TOC Finite Automata

The computation power of this model is rather weak. Because it can only use a limited (finite) memory. But it is also useful in some applications.

DFA

+ Deterministic finite automata $DFA$

A deterministic finite automata (DFA) is a 5-tuple $(Q,\mathrm{\Sigma },\delta ,q_0,F)$ , where

• $Q$ is a finite set called states,
• $\mathrm{\Sigma }$ is a finite set called alphabet,
• $\delta :Q\times \mathrm{\Sigma }\to Q$ is the transition function,
• $q_0\in Q$ is the start state (also called initial state), and
• $F\subseteq Q$ is the set of accept states (final states).

The transition function $\delta$ can also be represented as a transition table. The table has $|Q|$ rows and $|\mathrm{\Sigma }|$ columns. The value in each entry is a state. $\delta$ is a total function. So, every entry has a value. The input of a DFA is a string in the language ${\mathrm{\Sigma }}^{\ast }$. The output of a DFA is only accept or reject.

Suppose $w=w_1{w}_2\cdots w_n$ is a string in ${\mathrm{\Sigma }}^{\ast }$. The DFA $M$ accepts $w$ if and only if:

$$\delta (q_0,w_1)=q_{s_1}$$$$\delta (q_{s_1},w_2)=q_{s_2}$$$$⋮$$$$\delta (q_{s_{n-1}},w_n)=q_{s_n}\in F$$

Each $q_{s_i}$ is a state of $M$. Each $\delta (q_{s_i},w_{i+1})=q_{s_{i+1}}$ is a transition. $q_{s_n}$ is a final state in $F$.

The sequence of transitions can also be expressed as:

$$\delta (\cdots \delta (\delta (q_0,w_1),w_2)\cdots ,w_n)\in F$$

Remember, a DFA can be represented as a state graph. The features of the graph are listed:

• A state is in a circle.
• A transition is an arc from the current state to the next state.
• The input symbol required by a transition is the label of the arc.
• The initial state is pointed by an arrow.
• Each final state is double circled.

Example

Let $M_1=(\{q_0,q_1\},\{0,1\},\delta ,q_0,\{q_1\})$ be a DFA, where the transition function $\delta$ is:

	0	1
$q_0$	$q_0$	$q_1$
$q_1$	$q_0$	$q_1$

$$L(M_1)=\{w\in \{0,1{\}}^{\ast }\mid w\text{ end with 1’s}\}.$$

If $A$ is the set of all strings that $M$ accepts, then:

• $A$ is the language of machine $M$;
• $L(M)=A$;
• $M$ recognizes $A$.

For example:

$$L(M_1)=\{w\mid w\text{ ends with a }1\}.$$

A language is regular if it is recognized by some DFA (Deterministic Finite Automaton).

NFA

+ Nondeterministic finite automata $NFA$.

An NFA (Nondeterministic Finite Automaton) is a 5-tuple $(Q,\mathrm{\Sigma },\delta ,q_0,F)$, where:

• $Q$ is the set of states;
• $\mathrm{\Sigma }$ is the alphabet;
• $\delta :Q\times \mathrm{\Sigma }\to \mathcal{P}(Q)$ is the transition function;
• $q_0\in Q$ is the start state; and
• $F\subseteq Q$ is the set of final states.

Suppose $M$ is an NFA to recognize the language $A$.

• If $w$ is a string in the language $A$:

  • Then there exists a "way" to reach the final state;
  • It's also possible to reach a non-final state.

• But if $w$ is not a string in the language $A$:

  • Then there is no way to reach the final state.

Please pay attention to the quantifiers.

Different between DFA and NFA

The difference between a DFA and an NFA is the transition function $\delta$.

• DFA: $\delta :Q\times \mathrm{\Sigma }\to Q$
• NFA: $\delta :Q\times \mathrm{\Sigma }\to \mathcal{P}(Q)$
  • $\mathcal{P}(Q)$ is the power set of $Q$.

This difference allows:

• One state to transition to zero, one, or multiple states by taking one input symbol.

$\epsilon -Moves$

Automatas can be further upgraded by a special transition.

+ $\epsilon-Move$

An $ϵ-Move$ is a transition from one state to another state by taking no input symbol.

Such automata is defined.

