Ernst Leiss

Followers

Following

Co-authors

Public Views

Interests

Uploads

Papers by Ernst Leiss

Sources of Disappointments

Comanche – A Compiler-Driven I/O Management System

World Academy of Science, Engineering and Technology, International Journal of Computer, Electrical, Automation, Control and Information Engineering, May 22, 2008

Most scientific programs have large input and output data sets that require out-of-core programmi... more Most scientific programs have large input and output data sets that require out-of-core programming or use virtual memory management (VMM). Out-of-core programming is very error-prone and tedious; as a result, it is generally avoided. However, in many instance, VMM is not an effective approach because it often results in substantial performance reduction. In contrast, compiler driven I/O management will allow a program's data sets to be retrieved in parts, called blocks or tiles. Comanche (COmpiler MANaged caCHE) is a compiler combined with a user level runtime system that can be used to replace standard VMM for out-of-core programs. We describe Comanche and demonstrate on a number of representative problems that it substantially out-performs VMM. Significantly our system does not require any special services from the operating system and does not require modification of the operating system kernel.

Best Papers from the 2006 CLEI Conference, Santiago, Chile, August, 2006

Clei Electronic Journal, 2007

I/O Minimization in a Genetic Sequencing Framework

Parallel and Distributed Processing Techniques and Applications, 1997

Block Mapping - A Compiler Driven I/O Management Study

Parallel and Distributed Processing Techniques and Applications, 2000

On testing for immutability of codes (Corresp.)

IEEE Transactions on Information Theory, Nov 1, 1987

Ahstracf -Immutable codes have the property that information recorded with them on write-once mem... more Ahstracf -Immutable codes have the property that information recorded with them on write-once memories such as digital optical discs cannot be changed. Write-once memory permits changing a 0 into a 1, but once a 1 is written it cannot be changed back into a 0. Most commonly used codes do not have the property of immutability. After a very general definition of immutability is given, algorithms which test a given code for immutability are developed. Algorithms are presented for fixed-length codes as well as for variable-length codes.

Generalized language equations with multiple solutions

Theoretical Computer Science, 1986

A generalized language equation X = F(X) is an equation in a variable X over an alphabet A where ... more A generalized language equation X = F(X) is an equation in a variable X over an alphabet A where the language function F is expressible in terms of arbitrary set operations (including complementation) and concatenation from the left by a constant (i.e., a language). In previous work, these equations were studied provided an easily verifiable condition was met. In this case, solutions always exist and are always unique. In the present paper, the condition is dropped. As a consequence, our equations may have zero, one, or more (up to uncountably infinitely many) solutions. We give a representation theorem for these equations and use it to classify them. For equations in one variable, a complete classification is given, including parametric representations of all solutions. For systems of m equations in m variables, the notion of 1-reducible substitution graph is introduced and employed to reduce the problem of existence of solutions of the entire system to the same problem for a single equation. The paper concludes with a discussion of the implications if one assumes that all constants are regular languages, and poses some open problems.

Download

Parallel and vector computing: a practical introduction

McGraw-Hill, Inc. eBooks, Apr 1, 1995

Classification of parallelism commercial vector and parallel hardware programming parallel system... more

On generalized language equations

Theoretical Computer Science, 1981

Gi are functions from [P(A*)]" into P(A*), i = 1 .... , n, P(A*) denoting the set of all language... more Gi are functions from [P(A*)]" into P(A*), i = 1 .... , n, P(A*) denoting the set of all languages over A. Furthermore the G~ are expressible ir~ terms of set-operations, concatenations, and stars which involve the variables X, as well as certain fixed languages° In this note we investigate existence and uniqiuenes~ of solutions of a certain subclass of generalized language equations. Furthermore we show that a soiution is regular if all fixed languages are regular.

On equations for regular languages, finite automata, and sequential networks

Theoretical Computer Science, 1980

We consider systems of equations of the form Xi=,IJ,a l &~LJ& i=l,...,n where A is the underlying... more We consider systems of equations of the form Xi=,IJ,a l &~LJ& i=l,...,n where A is the underlying alphabet, the Xi are variables, the Pi.0 are boolean functions in the variables J&B and each & is either the empty word or the empty set. The symbolsand u denote concatenation and union of languages over A. We show that any such system has a unique solution which, moreover, is regular. These equations correspond to a type of automaton, called boolean automaton, which is a generalization of a nondeterministic automaton. The equations are then used to determine the language accepted by a sequential network; they are obtainable directly from the network.

Download

Implicit Equations with Union and Left-Concatenation

Springer eBooks, 1999

We begin our coverage of implicit equations by considering the exact analogue to the classical (e... more We begin our coverage of implicit equations by considering the exact analogue to the classical (explicit) equations solved in Chapter 3. Specifically, the operations are union and left-concatenation and the constants are over an arbitrary alphabet. We first establish that, in contrast to explicit equations, implicit ones need not have any solutions. Furthermore, there exist implicit equations with context-free constants whose only solutions are non-context-free. Then we show a general criterion for the existence of a solution. This criterion is constructive if the constants are regular. Furthermore, the criterion gives rise to a test whether there exist finitely or infinitely many solutions. If the constants are regular, solutions can be constructed effectively; moreover, if there are finitely many solutions, all solutions can be constructed, and if there are infinitely many, an arbitrarily large number of them can be constructed.

Statistical Database Security

A statistical database is a collection of data items each of which is to be considered confidenti... more A statistical database is a collection of data items each of which is to be considered confidential but access to statistical information about these data items is to be allowed. The central questions of this chapter can be formulated as follows: Is a given statistical database secure, i.e., using the responses to permitted queries is it possible to determine the previously unknown value of any of the confidential data items stored in the database, and more fundamentally, are there secure statistical databases?

