L. Potter",
  title =        "Using Ambiguous Grammar for Register Optimization",
  institution =  "Kent State University",
  number =       "CS-9010-29",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "This paper presents an infix operator grammar ({\bf
                 IOG}) which is an extension of the operator infix
                 grammar ({\bf OIG}) [3]. IOG describes infix operator
                 expressions (IOE). While OIG is unambiguous and allows
                 strictly left (right) to right (left) order of
                 evaluation of left (right) associative operators , IOG
                 allows both left to right and right to left evaluation
                 of left associative operators by introducing ambiguity
                 in the grammar. It also shows that the ambiguity does
                 not violate algebraic axioms on operations and it helps
                 in the efficient utilization of the
                 temporaries/registers needed for the evaluation of an
                 expression. This IOG forms the theoretical basis for a
                 new parallel algorithm, associative optimizing operator
                 precedence algorithm (AOOPA) [2]. AOOPA has been
                 implemented in ASC [6,7,8] which parses and generates
                 code for IOEs.",
}

@TechReport{Asthagiri90a,
  authorkey =    "AsthagiriCR PotterJL",
  author =       "Chandra R. Asthagiri and Jerry L. Potter",
  title =        "An Optimizing Parallel Algorithm for Evaluating Infix
                 Operator Expressions on an Associative {SIMD}
                 Computer",
  institution =  "Kent State University",
  number =       "CS-9010-30",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "This paper presents an efficient associative
                 optimizing operator precedence algorithm ({\bf AOOPA})
                 to evaluate infix operator expressions. AOOPA does not
                 use the standard shift-reduce concept in parsing an
                 expression, but uses massive parallel associative
                 search instead. While conventional approaches separate
                 the optimization of the number of registers needed for
                 the evaluation of an expression from the parsing of the
                 expression, AOOPA incorporates optimization into the
                 parsing process itself. A conventional algorithm takes
                 3 passes and 4n steps in compiling an expression of n
                 operators, AOOPA takes only one pass and $\lfloor 5n/2
                 \rfloor$ steps.",
}

@TechReport{Farrell90,
  authorkey =    "FarrellPA",
  author =       "Paul A. Farrell",
  title =        "{LATEX} at Kent - {TEX}niques and Tools,",
  institution =  "Kent State University",
  number =       "CS-9003-11",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "LATEX runs on a variety of computers at many different
                 sites. This document tells you how to use LATEX on the
                 machines in the Department of Mathematics and Computer
                 Science at Kent. The instructions are specifically for
                 LATEX on the Encore Multimax but similar instructions
                 are true for LATEX on the Vax 780, and the
                 Hewlett-Packard and Sun workstations although
                 occasionally the versions implemented on the different
                 types of machine may be different. This document is not
                 about LATEX itself, which is described by the
                 manual---LATEX: {\em A Document Preparation System},
                 published by Addison-Wesley, available at fine book
                 stores everywhere. Similar, instructions also hold for
                 running TEX, AMS-TEX, and BIBTEX which are also
                 installed on some or all of the machines above. None of
                 the TEX related programs are implemented on the Sequent
                 Balance or Warp machines, which are reserved for
                 research purposes. Table 1 give a list of TEX related
                 programs and the machines they are available on. If you
                 have a question that you can't answer by reading the
                 manual and this document, ask Paul Farrell. He should
                 also be informed of any possible LATEX bugs or
                 undocumented anomalies.",
}

@TechReport{Farrell90a,
  authorkey =    "FarrellP",
  author =       "Paul Farrell",
  title =        "{TEX}niques and Tools s(Volume 1) - Basic {LATEX} at
                 Kent",
  institution =  "Kent State University",
  number =       "CS-9009-25",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "LATEX runs on a variety of computes at many different
                 sites. This document tells ;you how to use LATEX on the
                 machines in the Department of Mathematics and Computer
                 Science at Kent. The instructions are specifically for
                 LATEX on the Encore Multimax but similar instructions
                 are true for LATEX on the Vax 780, and the
                 Hewlett-Packard and Sun workstations although
                 occasionally the versions implemented on the different
                 types of machine may be different. This document is not
                 about LATEX itself, which is described by the
                 manual---LATEX: {\em A Document Preparation System},
                 published by Addison-Wesley, available at fine book
                 stores everywhere. Similar, instructions also hold for
                 running TEX, AMS-TEX, and BIBTEX which are also
                 installed on some or all of the machines above. None of
                 the TEX related programs are implemented on the Sequent
                 Balance or Warp machines, which are reserved for
                 research purposes. Table 1 gives a slightly dated list
                 of TEX related programs and the machines they are
                 available on.",
}

@TechReport{Buoni91,
  authorkey =    "BuoniJ FarrellPA RuttanA",
  author =       "John Buoni and Paul A. Farrell and Arden Ruttan",
  title =        "Algorithms for {LU} Decomposition on a Shared Memory
                 Multiprocessor",
  institution =  "Kent State University",
  number =       "CS-9101-03",
  year =         "1991",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "In this paper we propose a number of algorithms for
                 the parallel LU decomposition of an upper Hessenberg
                 matrix on a shared memory multi-processor. We consider
                 the general case of $p$ processors, where $p$ is not
                 related to the size of the matrix problem. We show that
                 the LU decomposition of an $(M+1)$-banded Hessenberg
                 matrix can be achieved in $O(\frac{3nm^{2}}{p})$
                 operations, where $n$ is the dimension of the matrix
                 and $p$ is the number of processors. We show that for
                 the special case of tridiagonal matrices these
                 algorithms have a lower operation count than those in
                 the literature. In addition, they yield the best
                 existing algorithm for the solution of tridiagonal
                 systems of equations.",
}

