See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/242260585

Theoretical Computer Science, TCS

Article · November 1999

CITATION READS

1 4,229

4 authors, including:

Stefan Arnborg Viggo Kann

KTH Royal Institute of Technology KTH Royal Institute of Technology
83 PUBLICATIONS 4,815 CITATIONS 101 PUBLICATIONS 5,315 CITATIONS

SEE PROFILE SEE PROFILE

Jens Lagergren
KTH Royal Institute of Technology
126 PUBLICATIONS 4,166 CITATIONS

SEE PROFILE

All content following this page was uploaded by Viggo Kann on 31 January 2014.

The user has requested enhancement of the downloaded file.

Theoretical Computer Science, TCS
Stefan Arnborg and Johan Håstad

The objective of the research in Theoretical Computer Science is to inves- The objective of the research in
tigate methods of efficient computation in a mathematically precise sense, Theoretical Computer Science
and to find lower bounds on the computational resources required for a is to investigate methods of
computation. The types of computations studied are either chosen for their efficient computation ...
tractability or for their importance in practical applications. The different
areas of study are described below.
During the period 1993-1999 the group has been supported by Nutek, http://www.nada.kth.se/
BFR, TFR, NFR, SSF, and HSFR. theory/

Approximation algorithms
Johan Håstad, Viggo Kann, Jens Lagergren

A large number of the known NP-complete problems are in fact optimiza-

tion problems, and for some of these optimization problems there are fast
approximation algorithms, i.e. algorithms guaranteed to find close to
optimal solutions. For example, the travelling salesperson problem in the
plane is NP-complete, but in polynomial time it can be solved approxi-
mately within every constant, i.e. for any ε>0 one can find a trip of
length at most 1+ε times the shortest trip possible. On the other hand,
some NP-complete problems are extremely hard to approximate. For
example, unless P=NP the maximum independent set problem cannot in
polynomial time be approximated within n1-ε for any ε>0, where n is the
number of vertices in the input graph.
The main topic of this project is to investigate to what extent the
optimum value of important NP-complete problems can be approximated
effectively. These investigations are naturally divided into two types of ac-
tivities, namely to prove positive and negative approximation results. To
get a positive result—an upper bound of the approximability—one con-
structs an algorithm, proves that it is efficient and approximates the prob-
lem within a certain accuracy. To get a negative result—a lower bound—
one usually proves that approximating a given problem remains NP-hard.
We have studied the approximability of several basic and important
combinatorial problems. Positive or negative results have been found for
e.g. some generalizations of the maximum cut problem, maximum number
of satisfied linear equations mod p, maximum p-section, and the travelling
salesperson problem with weights 1 and 2, see e.g. [Amaldi and Kann, 1998],
[Andersson, 1999], [Andersson and Engebretsen, 1998], [Andersson,
Engebretsen, and Håstad, 1999], [Engebretsen, 1999], [Håstad, 1999], [Kann,
Lagergren, and Panconesi, 1998].

65
 Viggo Kann and Pierluigi Crescenzi have compiled a list of the best
lower and upper bounds known for more than two hundred well-studied
NP-complete optimization problems. We try to collect all new results in the
wide area of approximation in order to keep the problem list updated. The
list is included in a new text book on approximation [Ausiello et al., 1999],
and is also available for everybody on the web as http://www.nada.kth.se/
~viggo/problemlist/. It has frequently been used by researchers throughout
the world in the last five years, see [Kann and Crescenzi, 1998].

