Skip to main content

1990 | Buch

Structural Complexity II

verfasst von: Prof. Dr. José Luis Balcázar, Prof. Dr. Josep Díaz, Prof. Dr. Joaquim Gabarró

Verlag: Springer Berlin Heidelberg

Buchreihe : Monographs in Theoretical Computer Science. An EATCS Series

insite
SUCHEN

Über dieses Buch

This is the second volume of a two volume collection on Structural Complexity. This volume assumes as a prerequisite knowledge about the topics treated in Volume I, but the present volume itself is nearly self-contained. As in Volume I, each chapter of this book ends with a section entitled "Bibliographical Remarks", in which the relevant references for the chapter are briefly commented upon. These sections might also be of interest to those wanting an overview of the evolution of the field, as well as relevant related results which are not included in the text. Each chapter includes a section of exercises. The reader is encouraged to spend some time on them. Some results presented as exercises are occasionally used later in the text. A reference is provided for the most interesting and for the most useful exercises. Some exercises are marked with a • to indicate that, to the best knowledge of the authors, the solution has a certain degree of difficulty. Many topics from the field of Structural Complexity are not treated in depth, or not treated at all. The authors bear all responsibility for the choice of topics, which has been made based on the interest of the authors on each topic. Many friends and colleagues have made suggestions or corrections. In partic­ ular we would like to express our gratitude to Richard Beigel, Ron Book, Rafael Casas, Jozef Gruska, Uwe Schoning, Pekka Orponen, and Osamu Watanabe.