@TechReport{Spaces91,
  authorkey =    "SpacesB AronRM LohmanRH SuarezA",
  author =       "Banach Spaces and Richard M. Aron and Robert H. Lohman
                 and Antonio Su\'{a}rez",
  title =        "Rotundity, the {C}.{S}.{R}.{P}., and the
                 $\lambda$-Property in",
  institution =  "Kent State University",
  number =       "MA-9101-04",
  year =         "1991",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "Two open questions stemming from the
                 $\lambda$-property in Banach spaces are solved. The
                 following are shown to be equivalent in a Banach space
                 $X:(a)X$ has the $\lambda$-property; (b) every vector
                 in the closed unit ball of $X$ is expressible as a
                 convex series of extreme points of the unit ball of
                 $X$. Also, by exhibiting a class of nonrotund Orlicz
                 spaces for which the $\lambda$-function is identically
                 1 on the unit spheres, we answer negatively the
                 question of whether the $\lambda$-function
                 characterizes rotund Banach spaces.",
  note =         "(American Mathematical Society, Vol. 111, No. 1,
                 January 1991.)",
}

@TechReport{Peyravi91,
  authorkey =    "PeyraviH",
  author =       "Hassan Peyravi",
  title =        "On Distributed Adaptive Routing Schemes for
                 Circuit-Switched Networks",
  institution =  "Kent State University",
  number =       "CS-9102-08",
  year =         "1991",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "The objective of this research is to investigate the
                 performance of distributed non-hierarchical
                 circuit-switch networks for different adaptive routing
                 algorithms. We focus our attention on those routing
                 schemes which greatly influence the reconfiguration
                 process ({\em self-healing)} and improve the blocking
                 probability when the congestion level on a link exceeds
                 a threshold, a switching site fails, or a transmission
                 link gets disconnected. The adaptive routing
                 procedures, in this study, are virtually capable of
                 identifying the failures, isolating them, and
                 reconfiguring a new topology for the network. In this
                 study, we consider the elimination of the assumption of
                 independent traffic distributions on various links, in
                 the presence of redirected traffic. Consequently, we
                 confirm the behavior of these algorithms using
                 simulation techniques. Among different routing schemes
                 simulated in this study, the {\em Least Average Number
                 of Hops}, (LANG) algorithm is most attractive and leads
                 to a better performance. The simulated network is a
                 fault tolerant network and the proposed distributed
                 adaptive routing algorithms are capable of
                 self-reconfiguration during a critical period and when
                 an unexpected event happens.",
}

@TechReport{Farrell91,
  authorkey =    "FarrellP KhanK PeyraviH",
  author =       "Paul Farrell and Kazim Khan and Hassan Peyravi",
  title =        "On Blocking Distribution of Nonhierarchical Networks",
  institution =  "Kent State University",
  number =       "CS-9105-10",
  year =         "1991",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "It is shown in [19] that different types of routing
                 schemes for circuit-switched traffic in a
                 non-hierarchical network with uniform traffic intensity
                 lead to different performance trade-offs. It is also
                 assumed in [19] that the network is symmetric and the
                 redirected traffic on the links is Poisson. It is,
                 however, not clear how the trade-offs would be effected
                 by using different traffic intensity on links, i.e., if
                 nodes generate calls with different rates. This paper
                 analyzes the traffic behavior without assuming that the
                 redirected traffic is necessarily Poisson or that the
                 network is symmetric. We examine a statistical model to
                 obtain the {\em exact limiting distribution} and the
                 {\em end-to-end blocking} probability of a single link,
                 a group of links and the entire network.",
}

@TechReport{Frykman91,
  authorkey =    "FrykmanJ",
  author =       "James Frykman",
  title =        "$2^{m}3^{n}$ is Dense in ${R}^{+}$ an Application of
                 Elementary Real Analysis",
  number =       "MA-9106-10",
  year =         "1991",
}

@TechReport{Peyravi91a,
  authorkey =    "PeyraviH",
  author =       "Hassan Peyravi",
  title =        "Reliability of a Class of Real-time Adaptive Routing
                 Algorithms for Nonhierarchical Networks",
  institution =  "Kent State University",
  number =       "CS-9108-14",
  year =         "1991",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "The {\em end-to-end blocking} probability has been
                 used as a performance criterion to quantify the network
                 reliability. In this paper, we examine different
                 approaches to compute the reliability of a distributed
                 network system. First, we discuss an analytical
                 approach and a way to compute the end-to-end blocking
                 probability using Markov process. Second, we compare
                 the performance of a class of simple, yet adaptive
                 routing algorithms which improve the network
                 reliability. We focus our attention to those routing
                 schemes which greatly influence the reconfiguration
                 ({\em self-healing}) process of the network in
                 real-time. These algorithms are attractive since they
                 are simple and can be implemented in hardware with
                 minimum overhead. The adaptive routing algorithms, in
                 this study, are virtually capable of identifying and
                 isolating network failures, and reconfiguring a new
                 topology for the network. We confirm the behavior of
                 these algorithms using a simulation technique.",
}