+ NFA with $\epsilon-Move$

An NFA with ε-moves is the same as an NFA (without ε-moves) except the transition function:

$$\delta :Q\times \mathrm{\Sigma }\cup \{\epsilon \}\to \mathcal{P}(Q)$$

Later, we will see that ε-moves allow us to recursively construct NFAs, which is... elegant.

Equivalence of DFA and NFA

Power

Let ${\mathcal{M}}_1$ and ${\mathcal{M}}_2$ be two computation models. ${\mathcal{M}}_1$ is as powerful as ${\mathcal{M}}_2$ if:

• For all languages $L$, $L$ is recognized by some machines in ${\mathcal{M}}_1$ if and only if it is recognized by some machines in ${\mathcal{M}}_2$.

${\mathcal{M}}_1$ is more powerful than ${\mathcal{M}}_2$ if there is a language $L$ which can be recognized by some machines in ${\mathcal{M}}_1$ but no machine in ${\mathcal{M}}_2$.

Note: A computation model is a set of all machines of the same type.

+ Theorem

DFA, NFA, and NFA with $\epsilon$-moves are of the same computation power.

For convenience, we denote the ordering on the power by ${=}_{\text{pow}}$, ${<}_{\text{pow}}$, ${\le }_{\text{pow}}$, etc. Trivially,

+ Lemma

$$\text{DFA}{\le }_{\text{pow}}\text{NFA}{\le }_{\text{pow}}{\text{NFA}}_{\epsilon }$$

Next, we only need to show ${\text{NFA}}_{\epsilon }{\le }_{\text{pow}}\text{DFA}$.

It seems the content repeats the explanation for converting ${\text{NFA}}_{\epsilon }$ to a DFA and introduces the concept of $\epsilon$-closure more formally. Here's a summary in Markdown format for your reference:

The Intuition of the Proof:

• For every ${\text{NFA}}_{\epsilon }$, we can construct a DFA that recognizes the same language.
• This is equivalent to the process of converting an NFA to an equivalent DFA.

+ Def: $\varepsilon$-closure

Let $N=(Q,\mathrm{\Sigma }\cup \{\epsilon \},\delta ,q_0,F)$ be an ${\text{NFA}}_{\epsilon }$.
The $\epsilon$-closure of a state $q$, denoted $E(q)$, is defined as:

$$E(q)=\{q\}\cup \delta (q,\epsilon )\cup \delta (\delta (q,\epsilon ),\epsilon )\cup \dots$$

• $E(q)$ represents the set of all states that can be reached starting from $q$ only using $\epsilon$-transitions.
• This closure includes $q$ itself and any other states reachable via one or more $\epsilon$-moves.
• Alternative Name: The $\epsilon$-closure is sometimes referred to as the Kleene Star or Kleene Closure.

+ Def: $\upvarepsilon-Closure$ of a Set of States

Let $R$ be a set of states for a given NFA. The $\epsilon$-closure of $R$, denoted $E(R)$, is:

$$E(R)=\bigcup _{q\in R}E(q)$$

This means the $\epsilon$-closure of $R$ is the union of the $\epsilon$-closures of all individual states in $R$. $E(q)$ is the set of states reachable from $q$ using only $\epsilon$-transitions.

+ Def: **$a$-Move of a Set of States:**

Let $R$ be a set of states, and let $a$ be a symbol for a given NFA. The $a$-move of $R$, denoted $\delta (R,a)$, is:

$$\delta (R,a)=\bigcup _{q\in R}E(\delta (q,a))$$

This describes all the states reachable by:

1. First making an $a$-transition from any state $q$ in $R$, and
2. Then taking all $\epsilon$-transitions from those resulting states.

NOTE: $\delta (q,a)$ gives the set of states reachable by consuming the symbol $a$.

Construction Algorithm

Let $N=(Q,\mathrm{\Sigma },\delta ,q_0,F)$ be an ${\text{NFA}}_{\epsilon }$. We construct a DFA $D=(Q^{\prime },\mathrm{\Sigma },{\delta }^{\prime },q_0^{\prime },F^{\prime })$, where:

• $Q^{\prime }=\mathcal{P}(Q)$
• ${\delta }^{\prime }:\mathcal{P}(Q)\times \mathrm{\Sigma }\to \mathcal{P}(Q)$ such that:$${\delta }^{\prime }(R,a)=\{q\in Q\mid \mathrm{\exists }r\in R,q\in E(\delta (r,a))\}$$
• $q_0^{\prime }=E(q_0)$
• $F^{\prime }=\{R\in Q^{\prime }\mid R\cap F\ne \mathrm{\varnothing }\}$

---

States ($Q^{\prime }$):

• $Q^{\prime }=\mathcal{P}(Q)$, the power set of $Q$.
• Each state in $D$ is a subset of the states in $N$.