Complexity
Johan Håstad, Mikael Goldmann

Computational complexity involves studying the amount of computa-

tional resources required to solve certain types of problems. Resources
typically considered are, for instance, the time (number of steps) of an
algorithm, or the size and depth of Boolean circuits.
The general area in which we have worked is that of establishing
Computational complexity
lower bounds for natural functions in restricted computational models. Two
involves studying the amount of
models of computation where this program has been especially successful
computational resources re-
are small-depth circuits and monotone circuits. The most significant steps
quired to solve certain types of
in both models were taken in the 1980’s but developments during the last
problems.
decade have increased our understanding considerably.
Our main results in the area of monotone complexity are the sym-
metric approximation method introduced by Berg and Ulfberg (see Ulfberg’s
thesis, [Ulfberg, 1999]), and the lower bounds for the depth of monotone
circuits computing connectivity by [Goldmann and Håstad, 1998]. The
former very much simplifies the rather complicated combinatorial argu-
ments that were previously used for proving lower bounds for monotone
circuits.
For small-depth circuits a number of results were established using
similar methods. [Lai, Chen and Håstad, 1998] studied how bottom fan-in
affected the computational power of small-depth circuits, and Berg and
Ulfberg used circuit complexity techniques to construct oracles that sepa-
rate the levels of the PPPH hierarchy (see [Ulfberg, 1999]).
As for more general models, [Håstad, 1998] also established the
strongest lower bound on the formula size of any function in NP.

66
Cryptography
Johan Håstad, Mikael Goldmann, Mats Näslund

Our work has been mostly foundational. A basic concept in cryptography

is that of a one-way function, a function that is easy to compute and hard
to invert.
It has been known for a very long time that given a one-way function
with suitable structure (e.g., a function that is one-to-one) it is possible to
construct a strong crypto system. Results by [Håstad et al., 1998] show that
in fact no structure is needed, and any one-way function is sufficient to
construct a strong crypto system.
Strong crypto systems can be constructed through the use of good Strong crypto systems can be
pseudorandom generators. A very convenient way to construct a constructed through the use of
pseudorandom generator is to iterate a one-way function f and on each it- good pseudorandom generators
eration output one or more bits obtained from the current value of the iter-
ate. That is, starting from a seed x0 one iterates the function producing x1 =
f(x0), x2=f(x1) and so on, each time outputting a bit B(x0 ), B(x1 ), B(x2 ).... To
prove that such a scheme is secure (i.e., that the sequence of bits ”looks
random”) one would typically need that the bit output is a so-called hard-
core predicate. This means that B(x) is easy to compute given x but hard
when only given f(x).
Näslund proved that for any one-way function the standard pairwise
independent hash-functions do provide hard-core predicates. These are very
simple functions and Goldmann and Näslund proved that this is about as
simple as possible for functions which one can hope to have this property
(see Näslund’s thesis, [Näslund, 1998]). To be more specific, they proved
that any predicate that is hard-core for any one-way function requires, when
computed by unbounded fan-in circuits, circuits of almost logarithmic depth.
This work is currently being extended to a wider class of functions by
[Goldmann et al., 1999].
For specific functions one can of course have a simpler hard-core
predicate. The most famous function in cryptography is RSA-encryption
and it has been known for over a decade that the least significant bits give
hard-core predicates for this function. Settling a long standing conjecture
this was extended to each individual bit by [Håstad and Näslund, 1998].
The same result applied to the discrete logarithm function, the second most
used function in public-key cryptography.

67
 Decomposability
Stefan Arnborg, Jens Lagergren

The theory of tree-decomposable graphs became a foundation for, for in-

stance, BDD technology and Bayesian networks, two areas which have
numerous applications. Our recent work includes better bounds for obstruc-
tion sizes in [Lagergren, 1998], and an improved way of handling inde-
pendence assumptions in imprecise probability theory [Arnborg, 1999a].
Studies of dense decomposable graph families have been made, like graphs
of bounded clique-width and NLC-width (see [Johansson , 1998]). The first
progress in the (since 1985) open problem of complexity of decomposition
of dense decomposable families except co-graphs was obtained in
[Johansson, 1999], with a polynomial time algorithm for NLC2-decompo-
sition.