@TechReport{Peyravi91b,
  authorkey =    "PeyraviH",
  author =       "Hassan Peyravi",
  title =        "On Performance of a Class of Real-Time Adaptive
                 Routing Algorithms for Circuit Switched Networks",
  institution =  "Kent State University",
  number =       "CS-9108-15",
  year =         "1991",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "The objective of this research is to investigate the
                 performance of a class of real-time adaptive routing
                 algorithms for a distributed non-hierarchical
                 circuit-switched network. We focus our attention to
                 those routing schemes which greatly influence the
                 reconfiguration ({\em self-healing}) process of the
                 network in real-time. These algorithms are attractive
                 due to their simplicity and their hardware
                 implementation. They react to congestions and network
                 failures, and improve the blocking probability. The
                 adaptive routing algorithms, in this study, are
                 virtually capable of identifying and isolating network
                 failures, and reconfiguring a new topology for the
                 network. We confirm the behavior of these algorithms
                 using a simulation technique.",
}

@TechReport{Peyravi91c,
  authorkey =    "PeyraviH",
  author =       "Hassan Peyravi",
  title =        "Limiting Distribution of the Blocking Probability for
                 Circuit-switched Networks",
  institution =  "Kent State University",
  number =       "CS-9108-16",
  year =         "1991",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "The {\em end-end-end} blocking probability has been
                 used as a performance criterion to quantify network
                 reliability. Yum and Schwartz [19] have shown that
                 different types of routing schemes for circuit-switched
                 traffic in a non-hierarchical network, with uniform
                 traffic intensity, lead to different performance
                 trade-offs. It is also assumed that the network is
                 symmetric and the redirected traffic on the links is
                 Poisson. It is, however, not clear how the trade-offs
                 would be effected by using different traffic
                 intensities on the links, i.e., if the nodes generate
                 calls with different rates. This paper analyzes the
                 traffic behavior without assuming that the redirected
                 traffic is necessarily Poisson or that the network is
                 symmetric. We examine a statistical model to obtain the
                 {\em exact limiting distribution} and the {\em
                 end-to-end blocking} probability of a single specified
                 link, a group of links and the entire network.",
}

@TechReport{Liszka92,
  authorkey =    "LiszkaKJ BatcherKE",
  author =       "Kathy J. Liszka and Kenneth E. Batcher",
  title =        "A Generalized Bitonic Merging Network",
  institution =  "Kent State University",
  number =       "CS-9201-01",
  year =         "1992",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "The bitonic merging network will sort N-keys in
                 O$(\log^{2}N)$ time with O$(N\log^{2}N)$ comparators.
                 Improvements to the sorter enable the network to sort
                 pq-keys, a composite number. Two methods are presented
                 for completely generalizing the bitonic merger for an
                 arbitrary number of keys. They are based on different
                 decompositions of the network. The add-merge bitonic
                 sorter decomposes $N$ keys into a p-key bitonic sorter,
                 a q-key bitonic sorter and a merge of p keys with q
                 keys where $N=p+q$. The odd-merge bitonic sorter also
                 uses a p and q-key bitonic sorter, but the
                 decomposition is different and the final merge stage is
                 an interleave. The two decompositions are compared in
                 terms of efficiency and practicality.",
}

@TechReport{Scott92,
  authorkey =    "ScottSL BakerJW",
  author =       "Stephen L. Scott and Johnnie W. Baker",
  title =        "Embedding the Hypercube into the 3-Dimension Mesh",
  institution =  "Kent State University",
  number =       "CS-9203-05",
  year =         "1992",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "A constant time and space algorithm for embedding the
                 hypercube architecture into the {\rm 3-dimension} mesh
                 architecture is presented. This embedding enables the
                 cube$_i$ data exchange operation to be performed on the
                 embedded hypercube machine in a constant time when the
                 interprocessor communication function cube$_i$ is
                 defined on the embedded hypercube's PEs as
                 cube$_i$${(b_{n-1}\cdots b_{i+1}b_ib_{i-1}\cdots
                 b_1b_0)=b_{n-1}\cdots
                 b_{i+1}\overline{b}_ib_{i-1}\cdots b_1b_0}$, where
                 $\overline{b}_i$ is the binary complement of ${b}_i$.",
}

@TechReport{Johnson90,
  authorkey =    "JohnsonM NicholsonV TsaiCC",
  author =       "Mark Johnson and Victor Nicholson and Chun-Che Tsai",
  title =        "Sterographs: An Extension of Chemical Graphs that can
                 Distinguish Stereochemical and Conformational
                 Structures",
  institution =  "Kent State University",
  number =       "CS-9000-01",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "Stereographs are presented as new vehicles for
                 representing stereochemical and conformational
                 information. They extend the concept of a chemical
                 graph by adding ``tetrads'', each qualitatively
                 characterizing the three---dimensional arrangement of
                 four atoms into a clinicity and handedness component.
                 The characterization is sufficiently precise to
                 distinguish synperiplanar, synclinal, anticlinal and
                 antiperiplanar arrangements. Collectively, the tetrads
                 constitute the embedding graph which presents new
                 possibilities in displaying the stereochemical and
                 conformational features of a molecule. Potential
                 applications of stereographs in the areas of structural
                 representation, molecular symmetry analysis,
                 stereo-specific substructure searching, chemical
                 transformation analysis, and molecular proximity
                 analysis are briefly discussed.",
  note =         "(presented by MJ at the 1989 PacifiChem Mathematical
                 Chemistry Minisymposium.)",
}

@TechReport{Cavaretta90,
  authorkey =    "CavarettaAS DahmenW MicchelliCA",
  author =       "A. S. Cavaretta and W. Dahmen and C. A. Micchelli",
  title =        "The Volume of Restricted Moment Spaces",
  institution =  "Kent State University",
  number =       "MA-9000-02",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "In this paper we compute the volume of restricted
                 moment spaces under very general conditions on the
                 functions generating the moment space.",
  note =         "(Rendiconti Del Circolo Matematico Di Palermo, Serie
                 II, Tomo XXXVIII (1989), pp. 419-429.)",
}