Transition Function (${\delta }^{\prime }$):

${\delta }^{\prime }:\mathcal{P}(Q)\times \mathrm{\Sigma }\to \mathcal{P}(Q)$, defined as: $$\delta'(R, a) = { q \in Q \mid \exists r \in R, q \in E(\delta(r, a)) }$$

• For each state $R\subseteq Q$ and symbol $a\in \mathrm{\Sigma }$:
• Find all states $q$ reachable by an $a$-move from some $r\in R$, considering $\epsilon$-closures.

Start State ($q_0^{\prime }$):

• $q_0^{\prime }=E(q_0)$, the $\epsilon$-closure of the initial state $q_0$ of $N$.

Final States ($F^{\prime }$):

• $F^{\prime }=\{R\in Q^{\prime }\mid R\cap F\ne \mathrm{\varnothing }\}$.
• A subset $R$ of states is a final state in $D$ if it includes any final state of $N$.

Ensures $D$ captures all possible behaviors of $N$, including $\epsilon$-transitions. Uses the power set of $Q$ as the state space to simulate nondeterminism deterministically.

---

Pseudocode

Algorithm 2 Convert NFA$_\varepsilon$ to DFA

1:Input: An NFA$_\varepsilon$ $N = (Q, \Sigma, \delta, q_0, F)$

2:Output: A DFA $D = (Q', \Sigma, \delta', q_0', F')$

3:$Q' \gets \{E(q_0)\}$, $E(q_0)$ is unmarked, and $\delta' \gets \emptyset$

4:while there exists an unmarked state $R \in Q'$ do

5:Mark $R$

6:for each symbol $a \in \Sigma$ do

7:$U \gets E(\delta(R, a))$

8:if $U \notin Q'$ then

9:Add $U$ to $Q'$

10:end if

11:Add $\delta'(R, a) = U$ to $\delta'$

12:end for

13:end while

14:$q_0' \gets E(q_0)$

15:$F' \gets \{R \mid R \text{ has a final state in } F\}$

16:

17:return $D \gets (Q', \Sigma, \delta', q_0', F')$

Example

Let’s try to convert this $NFA$ to $DFA$.

Initially

Suppose the unmarked states are red.

$$Q^{\prime }=\{E(q_0)=\{q_0,q_2\}\},{\delta }^{\prime }=\{\}$$

Step 1

$E(\delta (\{q_0,q_2\},0))=E(\{q_0\})=\{q_0,q_2\}$$E(\delta (\{q_0,q_2\},1))=E(\{q_1\})=\{q_1\}$

Step 2

$E(\delta (\{q_1\},0))=E(\{q_1,q_2\})=\{q_1,q_2\}$$E(\delta (\{q_1\},1))=E(\{q_2\})=\{q_2\}$

Step 3

$E(\delta (\{q_1,q_2\},0))=E(\{q_0,q_1,q_2\})=\{q_0,q_1,q_2\}$$E(\delta (\{q_1,q_2\},1))=E(\{q_2\})=\{q_2\}$

Step 4

$E(\delta (\{q_2\},0))=E(\{q_0\})=\{q_0,q_2\}$$E(\delta (\{q_2\},1))=E(\{\})=\{\}$

Step 5

$E(\delta (\{q_0,q_1,q_2\},0))=E(\{q_0,q_1,q_2\})=\{q_0,q_1,q_2\}$$E(\delta (\{q_0,q_1,q_2\},1))=E(\{q_1,q_2\})=\{q_1,q_2\}$

$E(\delta (\{\},0))=E(\{\})=\{\}$$E(\delta (\{\},1))=E(\{\})=\{\}$

Finally

$\{q_0,q_2\}$ and $\{q_0,q_1,q_2\}$ are final states.

Rename the states to make the DFA more readable and beautiful: $q_0^{\prime },q_1^{\prime },q_2^{\prime },q_3^{\prime },q_4^{\prime },q_5^{\prime }$.

$q_5^{\prime }$ is a trap state because it has no valid transitions.