Formal methods based software testing

Karl Meinke

In software engineering, the classical approach to validating the correct-

ness of software consists of testing a system implementation against its
requirements. The testing process involves: (i) collecting an ad-hoc, often
random, collection of sample points from the entire input space, known as
the test set, (ii) executing the system on each input from the test set, and (iii)
evaluating the outcome of each such execution with reference to the soft-
ware requirements.
Step (ii) is uncontroversial. However, step (i) leads to a fundamental
problem known as the coverage problem – this is essentially the likelihood
that an implementation error is not discovered by the test set. Furthermore,
step (iii) leads to an equally fundamental problem, interpreting the outcome
of each test, and reaching one of the judgements: pass/fail/don’t know. This
is known as the oracle problem.
It seems fair to say that both the coverage and oracle problems are
still extremely poorly understood scientifically. This is despite the fact that
software testing is carried out intensively and widely within the IT industry.
The methodology of program testing seems to oppose the methodol-
ogy of formal methods, which advocates that programs be mathematically
proved correct (for all possible input values). Although formal program
verification has a completely sound scientific basis, it is often difficult to
put into practise for real programs, and for certain data types, notably the
floating point type.
In fact, it is possible to see program verification and program testing
as complementary approaches. Starting from a set R of program require-

68
ments, expressed in some logic, verification seeks a proof of R while test-
ing seeks a satisfying assignment for the negation of R. Thus (by Gödel’s
Completeness Theorem) testing and verification are in a strong sense dual.
During the period in question, we have carried out research into for-
mal methods based black box software testing. This research includes:
(i) Case studies of software requirements, particularly in the domains
of computational finance (Black-Scholes) and control systems design (au-
tomotive systems). Characteristic of these areas is the need for floating point
computation, and thus the limited applicability of present day formal verifi-
cation methods.
(ii) Algorithm design for automated testing. We develop decision al-
gorithms for the satisfiability problem for first-order logic over finite large
cardinality data types. First order formulas are used to define program pre-
and postconditions which can provide an oracle during testing. Our ap-
proach to algorithm design is based upon function approximation theory.
The basic strategy is to iteratively search for a satisfying assignment, using
the outcome of searches to construct an approximate model of the system,
from which we can prune the search space.
Future research in this area will include:
(iii) Theoretical study of the coverage problem. We are attempting to
find an appropriate stochastic model for the variance between the func-
tional behaviour of programs and their approximate models in the sense of
(ii). For example, is a random walk model appropriate? Such models could
be used to estimate coverage in a mathematically precise way.
Results in this project are presented in [Abdulla et al.,1999], [Berg et
al., 1999], and in [Meinke and Nielsen, 1999].
A description of our research project for the layman can be found on
the web at http://www.nada.kth.se/\~karlm/Testing.htm .

Computational biology
Jens Lagergren, Henrik Eriksson

Algorithms on strings (sequences of characters) have been studied since

the birth of computer science. In biology, perhaps the most important dis- In biology, perhaps the most
covery during this century is that of DNA and the genetic code. For a com- important discovery during this
puter scientist it is natural to formulate the most fundamental conclusion of century is that of DNA and the
these discoveries as: the genetic information of any being is a string over a genetic code.
four letter alphabet. Also proteins can be viewed as strings. Many other
structures are common to biology and computer science. For instance, evo-
lutionary trees are important in biology, and trees in general are fundamen-
tal to computer science.
Recent advances in molecular biology have lead to a large number of