@TechReport{Garner90,
  authorkey =    "GarnerJ DasSG MinBI WoodwardC BenedekR",
  author =       "J. Garner and S. G. Das and B. I. Min and C. Woodward
                 and R. Benedek",
  title =        "Comparison of Optimization Methods for
                 Electronic-Structure Calculations",
  institution =  "Kent State University",
  number =       "MA-9000-03",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "The performance of several local-optimization methods
                 for calculating electronic structure is compared. The
                 fictitious first-order equation of motion proposed by
                 Williams and Soler is integrated numerically by three
                 procedures: simple finite-difference integration,
                 approximate analytical integration (the Williams-Soler
                 algorithm), and the Born perturbation series. These
                 techniques are applied to a model problem for which
                 exact solutions are known, the Mathieu equation. The
                 Williams-Soler algorithm and the second Born
                 approximation converge equally rapidly, but the former
                 involves considerably less computational effort and
                 gives a more accurate converged solution. Application
                 of the method of conjugate gradients to the Mathieu
                 equation is discussed.",
  note =         "(The American Physical Society, Vol. 39, No. 17, p.
                 39, June 15, 1989.)",
}

@TechReport{Aron90,
  authorkey =    "AronRM ColeBJ GamelinTW",
  author =       "R. M. Aron and B. J. Cole and T. W. Gamelin",
  title =        "Spectra of Algebras of Analytic Functions on a Banach
                 Space",
  institution =  "Kent State University",
  number =       "MA-9000-05",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "Fix a Banach space \$X\$, with open unit ball \$B\$.
                 We are interested in studying the uniform algebra
                 \$H\^{}$\lbrace$$\backslash$infty$\rbrace$(B)\$ of
                 bounded analytic functions on \$B\$, and its spectrum
                 \$M=M(B)\$ consisting of the nonzero complex-valued
                 homomorphisms of
                 \$H\^{}$\lbrace$$\backslash$infty$\rbrace$(B)\$.",
}

@TechReport{Peyravi90,
  authorkey =    "PeyraviH KhanMK FarrellP",
  author =       "Hassan Peyravi and Mohammad K. Khan and Paul Farrell",
  title =        "Limiting Distributions of Some Circuit Switched
                 Nonhierarchical Networks",
  institution =  "Kent State University",
  number =       "CS-9002-07",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "Among the two major types of networks, packet switched
                 and circuit switched networks, we discuss the second
                 type of networks which are often used in telephone
                 traffic. By providing simple examples we show how one
                 can obtain the exact limiting distribution of the
                 network. This can be used to obtain the exact end to
                 end blocking probability of an incoming call along with
                 the system utilization statistics. We provide some
                 results for three node symmetric and nonsymmetric
                 networks. The generalizations to higher numbers of
                 nodes is analogous. Some examples of the system
                 utilization and blocking probability are provided under
                 various input and holding time parameters.",
}

@TechReport{Smith90,
  authorkey =    "SmithFA",
  author =       "F. A. Smith",
  title =        "On Certain Extensions of Partial Orders on Idempotent
                 and Abelian Semigroups,",
  institution =  "Kent State University",
  number =       "MA-9003-10",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "Partial orders on semigroups are examined. A necessary
                 and sufficient condition for extending a partial order
                 to a full order is given in both the idempotent and
                 cancellative case.
                 $\backslash$nd$\lbrace$tabular$\rbrace$",
}

@TechReport{Lu90,
  authorkey =    "LuCC DunhamJG",
  author =       "Cheng-Chang Lu and James George Dunham",
  title =        "A Universal Model Based on Minimax Average
                 Divergence",
  institution =  "Kent State University",
  number =       "CS-9001-13",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "Given a set of training samples, the commonly used
                 approach to determine a universal model is accomplished
                 by averaging the statistics over all training samples.
                 In this correspondence, we suggest to use average
                 divergence as a measurement for the effectiveness of a
                 universal model and propose a minimax universal model
                 which minimizes the maximum average divergence among
                 all training samples. Efficient searching algorithms
                 are developed and experimental results are also
                 presented.",
  note =         "(IEEE International Symposium on Information Theory,
                 San Diego.)",
}

@TechReport{Frykman90,
  authorkey =    "FrykmanJ",
  author =       "James Frykman",
  title =        "Practical Implementation of {RSA} Public Key
                 Encryption",
  institution =  "Kent State University",
  number =       "CS-9005-14",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "Despite its apparent advantages the public key
                 encryption system proposed by Rivest, Shamir, and
                 Adleman has not been adopted for widespread use. The
                 principal reason is that if a computer program is used
                 to do the computations for the RSA system then the
                 encryption/decryption proceeds at a rate of a few
                 hundred bits per second, far slower than the tens of
                 thousands of bits per second needed for practical
                 applications. This paper develops an algorithm as a
                 circuit on a single chip with some rough calculations
                 to show the feasibility of doing encryption/decryption
                 for practical RSA public key systems at rates
                 sufficiently high to be adopted for general use.",
}

