מכונת טיורינג
A computation model with tapes, read/write heads, and a control state.
Terminology is kept consistent across the entire course.
A computation model with tapes, read/write heads, and a control state.
A machine that must halt on every input and return accept or reject.
A complete instantaneous description of a machine run: state, tape contents, and head positions.
A graph whose vertices are configurations and whose edges are single legal transitions.
A representation of instances, machines, or numbers as strings over a finite alphabet.
A running-time bound by a polynomial in the encoded input length.
The class of languages decided by deterministic machines in $O(t(n))$ time.
The class of languages decided by deterministic machines using $O(s(n))$ workspace.
The union of all polynomial DTIME classes; languages decidable in polynomial time.
Decision problems solvable with polynomially bounded workspace.
A result showing that more resource (time or space) enables deciding strictly more languages.
Inflating input length without changing the encoded decision.
A machine that accepts if at least one computation branch accepts.
A tree whose nodes are configurations and whose children represent legal transition choices.
A short string that lets a verifier confirm a yes instance.
A polynomially bounded witness attached to an instance and checked by a verifier.
A polynomial-time machine checking an instance-certificate pair.
Languages with polynomially bounded certificates verifiable in polynomial time.
Languages whose complements are in NP; no instances are proved by short certificates.
A polynomial-time map $f$ such that $x\in A\iff f(x)\in B$.
A computable transformation preserving membership.
A standard, well-known reduction in the problem network, e.g. from 3SAT to IS.
The central NP-complete family (SAT, IS, CLIQUE, TSP, etc.) and the reduction network linking them.
If $A\le_m^p B$ and $B\le_m^p C$, then $A\le_m^p C$; composing reductions.
If $A\le_m^p B$ and $B\in\mathrm P$, then $A\in\mathrm P$; same idea for NP and coNP.
At least as hard as every language in NP under the stated reduction; need not be in NP.
A problem in NP to which every NP problem reduces.
SAT is NP-complete; every NP language reduces to it via a computation tableau.
A matrix whose rows are full configurations of a machine on a given input.
A subcomponent in a reduction encoding a local constraint (clause, variable, edge).
A two-direction proof: yes in the target $\iff$ yes in the source.
A Boolean formula as a product of clauses (ORs of literals).
SAT where every clause has exactly three literals.
In a graph: a set of vertices with no edges between them.
In a graph: a set of vertices pairwise adjacent.
A vertex set touching every edge at least once.
A graph with the same vertices and inverted edges: $\{u,v\}\in E(\overline G)\iff\{u,v\}\notin E(G)$.
A decision problem returns one bit; an optimization problem asks for the solution or objective value.
Polynomial in numeric values but exponential in binary encoding length.
Hardness that persists even when numeric values are polynomially bounded.
A bound comparing an algorithmic solution with the optimum.
A method that works well in practice but lacks approximation or optimality guarantees.
A machine with access to random bits; after choosing $r$ the run is deterministic.
A randomized algorithm with fixed runtime that may err with bounded probability.
A randomized algorithm always correct; runtime is random with bounded expectation.
Polynomial-time randomized decision with bounded two-sided error.
Reducing error probability by repeating a randomized algorithm.
Languages decidable in $O(\log n)$ workspace.
Languages decided nondeterministically in logarithmic space.
Reachability: is there a path from $s$ to $t$ in a directed graph?
A reduction computing output without storing it fully; a logarithmic-space transducer.
A machine that prints output while staying within a space bound.
A graph not stored explicitly; adjacency is generated or checked on demand.
$\mathrm{NSPACE}(s(n))\subseteq\mathrm{DSPACE}(s(n)^2)$; in particular $\mathrm{NL}\subseteq\mathrm L$.
A black-box language query answered in one computation step.
Classes $\Sigma_k^P$ and $\Pi_k^P$ with $k$ alternating blocks of existential and universal quantifiers.
A proof technique valid relative to every oracle; it limits separation methods.
Languages decidable by non-uniform polynomial-size circuits.
A network of AND, OR, NOT gates computing a Boolean function.