69
 biologically motivated algorithmic problems. For instance, sequencing of
DNA constitutes a partly algorithmic problem. Moreover, it gives rise to an
enormous amount of discrete data, and the analysis of this data often gives
algorithmic problems. Many of the most interesting problems are NP-com-
plete, and thus expected to be difficult to solve, but sometimes there are
ways around this dilemma. One can consider approximation algorithms,
fixed parameter variations etc.
One fundamental problem in One fundamental problem in computational biology is to measure
computational biology is to the similarity between two genomes. The Sequence Alignment problem is
measure the similarity between to, given two DNA sequences, insert blanks into the sequences in such a
to genomes. way that they become as similar as possible.
Say that we are given two genomes as sequences of genes. One can
measure the distance between these two genomes by counting the least
number of operations that transform one into the other. One example of an
operation that is reasonable to use (since it mimics an operation on the
genome that takes place during evolution) is reversal, that is any subse-
quence of the gene-sequence may be reversed. We have studied the algo-
rithmic part of a procedure to obtain the gene sequences of a chromosome –
namely Radiation Hybrid mapping see [Håstad et al., 1998], and [Ivansson,
and Lagergren, 1999].
Another fundamental problem Another fundamental problem in computational biology is that of
in computational biology is that inferring the evolutionary history of a set of species, given some represen-
of inferring the evolutionary tation of the species. The evolutionary history is represented by a
history of a set of species, gi- phylogenetic tree. The vertices of a phylogenetic tree are labelled with spe-
ven some representation of the cies, given species or inferred species (representing ancestors of the given
species. species). Two adjacent nodes of the phylogenetic tree should be labelled
with species such that one is the ancestor of the other.
The following strategy is often used for reconstructing the evolu-
tionary relationships of a set of species (i.e. their species tree): One begins
by constructing phylogenetic trees for a set of distinct gene families (i.e.
gene trees). Typically, these gene trees are built using one of the standard
techniques for constructing phylogenetic trees from molecular sequence
data. However, for many gene families, the gene tree differs from the spe-
cies tree (using another terminology, their topologies disagree). Hence, a
single gene tree is not considered sufficient for inferring a species tree. For
this reason, a set of gene trees is often used in order to increase the reliabil-
ity of the resulting species tree. We have studied this algorithmic problem,
that is, given a set of disagreeing gene trees find the species tree [Hallet and
Lagergren, 1999].

70
Algorithms in language engineering
Viggo Kann

The goal of this project is to combine algorithms that have a solid math- The goal of this project is to
ematical base with linguistic knowledge in order to construct tools for Swed- combine algorithms ... to
ish text processing; tools that are both accurate and capable of handling construct tools for Swedish text
large amounts of data without speed degradation. We have earlier success- processing
fully constructed tools for Swedish hyphenation and Swedish spelling error
detection and correction.
Currently we are focusing on Swedish grammar checking and proof
reading. We are constructing Granska, a hybrid system using both statisti-
cal and linguistical methods for checking and correcting Swedish text. Much
work has been devoted to constructing a part-of-speech tagger that tags
each word in a sentence with its word class and inflectional features, see
[Carlberger and Kann, 1999]. By attacking the problems of disambiguation
of ambiguous words and tagging of unknown words we have managed to
obtain a tagger that tags 97% of the words correctly.
We have also constructed and implemented an efficient and power-
ful artificial language to express grammatical errors and corrections. In the
IPLab part of the project, many rules in this language have been written
http://www.nada.kth.se/
and evaluated, and a graphical user interface has been constructed. There is
theory/projects/granska/
a web interface to Granska: http://www.nada.kth.se/theory/projects/granska/
demo.html
demo.html

Intelligent Data Analysis and Visualization

Stefan Arnborg

In a project within the Centre for Geoinformatics, Johannes Keukelaar de-

velops methods for rough set classification using a visual language. This
technique is useful, for instance, when a number of qualitatively different
classifications of geographical areas are merged, see [Keukelaar, 1999] and
[Ahlqvist et al., 1999].

Design and development of brain image databases

Per Svensson

The project is to perform research on architecture, design and usage quality The project is to perform re-
aspects of large-scale brain image databases and related data analysis man- search on architecture, design
agement systems. It combines research objectives from the subject areas of and usage quality aspects of
database technology and brain image analysis so as to contribute to the large-scale brain image
ECHBD image database and the BINS neuroinformatics analysis center database...
projects (see below), investigate the application of advanced object-oriented