Inhaltsverzeichnis

Frontmatter
Introduction
Abstract
The notion of algorithm is very rich and can be studied under many different approaches. One of them is given by structural complexity. In the sixties the notion of feasible algorithm was developed and gradually was identified with P problems. In the early seventies the class NP was raised and the polynomial time reductions were defined. From this moment on these ideas were enlarged and studied more deeply. In structural complexity we study and classify the inherent properties of the problems by means of resource-bounded reducibilities; special attention is paid to complete problems. It was necessary to bring together some of the most important or fundamental material in a “uniform” way.
José Luis Balcázar, Josep Díaz, Joaquim Gabarró
1. Vector Machines
Abstract
In recent times, parallel algorithms have become increasingly employed to solve certain problems. A strategy for increasing the speed of computing is to perform in parallel as much as possible of the desired computation. Many models of parallel machines have been studied. The next four chapters are devoted to the study of some of them.
José Luis Balcázar, Josep Díaz, Joaquim Gabarró
2. The Parallel Computation Thesis
Abstract
We shall study in this chapter the relation between sequential and parallel computations. To this we consider in detail other models of parallel computers like Array machines, SIMDAG’s machines or Tree machines.
José Luis Balcázar, Josep Díaz, Joaquim Gabarró
3. Alternation
Abstract
In this chapter we shall study another computational model: the alternating Turing machine. This model combines the power of nondeterminism with parallelism by alternating the existential and universal states. Alternation connects time and space complexity rather well, in the sense that alternating polynomial time equals deterministic polynomial space, and alternating linear space equals deterministic exponential time. Analogous relationships hold for other time and space bounds. In the next chapter, we shall see how alternation links with the Parallel Computation Thesis presented in the previous chapters.
José Luis Balcázar, Josep Díaz, Joaquim Gabarró
4. Uniform Circuit Complexity
Abstract
In the previous chapters we have studied some models of parallel machines. In this chapter, we are going to define some parallel complexity classes. In modelling parallel computation, we wish to model the situation in which the number of processors is greater than the length of the input; however it is clear that any “real world” parallel computer should have a feasible number of processors. For this reason most of the theoretical research on parallelism has focused on the case in which the number of processors is bounded by a polynomial on the size of the input. As we have mentioned in the Chapters one and two, the number of theoretical models of parallel computation is very large; we shall define complexity classes in terms of uniform families of circuits, which seems to be a unifying framework for most of the parallel models, as well as a natural measure to count the size of the hardware. In the second section, we introduce the basic definitions and properties. In the following section we prove the robustness of parallel complexity classes with respect to other models of parallel computers studied in Chapters 1 and 2; in particular we prove the equivalence of uniform circuits with vector machines. Then we discuss other measures of uniformity and its incidence in the definition of parallel classes. In section 4, we establish the relationship of uniform circuits with alternating machines, and prove a characterization of the defined parallel classes in terms of alternating machines.
José Luis Balcázar, Josep Díaz, Joaquim Gabarró
5. Isomorphism and NP-completeness
Abstract
An interesting area of research in Complexity Theory is the study of the structural properties of the complete problems in NP. The analogy with the class of the recursively enumerable sets, provided by the polynomial time hierarchy, suggests that they might have properties similar to those of the r.e.-complete sets, like being pairwise isomorphic, in the appropriate sense; in fact, sufficient conditions are known for certain sets being isomorphic under polynomial time computable, polynomially invertible bijections, and no NP-complete set has been proved to be non-isomorphic to SAT in this sense. It was conjectured that all NP-complete sets are polynomially isomorphic, a statement which is known as the BermanHartmanis conjecture. Several consequences follow from this conjecture; among them, of course, P ≠ NP.
José Luis Balcázar, Josep Díaz, Joaquim Gabarró
6. Bi-Immunity and Complexity Cores
Abstract
In this chapter we present several interesting results regarding bi-immunity and related notions. We characterize the sets which are bi-immune for P by several different properties. A stronger notion is defined, and a construction of a set having this property is presented. Then it is shown that several properties of polynomial time m-reducibility and of the sets complete for this reducibility can be deduced from the existence of bi-immune sets.
José Luis Balcázar, Josep Díaz, Joaquim Gabarró
7. Relativization
Abstract
Many of the questions we address in this book can be thought of as questions about the properties of the resource-bounded reducibilities. As examples, the important concept of completeness (or incompleteness) in complexity classes is defined in terms of reducibility, and the polynomial time hierarchy is the closure of P under nondeterministic polynomial time Turing reducibility. We have shown in the previous chapter that polynomial time m-reducibility and T-reducibility differ. It is time to address a similar question: Do the deterministic and nondeterministic polynomial time Turing reducibilities differ? Similarly: do deterministic polynomial time reducibility and deterministic polynomial space reducibility differ?
José Luis Balcázar, Josep Díaz, Joaquim Gabarró
8. Positive Relativizations
Abstract
This chapter presents an attempt to fully understand the implications of the results about contradictory ways of relativizing statements involving complexity classes. The key point of this chapter is given by the following question: what do the proofs in the previous chapter really say? What are the properties of complexity classes that make the proofs work? Are there intrinsic characteristics of complexity classes that explain the meaning of these results?
José Luis Balcázar, Josep Díaz, Joaquim Gabarró
9. The Low and the High Hierarchies
Abstract
In this chapter we shall study two hierarchies of classes of sets, obtained by formalizing the concept of the information content of a set, considered as an oracle. This should not be confused with the information contents of a string, or Kolmogorov complexity, which will be studied in the next chapter. We want to be able to evaluate the amount of information provided by the set to the oracle Turing machines using it, so that questions such as “is this set useful?” or “is this set almost useless?” do make sense. The grading scale for comparing the power of different sets, used as oracles, will be the polynomial time hierarchy. We will compare classes Σ i (A) with unrelativized classes Σ j , looking for a nontrivial equality or inclusion. We define first a low and high hierarchy within NP. Later we present generalizations of this work.
José Luis Balcázar, Josep Díaz, Joaquim Gabarró
10. Resource-Bounded Kolmogorov Complexity
Abstract
There is a classical approach to the problem of classifying strings with respect to the difficulty of computing them; in some sense, we look for a definition of the amount of information coded by the string. Intuitively, a difficult string contains a high amount of compactly coded information, and one has to know all this information in order to write down the string. Conversely, an easy string has low information content, and one can describe the string in a compact way, so that from this small description one can obtain enough information to reconstruct the string and write it down. This idea can be formalized as we do in this chapter, giving rise to a different concept of complexity, known as Kolmogorov complexity, which has been succesfully used for proving lower bounds in a more concrete approach to complexity theory.
José Luis Balcázar, Josep Díaz, Joaquim Gabarró
11. Probability Classes and Proof-Systems
Abstract
In this last chapter, we bring together many of the concepts previously treated in this volume and in Volume I, to introduce proof-systems and frame them in the more general theory of probability classes. In the second section, taking as starting point the concept of NP, we introduce the interactive proof-systems. In the following two sections, we define another kind of proof-system, the Arthur-Merlin games, and prove some structural properties of these proof-systems, by considering them as probabilistic classes. The reader will recognize some of the arguments presented in Section 11.4 as generalizations of results from Chapters 6 and 8 in Volume I. Finally, in the last section, we prove that for a bounded number of interactions, interactive proof-systems classes and Arthur-Merlin classes coincide.
José Luis Balcázar, Josep Díaz, Joaquim Gabarró
Backmatter
Metadaten
Titel
Structural Complexity II
verfasst von
Prof. Dr. José Luis Balcázar
Prof. Dr. Josep Díaz
Prof. Dr. Joaquim Gabarró
Copyright-Jahr
1990
Verlag
Springer Berlin Heidelberg
Electronic ISBN
978-3-642-75357-2
Print ISBN
978-3-642-75359-6
DOI
https://doi.org/10.1007/978-3-642-75357-2