Nested stack automata Source: en.wikipedia.org/wiki/Nested_stack_automata

In automata theory, a nested stack automaton is a finite automaton that can make use of a stack containing data which can be additional stacks.^[1] Like a stack automaton, a nested stack automaton may step up or down in the stack, and read the current symbol; in addition, it may at any place create a new stack, operate on that one, eventually destroy it, and continue operating on the old stack. This way, stacks can be nested recursively to an arbitrary depth; however, the automaton always operates on the innermost stack only.

A nested stack automaton is capable of recognizing an indexed language,^[2] and in fact the class of indexed languages is exactly the class of languages accepted by one-way nondeterministic nested stack automata.^[1][3]

Nested stack automata should not be confused with embedded pushdown automata, which have less computational power.^{[citation needed]}

A (nondeterministic two-way) nested stack automaton is a tuple ⟨Q,Σ,Γ,δ,q₀,Z₀,F,[,],]⟩ where

• Q, Σ, and Γ is a nonempty finite set of states, input symbols, and stack symbols, respectively,
• [, ], and ] are distinct special symbols not contained in Σ ∪ Γ,
  • [ is used as left endmarker for both the input string and a (sub)stack string,
  • ] is used as right endmarker for these strings,
  • ] is used as the final endmarker of the string denoting the whole stack.^{[note 1]}
• An extended input alphabet is defined by Σ' = Σ ∪ {[,]}, an extended stack alphabet by Γ' = Γ ∪ {]}, and the set of input move directions by D = {-1,0,+1}.
• δ, the finite control, is a mapping from Q × Σ' × (Γ' ∪ [Γ' ∪ {], []}) into finite subsets of Q × D × ([Γ^* ∪ D), such that δ maps^{[note 2]}

Informally, the top symbol of a (sub)stack together with its preceding left endmarker "[" is viewed as a single symbol;^[4] then δ reads

• the current state,
• the current input symbol, and
• the current stack symbol,

and outputs

• the next state,
• the direction in which to move on the input, and
• the direction in which to move on the stack, or the string of symbols to replace the topmost stack symbol.

• q₀ ∈ Q is the initial state,
• Z₀ ∈ Γ is the initial stack symbol,
• F ⊆ Q is the set of final states.

A configuration, or instantaneous description of such an automaton consists in a triple ⟨ q, [a₁a₂...a_i...a_n-1], [Z₁X₂...X_j...X_m-1] ⟩, where

• q ∈ Q is the current state,
• [a₁a₂...a_i...a_n-1] is the input string; for convenience, a₀ = [ and a_n = ] is defined^{[note 3]} The current position in the input, viz. i with 0 ≤ i ≤ n, is marked by underlining the respective symbol.
• [Z₁X₂...X_j...X_m-1] is the stack, including substacks; for convenience, X₁ = [Z₁ ^{[note 4]} and X_m = ] is defined. The current position in the stack, viz. j with 1 ≤ j ≤ m, is marked by underlining the respective symbol.

An example run (input string not shown):

When automata are allowed to re-read their input ("two-way automata"), nested stacks do not result in additional language recognition capabilities, compared to plain stacks.^[5]

Gilman and Shapiro used nested stack automata to solve the word problem in virtually free groups, similarly to the Muller–Schupp theorem.^[6]