@TechReport{Bansal??,
  authorkey =    "BansalAK PotterJL",
  author =       "Arvind K. Bansal and Jerry L. Potter",
  title =        "Exploiting Data Parallelism for Efficient Execution of
                 Logic Programs with Large Knowledge Base",
  institution =  "Kent State University",
  number =       "CS-9005-20",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "A model is presented which is designed to exploit data
                 parallelism present in associative computers for the
                 efficient execution of logic program. Data parallel
                 model allows efficient handling of very large knowledge
                 bases. A scheme for logical data structure
                 representation incorporating direct interface between
                 lists and vectors is described which allows the
                 efficient integration of symbolic and numerical
                 computation on existing associative supercomputers. An
                 algorithm for data parallel goal reduction which is
                 almost independent of number of clauses is discussed.
                 The associative goal reduction scheme efficiently
                 prones non-unifiable clauses and performs data parallel
                 binding of variables with single occurrence. The
                 association property in the model effectively reduces
                 the cost of shallow backtracking, deep backtracking,
                 and garbage collection.",
}

@TechReport{Aron90a,
  authorkey =    "AronRM GlobevnikJ",
  author =       "Richard M. Aron and Josip Globevnik",
  title =        "Analytic Functions on \$c\_$\lbrace$0$\rbrace$\$",
  institution =  "Kent State University",
  number =       "MA-9001-24",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "Let \$$\backslash$cal F\$ be a space of continuous
                 complex valued functions on a subset of
                 \$c\_$\lbrace$0$\rbrace$\$ which contains the standard
                 unit vector basis \$$\backslash$$\lbrace$
                 e\_$\lbrace$n$\rbrace$$\backslash$$\rbrace$\$. Let
                 \$R:$\lbrace$$\backslash$cal F$\rbrace$
                 $\backslash$rightarrow
                 C\^{}$\lbrace$$\backslash$infty$\rbrace$\$ be the
                 restriction map, given by \$R(f)\~{} =
                 \~{}(f(e\_$\lbrace$1$\rbrace$),$\backslash$ldots,f(e\_$\lbrace$n$\rbrace$),$\backslash$ldots).\$
                 We characterize the ranges \$R($\lbrace$$\backslash$cal
                 F$\rbrace$)\$ for various ``nice'' spaces
                 \$$\lbrace$$\backslash$cal F$\rbrace$.\$ For example,
                 if \$$\lbrace$$\backslash$cal F$\rbrace$ =
                 $\lbrace$$\backslash$cal
                 P$\rbrace$(\^{}$\lbrace$n$\rbrace$E)\$, then
                 \$R($\lbrace$$\backslash$cal F$\rbrace$) =
                 $\backslash$ll\_$\lbrace$1$\rbrace$,\$ and if
                 \$$\lbrace$$\backslash$cal F$\rbrace$ =
                 A\^{}$\lbrace$$\backslash$infty$\rbrace$(B\_$\lbrace$c\_$\lbrace$0$\rbrace$$\rbrace$),\$
                 then \$R($\lbrace$$\backslash$cal F$\rbrace$) =
                 $\backslash$ll\_$\lbrace$$\backslash$infty$\rbrace$.\$",
}

@TechReport{Shih90,
  authorkey =    "ShihCJ BatcherKE",
  author =       "C. Jimmy Shih and Kenneth E. Batcher",
  title =        "Adding Multiple-Fault Tolerance to Generalized Cube
                 Networks",
  institution =  "Kent State University",
  number =       "CS-9012-31",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "Most generalized cube networks are limited to
                 single-fault tolerance in the criterion of permutation
                 connections. We are therefore proposing our 2 and
                 3-fault tolerance schemes. In this paper we first
                 present the proposed structure which picks many
                 appropriate connections and adds these connections to
                 an original network. As faults occur, the faulty
                 connections will be removed and remaining connections
                 will form a substitute network called physical network.
                 Since the physical network is no longer the same as the
                 original network, a physical-to-logical mapping
                 mechanism is required so that the physical network can
                 be used in the same way as before. We prove that the
                 mapping function is an isomorphism, therefore, the
                 reconfiguration process we use is correct. We then go
                 through the reconfiguration process with an example.",
}

@TechReport{Baker90,
  authorkey =    "BakerJW MillerAR",
  author =       "Johnnie W. Baker and Andres R. Miller",
  title =        "A Parallel Production System Extending {OPS5}",
  institution =  "Kent State University",
  number =       "CS-9010-32",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "This paper shows that by assuming a common model of
                 SIMD computing and extending the OPS5 production system
                 formalism to take advantage of the SIMD model, an
                 entirely new class of problems lends itself to solution
                 through the intuitive use of a Parallel Production
                 System language. The parallel Production System, OPS,
                 allows actions on the right hand sides of a rule to act
                 on classes of working memory elements rather than the
                 single element per action scheme used by OPS5. New
                 methods of conflict resolution have also been provided.
                 Parallel conflict resolution will resolve only the
                 ``potential conflicts.'' The initial conflict set is
                 formed traditionally, but a maximal set of rule
                 instantiations in the conflict set are executed.
                 Traditional OPS5 conflict resolution is used for any
                 rules in the conflict set with conflicting
                 instantiations, thus multiple directions of reasoning
                 may be followed concurrently. This facilitates the use
                 of production systems to implement asynchronous
                 concurrent systems using a SIMD computer.",
}

@TechReport{Batcher90,
  authorkey =    "BatcherKE",
  author =       "Kenneth E. Batcher",
  title =        "On Bitonic Sorting Networks",
  institution =  "Kent State University",
  number =       "CS-9001-34",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "A \$(kn)\$-key bitonic sorting network can be build
                 from \$n\$ \$k\$-key sorters and \$k\$ \$n\$-key
                 sorters; costs and delays of networks using different
                 values of \$n\$ are examined. For merging one need only
                 sort ascending-descending sequences of keys instead of
                 more general bitonic sequences; the construction of
                 networks to sort ascending-descending sequences of any
                 size is shown.",
  note =         "(1990 International Conference on Parallel
                 Processing.)",
}