Constructing stochastic networks via β-RBF networks

Proceedings of International Conference on Neural Networks (ICNN'96), Dec 24, 2002

Without considering spatial, stochastic, and temporal features inherent in natural neural systems... more Without considering spatial, stochastic, and temporal features inherent in natural neural systems, the computational power of conventional artificial neural networks (ANNs) is limited. In the present paper, we look at the stochastic complexity and construct a stochastic ANN by modeling stochastic fluctuations in the environmental stimuli such that all stimuli are prone to be corrupted by noise or even outliers and to break networks down; therefore, a positive-breakdown network is required. We investigate the stochasticity in the domain of function approximation (estimation) in the framework of radial basis function networks (RBFNs) and propose a robust RBFN, /spl beta/-RBFN, by applying the breakdown point approach in robust regression. Experimental results demonstrate the advantages of the proposed networks in robustness and simplicity over the plain RBFNs.

The worst parallel Hanoi graphs

Theoretical Computer Science, 2018

The "Towers of Hanoi" is a problem that has been extensively studied and frequently generalized. ... more The "Towers of Hanoi" is a problem that has been extensively studied and frequently generalized. In particular, it has been generalized to be played on arbitrary directed graphs and using parallel moves of two types. We ask what is the largest number of parallel moves, in either of the two models, that is required to move n disks from the starting node to the destination node. Not all directed graphs allow solving this problem; we will call those graphs that do Hanoi graphs. In previous work, we settled the question of what are the worst sequential Hanoi graphs, that is, those graphs that require the largest number of sequential moves. We also demonstrated that the characterization of sequential Hanoi graphs carries over the parallel Hanoi graphs. Here, we determine the worst Hanoi graphs provided parallel moves are allowed. It turns out that for one of the two models of parallel moves, the worst graphs are quite different from the worst sequential graphs, while in the other model of parallelism, there is little difference with the sequential situation.

Compile time data transfer analysis

The design and implementation of compilers for high performance compiler systems require a thorou... more The design and implementation of compilers for high performance compiler systems require a thorough understanding of the target architecture to deliver the highest level of performance. It is important to know the I/O behavior on memory hierarchies. Compile time data transfer analysis computes an accurate approximation on the number of data transfers between main and secondary memory when the Least Recently Used replacement policy is applied. This analysis is performed at compile time; since some details are unknown, simplifying assumptions have to be accepted from the outset. The analyzer estimates reasonably precisely the number of page transfers that a program needs for execution, based on the parameters of the system available at compile time. The results reflect the real behavior of the system when executing the program and can therefore be used to compare situations derived for the same program after code restructuring.

Download

Panel

Increasingly, techniques from data analytics fields of statistics, machine learning, data mining,... more Increasingly, techniques from data analytics fields of statistics, machine learning, data mining, and natural language processing are being employed for challenges in cyber-security and privacy. This panel examines which techniques from these fields are essential for current and future cyber-security practioners and what are the related considerations involved in successfully solving security and privacy challenges of the future.

Compiler driven i/o management

We show that a compiler combined with a user level runtime system can replace and out-perform sta... more We show that a compiler combined with a user level runtime system can replace and out-perform standard virtual memory management (VMM) for out-of-core problems. The execution time for a single process and the overall impact on system resource utilization is lower for our compiler/runtime system combination than for VMM. Another advantage of our model is that it does not require any special services from the operating system and does not require modification of the operating system kernel. We describe our architecture and present the results of a prototype implementation. We also present techniques for detecting and reducing dead weight in programs (data that are transferred but not used). We present a case study of a genetic sequencing program and a tool for detecting dead weight in virtual memory page transfers.

Implicit language equations: existence and uniqueness of solutions

Theoretical Computer Science, Jul 1, 1995

We introduce a new type of language equation, namely implicit equations, and derive a number of p... more We introduce a new type of language equation, namely implicit equations, and derive a number of properties which clearly demonstrate that the behavior of implicit equations differs substantially from that of conventional or explicit ones. Explicit language equations express each variable as a regular expression in terms of their variables and constant languages; implicit language equations express constant languages as regular expressions in terms of their variables and constant languages. An algorithm is given for determining a solution of a given system of implicit language equations, if one exists, provided all constant languages are regular and the operations permitted in the regular expressions are unrestricted union and concatenation from the left by a constant. A characterization of the uniqueness of a solution of such a system and a method for determining all solutions are also provided.

Download

Randomizing, A Practical Method for Protecting Statistical Databases Against Compromise

Very Large Data Bases, Sep 8, 1982

This paper reports on a method for protecting statistical databases against inference, developed ... more This paper reports on a method for protecting statistical databases against inference, developed over the past four years. The method, called randomizing, is conceptually elegant, easy to implement at very small additional cost, and requires no on-going maintenance. It can be equally applied to small and medium-size as well as large databases, because it does not depend on sampling techniques in the conventional sense. Thus we feel that randomizing is an eminently practical and effective method if the protection of statistical data against inference is of concern.

Download

Division and sign detection algorithms for residue number systems

Computers & mathematics with applications, 1984

This paper is concerned with the operations of division and sign detection in residue number syst... more This paper is concerned with the operations of division and sign detection in residue number systems. CORDIC, a well-known iterative algorithm originally designed for conventional floating point representation, is adapted for residue division. Since sign detection is required by the division algorithm we have also considered this problem. A modified sign detection algorithm which yields a high degree of parallelism is introduced resulting in time efficient division and sign detection.

Download

Ernst Leiss

Uploads

Papers by Ernst Leiss

Log In