71
 methodology in heterogeneous, networked scientific image analysis sys-
tems, and develop improved technology for meta-research and data mining
in large-scale brain image databases.
The European Computerized Human Brain Database (ECHBD)
project is supported by an EC Biotech grant and the Brain Image
Neuroinformatics System (BINS) by an SSF grant. Both projects were ini-
tiated and are led by professor Per Roland at the Division of Human Brain
Research of the Karolinska Institute in Stockholm.
There is no mature methodology for management of 3D spatial and
spatiotemporal image databases. Therefore, one objective of the project is
to evaluate and refine emerging such methodology, in particular the
RasDaMan system. We will also investigate the use of mediator technology
(the AMOS II system) in a large and complex scientific database environ-
ment.
The project has a neuroinformatics usage perspective involving man-
agement of very large, inhomogeneous data collections consisting mainly
of 3D raw data images, logical and statistical operations and queries on
sequences of such images, performance requirements arising from interac-
tive display 3D images, and web distribution and access of images.
From this usage perspective, the project investigates multidatabase
architecture and data modeling issues, data mining methods for functional
PET and MRI brain imagery, use of a supercomputing center as a backend
storage and computing resource, and performance issues in storage, access,
query, and display of brain imagery.
A brain image database system, based on the RasDaMan raster data-
base manager and the O2 object-oriented database management system,
was developed for the ECHBD project. The first working version of the
system was made available to the ECHBD project partners in September
1999.
A research group has been formed whose first task is to design the
general architecture of the BINS raw image database system.
Some of the groups results can be found in [Fredriksson, 1999] and
in [Fredriksson et al., 1999].

Object-relational language and system

for spatial analysis (AMOROSE)
Per Svensson

This project researches state-of-the-art spatial database technology to im-

prove the interface between users and Geographical Information Systems
(GIS), as well as between GIS and databases.

72
 Current GIS’s lack efficient facilities for integrating the base system
and a domain model for spatial analysis. Such facilities may be provided by
an object-oriented database language, whose primitives represent funda-
mental operations of spatial analysis.
The use of a declarative object-oriented query language for domain
modelling offers several advantages over conventional imperative program-
ming techniques. It permits clear expression of ad-hoc analysis problems
on a high level of abstraction. Declarative models are easy to define, in-
spect and understand and lead to compact, easily reusable, and powerful
domain models. Object-oriented query languages provide object views which
are invoked independently of whether they represent stored or derived data.
This feature supports data independence and schema evolution. The Amorose project aims to
The Amorose project aims to develop a prototype spatial analysis develop a prototype spatial
and database management system based on the AMOS II main-memory analysis and database manage-
object mediator system (Risch et al., Uppsala University) with spatial ex- ment system...
tensions from the ROSE library (Gueting et al., Fernuniversitaet Hagen) ,
the prototype will be used to evaluate the prototype over a representative
set of spatial data analysis tasks.
In the first phase of the project (1996–98), the Rose library was
brought to a state where it satisfies the basic needs of the Amorose project.
As a result of this work, Rose was used as the basic computational tool in a
study of the concept of rough sets for uncertainty representation in spatial
data classification, see [Ahlqvist et al., 1999].
The second phase involved the preliminary integration of Rose into
Amos. This first integrated Amorose prototype showed how complex spa-
tial queries could be formulated in AmosQL and executed within Amorose.
The ongoing third development phase aims at close integration of
the modified ROSE library and the Windows NT-based Amos II mediator
system. The resulting Amorose system will provide a scalable memory in-
tegration mechanism extending the Amos query optimizer and making use
of query optimization techniques for efficient query evaluation.

Inference from data

Stefan Arnborg

Under this heading we study fundamental and applied inference problems.

The fundamental studies are related to ongoing discussions about the role
and possible adaptation of Bayesianism to applications in semi-intelligent
computer systems. The applied studies concern the possibilities for advanced www.ki.se/cns/hubin/.
data mining techniques to help investigating major medical puzzles, like
explanation and treatment of schizophrenia. See [Arnborg, 1999a] and
Arnborg [1999b].

73
 References–TCS

Book
[Ausiello et al., 1999] Ausiello, G., Crescenzi, P., Gambosi, G., Kann, V.,
Marchetti Spaccamela, A., and Protasi, M. (1999), Complexity and
Approximation— combinatorial optimization problems and their
approximability properties, Springer Verlag.

Theses
[Engebretsen, 1998] Engebretsen, L. (1998), Approximating
Generalizations of Max Cut, Licentiate thesis.
[Keukelaar, 1999] Keukelaar, J. (1999), A Visual Programming Language
for the Analysis of Uncertain Spatial Data, Licentiate thesis.
[Näslund, 1998] Näslund, M. (1998), Bit Extraction, Hard-Core Predicates,
and the Bit Security of RSA, PhD thesis.
[Ulfberg, 1999] Ulfberg, S (1999), On Lower Bounds for Circuits and
Selection, PhD thesis.