@TechReport{Seville90,
  authorkey =    "Seville",
  author =       "1990). June Richard M. Aron (Seville",
  title =        "Spectra of Analytic Functions in Infinite Dimensions",
  institution =  "Kent State University",
  number =       "MA-9006-35",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "This talk describes part of the work by R.M. Aron, B.
                 Cole, and T. Gamelin, entitle ``Spectra of analytic
                 functions on a Banach space,'' which will appear in
                 [A-C-G].",
}

@TechReport{Farrell90,
  authorkey =    "FarrellPA",
  author =       "Paul A. Farrell",
  title =        "Accessing the Library Catalogue Over the Campus
                 Network at Kent State",
  institution =  "Kent State University",
  number =       "CS-9004-12",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "Do you find it irritating to walk to the library just
                 to find out if they have a book in the KSU collection.
                 Then read on. The library on-line catalogue can be
                 accessed directly from any place on the campus with
                 network access. This document outlines how this can be
                 done - in particular from Merrill Hall.",
}

@TechReport{Farrell90a,
  authorkey =    "FarrellP",
  author =       "Paul Farrell",
  title =        "National Internet-Accessible Library Catalogs
                 Databases",
  institution =  "Kent State University",
  number =       "CS-9010-26",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "For those of you interested in accessing library
                 catalogues at remote sites, outside the
                 Ohio-Pennsylvania region, the following is a list of
                 sites currently accessible via the internet. This
                 document outlines how this can be done---in particular
                 from the Mathematics and Computer Science department
                 machines in Merrill Hall. you DO NOT NEED to have an
                 account at the remote site to do this.",
}

@TechReport{Farrell90b,
  authorkey =    "FarrellP",
  author =       "Paul Farrell",
  title =        "Regional Internet-Accessible Library Catalogs
                 Databases",
  institution =  "Kent State University",
  number =       "CS-9010-27",
  year =         "1990",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "For those of you interested in accessing library
                 catalogues at remote sites, outside the
                 Ohio-Pennsylvania region, the following is a list of
                 sites currently accessible via the internet. This
                 document outlines how this can be done---in particular
                 from the Mathematics and Computer Science department
                 machines in Merrill Hall. you DO NOT NEED to have an
                 account at the remote site to do this.",
}

@TechReport{Shih91,
  authorkey =    "ShihCJ BatcherKE",
  author =       "C. Jimmy Shih and Kenneth E. Batcher",
  title =        "Multiple-Fault Tolerant Cube-Connected Cycles
                 Networks",
  institution =  "Kent State University",
  number =       "CS-9101-02",
  year =         "1991",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "Most fault-tolerance schemes for Cube-Connected Cycles
                 (CCC) networks are limited to single-fault tolerance in
                 the criterion of permutation connections. We are
                 therefore proposing our 2 and 3-fault tolerance
                 schemes. In this paper we first present the proposed
                 structure which picks many appropriate connections and
                 adds these connections and their associated PEs to an
                 original network. As faults occur, the faulty
                 connections and PEs will be removed and remaining
                 connections will form a substitute network called
                 physical network. Since the physical network is no
                 longer the same as the original network, a
                 physical-to-logical mapping mechanism is required so
                 that the physical network can be used in the same way
                 as before. We prove that the mapping function is an
                 isomorphism, therefore, the reconfiguration process we
                 use is correct. We then demonstrate the reconfiguration
                 process with an example.",
}

@TechReport{Ansemil91,
  authorkey =    "AnsemilJM AronR PonteS",
  author =       "J. M. Ansemil and R.m. Aron and S. Ponte",
  title =        "Spaces of Holomorphic Functions and Germs on
                 Quotients",
  institution =  "Kent State University",
  number =       "MA-9104-07",
  year =         "1991",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "Let \$E\$ be a complex locally convex space, \$F\$ a
                 closed subspace of \$E\$ and let \$$\backslash$Pi\$ be
                 the canonical quotient mapping from \$E\$ onto \$E/F\$.
                 If \$U\$ is an open subset of \$E\$ and \$K\$ is a
                 compact subset of \$E,$\backslash$,$\backslash$Pi\$
                 induces a linear injective mapping
                 \$$\backslash$Pi$\backslash$ast\$ from
                 \$$\lbrace$$\backslash$cal
                 H$\rbrace$[$\backslash$Pi(U)]\$ (resp.
                 \$$\lbrace$$\backslash$cal
                 H$\rbrace$[$\backslash$Pi(K)]\$ into
                 \$$\lbrace$$\backslash$cal H$\rbrace$(U)\$ (resp.
                 \$$\lbrace$$\backslash$cal H$\rbrace$(K)\$), where
                 \$$\lbrace$$\backslash$cal H$\rbrace$(V)\$ (resp.
                 \$$\lbrace$$\backslash$cal H$\rbrace$(L)\$ stands for
                 the space of holomorphic functions (resp. germs) on the
                 open subset \$V\$ (resp. the compact subset \$L\$) of a
                 complex locally convex space. This mapping
                 \$$\backslash$Pi$\backslash$ast\$ is defined by
                 \$$\backslash$Pi$\backslash$ast(g)=g$\backslash$circ
                 $\backslash$Pi;$\backslash$,$\backslash$,
                 g$\backslash$in $\lbrace$$\backslash$cal
                 H$\rbrace$[$\backslash$Pi(U)]\$ (resp.
                 \$$\lbrace$$\backslash$cal
                 H$\rbrace$[$\backslash$Pi(K)]\$). When we consider the
                 natural topologies on the spaces of holomorphic
                 functions and holomorphic germs,
                 \$$\backslash$Pi$\backslash$ast\$ is always continuous.
                 In some cases \$$\backslash$Pi$\backslash$ast\$ is also
                 a topological isomorphism onto its image (embedding)
                 but this is not always true. The aim of this paper is
                 to give a survey of the results concerning when
                 \$$\backslash$Pi$\backslash$ast\$ is, and is not, an
                 embedding. We also include some new results with their
                 proofs.",
}

