This is a survey of space-bounded probabilistic computation, summarizing the present state of knowledge about the relationships between the various complexity classes associated with such computation. The survey especially emphasizes recent progress in the construction of pseudorandom generators that fool probabilistic space-bounded computations, and the application of such generators to obtain deterministic simulations.

Let M be a logarithmic space Turing machine (or a polynomial width branching program) that uses up to k 2 p log n (read once) random bits. For a fixed input, let P i (S) be the probability (over the random string) that at time i the machine M is in state S, and assume that some weak estimation of the probabilities P i (S) is known or given or can be easily computed. We construct a logarithmic space pseudo-random generator that uses only logarithmic number of truly random bits and outputs a sequence of k bits that looks random to M . This means that a very weak estimation of the state probabilities of M is sufficient for a full derandomization of M and for constructing pseudo-random sequences for M . We have several applications of the main theorem, as stated within. To prove our theorem, we introduce the idea of recycling the state S of the machine M at time i as part of the random string for the same machine at later time. That is, we use the entropy of the random variable S in o...

. The short term behavior of random walks on graphs is studied, in particular, the rate at which a random walk discovers new vertices and edges. A conjecture by Linial that the expected time to find N distinct vertices is O(N 3 ) is proved. In addition, an upper bound of O(M 2 ) on the expected time to traverse M edges, and O(MN ) on the expected time to either visit N vertices or traverse M edges (whichever comes first) is proved. Key words. random walk, graph, Markov chain AMS subject classification. 60J15 1. Introduction. Consider a simple random walk on G, an undirected graph with n vertices and m edges. At each time step, if the walk is at vertex v, it moves to a vertex chosen uniformly at random from the neighbors of v. Random walks have been studied extensively, and have numerous applications in theoretical computer science, including space-efficient algorithms for undirected connectivity [4, 8], derandomization [1], recycling of random bits [10, 15], approximation algori...

We present fast and e#cient parallel algorithms for finding the connected components of an undirected graph. These algorithms run on the exclusive-read, exclusive-write (EREW) PRAM. On a graph with<F3.492e+05> n<F3.822e+05> vertices and<F3.492e+05> m<F3.822e+05> edges, our randomized algorithm runs in<F3.492e+05><F3.822e+05> O(log<F3.492e+05><F3.822e+05> n) time using<F3.492e+05> (m<F3.822e+05> +<F3.492e+05> n<F2.77e+05><F2.072e+05> 1+#<F3.822e+05><F3.492e+05> )/<F3.822e+05> log<F3.492e+05> n<F3.822e+05> EREW processors (for any fixed<F3.492e+05> # ><F3.822e+05> 0). A variant uses<F3.492e+05> (m<F3.822e+05> +<F3.492e+05><F3.822e+05><F3.492e+05> n)/<F3.822e+05> log<F3.492e+05> n<F3.822e+05> processors and runs in<F3.492e+05><F3.822e+05> O(log<F3.492e+05> n<F3.822e+05> log log<F3.492e+05><F3.822e+05> n) time. A deterministic version of the algorithm runs in<F3.492e+05><F3.822e+05> O(log<F2.77e+05><F2.072e+05><F2.77e+05> 1.5<F3.492e+05><F3.822e+05> n) time using<F3.492e+...

We present a Logspace, many-one reduction from the undirected st-connectivity problem to its complement. This shows that SL = co - SL.

We present two techniques for approximating probability distributions. The first is a simple method for constructing the small-bias probability spaces introduced by Naor & Naor. We show how to efficiently combine this construction with the method of conditional probabilities to yield improved NC algorithms for many problems such as set discrepancy, finding large cuts in graphs, finding large acyclic subgraphs etc. The second is a construction of small probability spaces approximating general independent distributions, which is of smaller size than the constructions of Even, Goldreich, Luby, Nisan & Velickovi'c. Such approximations are useful, e.g., for the derandomization of certain randomized algorithms. Keywords. Derandomization, parallel algorithms, discrepancy, graph coloring, small sample spaces, explicit constructions. 1 Introduction Derandomization, the development of general tools to derive efficient deterministic algorithms from their randomized counterparts, has blossomed ...

This paper concerns some of the theoretical complexity aspects of the reconfigurable network model. The computational power of the model is investigated under several variants, depending on the type of switches (or switch operations) assumed by the network nodes. Computational power is evaluated by focusing on the set of problems computable in constant time in each variant. A hierarchy of such problem classes corresponding to different variants is shown to exist and is placed relative to traditional classes of complexity theory. Department of Mathematics and Computer Science, The Haifa University, Haifa, Israel. E-mail: yosi@mathcs2.haifa.ac.il y Department of Computer Science, Technische Universitat Munchen, 80290 Munchen, Germany. E-mail: lange@informatik.tu-muenchen.de z Department of Applied Mathematics and Computer Science, The Weizmann Institute, Rehovot 76100, Israel. E-mail: peleg@wisdom.weizmann.ac.il. Supported in part by an Allon Fellowship, by a Bantrell Fellowship an...

We point out how the methods of Nisan [Nis90, Nis92], originally developed for derandomizing space-bounded computations, may be applied to obtain polynomial-time and NC derandomizations of several probabilistic algorithms. Our list includes the randomized rounding steps of linear and semi-definite programming relaxations of optimization problems, parallel derandomization of discrepancy-type problems, and the Johnson--Lindenstrauss lemma, to name a few.

Directed s-t connectivity is the problem of detecting whether there is a path from vertex s to vertex t in a directed graph. We present the first known deterministic sublinear space, polynomial time algorithm for directed s-t connectivity. For n-vertex graphs, our algorithm can use as little as n=2 \Theta( p log n) space while still running in polynomial time.

We clarify the computational complexity of planarity testing, by showing that planarity testing is hard for L, and lies in SL. This nearly settles the question, since it is widely conjectured that L = SL [25]. The upper bound of SL matches the lower bound of L in the context of (nonuniform) circuit complexity, since L/poly is equal to SL/poly. Similarly, we show that a planar embedding, when one exists, can be found in FL SL . Previously, these problems were known to reside in the complexity class AC 1 , via a O(log n) time CRCW PRAM algorithm [22], although planarity checking for degree-three graphs had been shown to be in SL [23, 20].