Nondeterministic Finite Automata—Recent Results

on the Descriptional and Computational

Complexity

1 2
Markus Holzer and Martin Kutrib
1
Institut f¨ur Informatik, Technische Universit¨at M¨unchen,
Boltzmannstr. 3, 85748 Garching bei M¨unchen, Germany
holzer@in.tum.de
2
Institut f¨ur Informatik, Universit¨at Giessen,
Arndtstr. 2, 35392 Giessen, Germany
kutrib@informatik.uni-giessen.de

Abstract. Nondeterministic finite automata (NFAs) were introduced in [67],

where their equivalence to deterministic finite automata was shown. Over the
last 50 years, a vast literature documenting the im-portance of finite automata
as an enormously valuable concept has been developed. In the present paper, we
tour a fragment of this literature. Mostly, we discuss recent developments
relevant to NFAs related prob-lems like, for example, (i) simulation of and by
several types of finite automata, (ii) minimization and approximation, (iii) size
estimation of minimal NFAs, and (iv) state complexity of language operations.
We thus come across descriptional and computational complexity issues of
nondeterministic finite automata. We do not prove these results but we merely
draw attention to the big picture and some of the main ideas involved.

1 Introduction

Nondeterministic finite automata (NFAs) are probably best known for being
equivalent to right-linear context-free grammars and, thus, for capturing the lowest
level of the Chomsky-hierarchy, the family of regular languages. It is well known that
NFAs can offer exponential saving in space compared with deterministic finite
automata (DFAs), that is, given some n-state NFA one can always construct a
n
language equivalent DFA with at most 2 states [67]. This so-called powerset
construction turned out to be optimal, in general. That is, the bound on the number of
states is tight in the sense that for an arbitrary n there is always some n-state NFA
n
which cannot be simulated by any DFA with less than 2 states [63,64]. These two
milestones from the early days of automata theory form part of an extensive list of
equally striking problems of NFA related problems, and are the basis of descriptional
complexity. Moreover, it initiated the study of the power of resources and features
given to finite automata, see, e.g., [21] for a survey on limited resources for finite
automata.

O.H. Ibarra and B. Ravikumar (Eds.): CIAA 2008, LNCS 5148, pp. 1–16, 2008.
c Springer-Verlag Berlin Heidelberg 2008
2 M. Holzer and M. Kutrib

Our tour on the subjects listed in the abstract of NFAs related problems cover some
(recent) results in the field of descriptional and computational complexity. It
obviously lacks completeness, as NFAs fall short of exhausting the large selec-tion of
finite automata related problems considered in the literature. We give our view of
what constitute the most recent interesting links to the considered prob-lem areas. Our
nomenclature of finite automata is as follows: A nondeterministic finite automaton
(NFA) is a quintuple A = (Q, Σ, δ, q0, F ), where Q is the finite
set of states, Σ is the finite set of input symbols, q0 ∈ Q is the initial state, F ⊆ Q
Q
is the set of accepting states, and δ : Q × Σ → 2 is the transition function. A
finite automaton is deterministic (DFA) if and only if |δ(q, a)| = 1, for all states q ∈
Q and letters a ∈ Σ. The language accepted by the finite automaton A is defined as
∗
L(A) = { w ∈ Σ | δ(q0 , w) ∩ F = ∅ }, where the transition function is recursively
∗ Q
extended to δ : Q × Σ → 2 . For further details we refer to [36].

2 Determinization and Simulations of NFAs

Since regular languages have many representations in the world of finite au-tomata, it
is natural to investigate the succinctness of their representation by di fferent types of
automata in order to optimize the space requirements. Here we measure the costs of
representations in terms of the states of a minimal au-tomaton accepting a language.
More precisely, the simulation problem is defined as follows:

– Given two classes of finite automata C1 and C2.

– How many states are sufficient and necessary in the worst case to simulate n-state
automata from C1 by automata from C2 ?
Probably the most famous simulation problem in the world of finite automata is the
simulation of NFAs by DFAs, which is widely known as determinization.
Theorem 1 (NFA Determinization). Let A be an n-state nondeterministic finite
n
automaton. Then 2 states are sufficient and necessary in the worst case for
a deterministic finite automaton to accept L(A).
For the particular case of finite and unary regular languages the situation is
significantly different. The determinization problem for finite languages over a
n
k-letter alphabet was solved in [70] with a tight bound of Θ(k 1+log2 k ). Thus,
n

for finite languages over a two-letter alphabet only Θ(2 2 ) states are sufficient and
necessary in the worst case for a DFA to accept a language specified by an n-state
NFA. The situation is similar when we turn to the second important special case, the
unary languages, that is discussed in more detail a bit later. Unary NFAs can be much
more concise than DFAs, but yet not as much as for the general case. For languages
that are unary and finite this is not the case, since in [60] it was proven that
nondeterminism does not help in this case. Unary DFAs are up to one additional state
are as large as equivalent minimal NFAs.
In the following, we concentrate on simulations between finite automata that may
or may not have the features nondeterminism and two-way head motion.