@TechReport{Asthagiri91,
  authorkey =    "AsthagiriCR PotterJL",
  author =       "Chandra R. Asthagiri and Jerry L. Potter",
  title =        "Parallel Lexing on an Associative {SIMD} Parallel
                 Computer",
  institution =  "Kent State University",
  number =       "CS-9106-11",
  year =         "1991",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "This paper presents a near constant time associative
                 parallel lexing (APL) algorithm. The best time
                 complexity thus far claimed on parallel lexing is
                 \$$\backslash$bigcirc\$(log \$n\$), where \$n\$ is the
                 number of input characters. The time complexity
                 reported is for the parallel prefix lexing (PPL)
                 algorithm. The linear state recording step in the PPL
                 algorithm, which needs to be done only once for each
                 grammar, has been ignored in claiming the long \$n\$
                 time complexity for the PPL algorithm. Furthermore, the
                 PPL algorithm does not consider recording line numbers
                 for the tokens and distinguishing identifier tokens as
                 keyword or user-identifier. The APL algorithm performs
                 all of the functions missed in the prefix algorithm.
                 Thus, without considering the efforts spent on these
                 functions, the APL algorithm takes constant time since
                 every step depends on the field width of the tokens,
                 not on the length of the input. Generalizing and
                 including these extra functions, the APL algorithm
                 takes near constant time.
                 $\backslash$nd$\lbrace$tabular$\rbrace$",
}

@TechReport{Rothstein91,
  authorkey =    "RothsteinM ZassenhausH",
  author =       "M. Rothstein and H. Zassenhaus",
  title =        "Deterministic Analysis of Aleatoric Methods of
                 Polynomial Factorization Over Finite Fields",
  institution =  "Kent State University",
  number =       "CS-9108-12",
  year =         "1991",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "We analyze the problem of finding the roots of an
                 arbitrary polynomial over a finite field (equivalent to
                 factoring an arbitrary polynomial over the field) and
                 propose a deterministic approach which leads to a
                 combinatorial problem whose satisfactory solution would
                 yield a factoring method which is polynomial in the
                 logarithm of the size of the field and the degree of
                 the input polynomial.",
}

@TechReport{Johnson91,
  authorkey =    "JohnsonM Tsai NicholsonV",
  author =       "M. Johnson and Chun-che Tsai and Victor Nicholson",
  title =        "Chemical Stereographs: An Extension of Chemical Graphs
                 for Representing Stereochemical and Conformational
                 Structures",
  institution =  "Kent State University",
  number =       "MA-9108-13",
  year =         "1991",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "Chemical stereographs are presented as vehicles for
                 representing qualitative three-dimensional features of
                 molecules that put stereochemical and conformational
                 distinctions in a common graph-theoretic formalism.
                 They extend the concept of a chemical graph by adding
                 ``tetrads'', each qualitatively characterizing the
                 three-dimensional arrangement of four atoms with
                 respect to its clinicity and handedness components. The
                 characterization is sufficiently precise to distinguish
                 synperiplanar, synclinal (gauche), anticlinal and
                 antiperiplanar relationships between vicinal atoms of
                 various conformers. Collectively, the tegrads
                 constitute the embedding graph which presents new
                 possibilities in displaying the stereochemical and
                 conformational features of a molecule. A chemical graph
                 and one of its possible embedding graphs constitute a
                 chemical stereograph. Potential applications of
                 chemical stereographs in the areas of structural
                 representation, molecular symmetry analysis, and
                 stereo-specific substructure searching are discussed.",
}

@TechReport{Sridhar91,
  authorkey =    "SridharK",
  author =       "Kanneganti Sridhar",
  title =        "Congestion in Perfect Shuffle Networks",
  institution =  "Kent State University",
  number =       "CS-9108-17",
  year =         "1991",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "A perfect shuffle network interconnecting
                 \$2\^{}$\lbrace$n$\rbrace$\$ processors is bound to
                 have some inter processor links more heavily congested
                 than others. This paper investigates the congestion in
                 different size perfect shuffle networks.",
}

@TechReport{Aron91,
  authorkey =    "AronR ColeB GamelinT",
  author =       "R. Aron and B. Cole and T. Gamelin",
  title =        "Spectra of Analytic Functions in Infinite Dimensions",
  number =       "MA-9100-18",
  year =         "1991",
}

@TechReport{Aron91a,
  authorkey =    "AronRM ZalduendoI",
  author =       "R. M. Aron and I. Zalduendo",
  title =        "Polynomial Norms and Coefficients",
  number =       "MA-9100-19",
  year =         "1991",
}

@TechReport{Aron91b,
  authorkey =    "AronRM ColeBJ GamelinTW",
  author =       "R. M. Aron and B. J. Cole and T. W. Gamelin",
  title =        "Spectra of Algebras of Analytic Functions on a Banach
                 Space",
  number =       "MA-9100-21",
  year =         "1991",
}