Journal publications
[Ahlqvist et al., 1999] Ahlqvist, O., Keukelaar, J. H. D., and Oukbir, K.
(1999), Rough classification and accuracy assessment, accepted to
Internat. J. Geographic Information Science.
[Amaldi and Kann, 1998] Amaldi, E. and Kann, V. (1998), On the
approximability of minimizing nonzero variables or unsatisfied
relations in linear systems, Theoretical Comput. Sci. 209, 237–260.
[Andersson and Engebretsen, 1998] Andersson, G. and Engebretsen, L.
(1998), Better approximation algorithms and tighter analysis for Set
Splitting and Not-All-Equal Sat, Inform. Process. Lett., 65:6, 305–
311.
[Berg and Ulfberg, 1998] Berg, C. and Ulfberg, S. (1998), A lower bound
for perceptrons and an oracle separation of the PPPH
hierarchy, J. Comput. and Syst. Sci. 56, 263–271.
[ Cai et al., 1998] Cai, L., Chen, J., and Håstad, J. (1998), Circuit bottom
fan-in and computational power, SIAM J. Computing 27:2, 341–
355.
[Carlberger and Kann, 1999] Carlberger, J. and Kann, V. (1999),
Implementing an efficient part-of-speech tagger, Software Practice
and Experience, 29, 815–832.
[Crescenzi and Kann, 1998] Crescenzi, P. and Kann, V. (1998), How to find
the best approximation results — a follow-up to Garey and Johnson,
SIGACT News 29:4, 90—97.

74
[Fernández-Baca and Lagergren, 1998] Fernández-Baca, D. and Lagergren,
J. (1998), On the approximability of the Steiner tree problem in
phylogeny, J. Discrete and Applied Math., Special issue on comp.
molecular biology, 88:127–143.
[Goldmann and Håstad, 1998] Goldmann, M. and Håstad J. (1998).
Monotone Circuits for Connectivity Have Depth (log n)2-o(1)
SIAM J. Computing, 27:5, 1283–1294.
[Goldmann and Karpinski, 1998] Goldmann, M. and Karpinski, M. (1998),
Simulating threshold circuits by majority circuits, SIAM J.
Computing, 27:1, 230–246.
[Goldreich and Håstad, 1998] Goldreich, O. and Håstad, J. (1998), On the
complexity of interactive proof with bounded communication, Inform.
Process. Lett. 67:4, 205–214.
[Håstad, 1998] Håstad, J. (1998), The shrinkage exponent of De Morgan
formulas is 2, SIAM J. Computing, 27:1, 48–64.
[Håstad, 1999] Håstad, J. (1999), Clique is hard to approximate within
n1-ε, Acta Mathematica, 182, 105–142.
[Håstad et al., 1999] Håstad, J., Imagliazzo, R., Levin, L., and Luby, M.
(1999), A Pseudorandom Generator from any one-way function,
SIAM J. Computing, 28:4, 1364–1396.
[Kann et al., 1999] Kann, V., Domeij, R., Hollman J., and Tillenius M.
(1999), Implementation aspects and applications of a spelling
correction algorithm, in R. Koehler, L. Uhlirova, G. Wimmer: Text
as a Linguistic Paradigm: Levels, Constituents, Constructs. Festschrift
in honour of Ludek Hrebicek.
[Kann et al., 1998] Kann, V., Lagergren, J., and Panconesi, A. (1998),
Approximate MAX k-CUT with subgraph guarantee, Inform. Pro
cess. Lett. 65, 145–150.
[Lagergren, 1998] Lagergren, J. (1998), Upper bounds on the size of
obstructions and intertwines. Journal of Combinatorial Theory
Series B, 73:1, 7–40.
[Johansson, 1998] Johansson, Ö. (1998), Clique-decomposition, NLC-
decomposition, and modular decomposition — relationships and
results for random graphs, CGTC 1998, Congressus Numerantium,
132, 39–60.
[Johansson, 1999] Johansson, Ö. (1999), Simple distributed ∆+1-
coloring of graphs,Inform. Process. Lett. 70:5, 229–232.