@TechReport{diestel91,
  authorkey =    "diestelJ",
  author =       "Joe diestel",
  title =        "Uniform Integrability: An Introduction",
  number =       "MA-9110-22",
  year =         "1991",
}

@TechReport{Batcher92,
  authorkey =    "BatcherKE",
  author =       "Kenneth E. Batcher",
  title =        "Low-Cost Flexible Simulation with the Static Perfect
                 Shuffle Network",
  institution =  "Kent State University",
  number =       "CS-9202-02",
  year =         "1992",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "The static perfect shuffle network is based on Stone's
                 perfect shuffle network; the two-by-two switches are
                 removed and replaced by exchange links between
                 processors. The network could be used in a low-cost
                 flexible simulator for other networks such as
                 multistage cube networks and hypercubes.",
}

@TechReport{Bansal92,
  authorkey =    "BansalAK PotterJ",
  author =       "Arvind K. Bansal and Jerry Potter",
  title =        "Exploiting Data Parallelism for Large Knowledge
                 Bases",
  institution =  "Kent State University",
  number =       "CS-9202-03",
  year =         "1992",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "A model is presented which exploits data level massive
                 parallelism present in associative computers for the
                 efficient execution of logic programs with large
                 knowledge bases. The exploitation of data parallelism
                 in goal reduction efficiently prunes non-unifiable
                 clauses resulting into effective reduction of shallow
                 backtracking, and marks the potential bindings for the
                 variables with single occurrence in a manner which is
                 independent of the number of clauses. During deep
                 backtracking, bindings are released simultaneously
                 using associative search resulting in a significant
                 reduction in execution time overhead of backtracking
                 and garbage collection. A scheme for a logical data
                 structure representation incorporating direct interface
                 between lists and vectors is described. This allows the
                 efficient integration of symbolic computation and a
                 large class of vectorizable numerical computation on
                 associative supercomputers.",
}

@TechReport{Bansal92a,
  authorkey =    "BansalAK PotterJL PrasadLV",
  author =       "Arvind K. Bansal and Jerry L. Potter and Lokam V.
                 Prasad",
  title =        "An Associative Compilation Model to Exploit
                 Date-parallelism in Function-free Logic Programs",
  institution =  "Kent State University",
  number =       "CS-9203-04",
  year =         "1992",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "This paper describes a model for the compilation of
                 function free logic programs to exploit the
                 data-parallelism present in associative supercomputers.
                 The compilation process maps sets of clause-heads to
                 associations of vectors, and subgoals to low level code
                 analogous to WAM to achieve a synergistic efficiency.
                 The exploitation of data-parallelism on associative
                 supercomputers allows goal reduction with all unifiable
                 clauses almost independent of the number of clauses,
                 and minimizes shallow backtracking. The use of
                 associative searching facilitates deep backtracking,
                 intelligent backtracking, and the handling of bindings
                 of aliased variables with very low overhead. The
                 representation of subgoals as low level code reduces
                 the excessive data-transfer present in conventional
                 data-parallel models [2,3] by making use of an
                 associative analogue of global registers. The compiler
                 also uses compile-time data flow analysis to derive
                 groundness and width information. This information is
                 used for efficient code optimization, and allows
                 efficient data-matching.",
}

@TechReport{Masri92,
  authorkey =    "MasriI TongeA",
  author =       "Ibrahim Masri and Andrew Tonge",
  title =        "Norm Estimates for Random Multilinear Hankel Forms",
  institution =  "Kent State University",
  number =       "MA-9203-06",
  year =         "1992",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "We obtain sharp estimates for average norms of
                 multilinear Hankel forms on complex Hilbert space whose
                 entries are independent Rademacher functions.",
}

@TechReport{Lu92,
  authorkey =    "LuCC OmarN",
  author =       "Cheng-Chang Lu and Norhanim Omar",
  title =        "A Modified Short-Kernel Filter Pair for Perfect
                 Reconstruction of {HDTV} Signals",
  institution =  "Kent State University",
  number =       "CS-9204-07",
  year =         "1992",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "For video transmission between broadcasting stations
                 or studios very high quality and even distortion-free
                 pictures are usually desired. Recently, subband coding
                 of images has received considerable attentions and
                 become an effective tool for image compression. In
                 recent work by Irie and Kishimoto [1], short-kernel
                 filters have been successfully applied to compress HDTV
                 signals. Perfect reconstruction is obtained with very
                 low average entropy. This paper presents further
                 improvements on average entropy for perfect
                 reconstruction of HDTV signals with a modified
                 short-kernel filter pair. The simulation results show
                 that the proposed filter pair with interband prediction
                 can greatly reduce the entropy compared to that
                 obtained with other approaches.",
}

@TechReport{White92,
  authorkey =    "WhiteDL",
  author =       "Donald L. White",
  title =        "Brauer Trees of \$Sp(4,q)\$",
  institution =  "Kent State University",
  number =       "MA-9203-08",
  year =         "1992",
  url =          "mailto:sommers@mcs.kent.edu",
  abstract =     "The Brauer trees of the symplectic group \$Sp(4,q)\$
                 are computed in all non-defining characteristics. This
                 completes the determination of all modular characters
                 of the groups \$Sp(4,q)\$ in non-defining
                 characteristics.",
  note =         "(Comm.$\backslash$, Alg.$\backslash$, 20 (March 1992),
                 645-653.)",
}