75
 Conference publications
[Abdulla et al., 1999] Abdulla, P., Ciapessoni, E., Marmo, P., Meinke, K.,
and Ratto, E. (1999), FAST: an integrated tool for verification and
validation of real time system requirements, in P.G. Larsen (ed) Proc.
Fifth FMRail Workshop, electronic publication with LNCS 1708/
1709, Proc. FM ’99, Springer Verlag.
[Andersson, 1999] Andersson, G. (1999), An approximation algorithm for
Max p-section, STACS 1999.
[Andersson and Engebretsen, 1998] Andersson, G. and Engebretsen, L.
(1998), Sampling methods applied to dense instances of non-Boolean
optimization problems, RANDOM 1998, LNCS 1518, 357–368.
[Andersson et al., 1999] Andersson, G., Engebretsen, L., and Håstad, J.
(1999), A new way to use semidefinite programming with
applications to linear equations mod p, SODA 1999, 41–50.
[Arnborg, 1999a] Arnborg, S. (1999a), Learning in Prevision Space, ISIPTA-
99.
[Aumann et al., 1999] Aumann, Y., Håstad, J., Rabin, M., and Sudan, M.
(1999), Linear consistency testing, Proceedings of Workshop on
Randomization and Approximation Techniques in Computer Science,
Berkeley CA.
[Engebretsen, 1999] Engebretsen, L. (1999), An explicit lower bound for
TSP with distances one and two, STACS 1999.
[Fredriksson, 1999] Fredriksson, J. (1999), Design of an Internet
accessible visual human brain database system, IEEE International
Confer-ence on Multimedia Computing and Systems (ICMCS ’99),
Florence, Italy.
[Fredriksson et al., 1999] Fredriksson, J., Roland, P. and Svensson, P. (1999),
Rationale and design of the European computerized human brain
database, System, Scientific and Statistical Database Management
(SSDBM ’99), Cleveland.
[Goldmann and Russell, 1999] Goldmann, M. and Russell, A. (1999), The
complexity of solving equations over finite groups, CCC 1999.
[Håstad et al., 1998] Håstad, J., Ivansson, L., and Lagergren, J. (1998),
Fitting points on the real line and its application to RH mapping,
ESA 1998, 465–476.
[Håstad and Näslund, 1998] Håstad, J. and Näslund, M. (1998), The
security of individual RSA bits, FOCS 1998, 510–519.
[Johansson, 1999] Johansson, Ö., NLC2-decomposition in polynomial
time, WG’99.
[Näslund and Russell, 1998] Näslund, M. and Russell, A. (1998)
Extraction of optimally unbiased bits from a biased source, IEEE
ITW ’98, 90–91.

76
 Misc
[Arnborg, 1999b] Arnborg, S. (1999b), A survey of Bayesian Data Mining
–Part I: Discrete and semi-discrete Data Matrices, SICS
Technical Report.
[Berg et al., 1999] Berg, C., Ekström, R., and Meinke, K. (1999),
Automatic test case generation for derivative trading software,
submitted to: TEST Congress 2000.
[Goldmann et al., 1999] Goldmann, M., Näslund, M., and Russell, A. (1999),
Spectral bounds on general hard core predicates, manuscript.
[Hallet and Lagergren, 1999] Hallet, M.T. and Lagergren, J, (1999), New
Algorithms for the Duplication- Loss Model, manuscript.
[Ivansson and Lagergren, 1999] Ivansson, and Lagergren (1999), Algorithms
for RH Mapping: New Ideas and Improved Analysis, manuscript.
[Meinke and Nielsen, 1999] Meinke, K. and Nielsen, J. (1999), User
requirements capture with tabular TRIO: a case study of a cruise
controller, submitted to Formal Aspects of Computing.

77

View publication stats