This article was accepted into the corpus but its outbound wikilinks were never NER-processed — typical at the deepest BFS hop or when the run's entity cap was reached. No expansion funnel to show.
| AC0 | |
|---|---|
| Name | AC0 |
| Type | Circuit complexity class |
| Defined | 1980s |
| Members | constant-depth polynomial-size Boolean circuits with AND, OR, NOT |
| Notable problems | Parity (not in AC0), Majority (not in AC0) |
AC0
AC0 is the class of decision problems computable by families of Boolean circuits of constant depth and polynomial size using unbounded-fan-in AND, OR, and NOT gates. The class was introduced in the study of Boolean circuit complexity alongside classes such as NC^1, P/poly, and AC^i and has pivotal connections to work by researchers associated with Furst-Saxe-Sipser, Ajtai, Yao, and Håstad. AC0 serves as a baseline for separating complexity classes in results by authors linked to institutions like Princeton University, MIT, and UC Berkeley.
Formally, AC0 consists of languages L for which there exists a family {C_n} of Boolean circuits where each C_n has size polynomial in n, depth bounded by a constant d independent of n, inputs x_1,...,x_n, unbounded-fan-in AND and OR gates, and NOT gates allowed only at the inputs. The uniformity of families is often measured with notions such as DLOGTIME-uniformity, LOGSPACE-uniformity, and P-uniformity as used in complexity comparisons with classes like P, NP, and L. Variants include nonuniform AC0, which corresponds to families without a uniformity constraint and relates to P/poly.
AC0 admits several equivalent descriptions: as constant-depth polynomial-size circuits with unbounded fan-in AND/OR/NOT; as families of formulas when depth is counted differently compared with NC^0 and NC^1; and via descriptive complexity characterizations connecting to fragments of first-order logic with bounded quantifier alternation over ordered structures, paralleling results that relate NL and FO fragments. Low-depth circuit characterizations compare AC0 to subclasses defined by gate types and depth such as AC^0[p], which augments AC0 with MOD_p gates linked to modular counting, studied in work tied to Smolensky and Razborov.
AC0 can compute basic symmetric functions like AND, OR, Majority in restricted forms at constant depth only when depth grows modestly; however, canonical lower bounds show that parity functions such as Parity and MOD_2 are not in AC0. This separation leverages techniques developed in papers by researchers at IBM Research, Microsoft Research, and academic groups at Harvard University and Stanford University. AC0 is properly contained in classes like NC^1 and P/poly; oracle separations and circuit lower bounds connect AC0 to notions in cryptography where hardness assumptions often reference AC0 limitations for constructing pseudorandom generators studied by teams at Bell Labs and I3P.
Measures for AC0 families include size (number of gates), depth (constant d), and fan-in (unbounded). Within the low-depth hierarchy, AC0 is the base level that contrasts with parameterized families like AC^i for i ≥ 1 and NC^i; comparisons use reductions preserving depth and size constraints as in work from Carnegie Mellon University and ETH Zurich. Stratifications of AC0 by depth and size lead to trade-off results: increasing depth allows computing more complex functions, connecting to hierarchies studied by researchers from Rutgers University and Columbia University. Randomized and quantum analogues such as classes studied at Google DeepMind and IBM explore how quantum circuits of constant depth compare to classical AC0.
AC0 lacks natural complete problems under many standard reductions due to its nonuniformity and constant-depth constraint; however, certain promise problems and projection reductions identify AC0-hard instances. Reductions between AC0 languages commonly use AC0-computable projections or NC^0 reductions, techniques analyzed in literature from University of Toronto and University of Cambridge. Complete problem notions are more robust in related classes like AC^0[p] where modular counting problems like MOD_p become central and reductions are studied by groups at Tel Aviv University and University of Illinois Urbana–Champaign.
Proving lower bounds against AC0 has relied on techniques such as the polynomial method (approximating Boolean functions by low-degree polynomials), random restriction arguments exemplified in the Håstad switching lemma, and communication complexity methods that draw on ideas from collaborators at Bell Labs and Microsoft Research. Håstad's switching lemma yields exponential size lower bounds for depth-d circuits computing parity, and Smolensky's algebraic techniques give separations for AC0 augmented with MOD_p gates. Other tools include Fourier analysis on the Boolean cube used in work connected to MIT and Princeton.
AC0 plays a foundational role in theoretical computer science and has applications in finite model theory, circuit design, and proof complexity; it appears in studies linking to SAT solvers, pseudorandomness, and derandomization projects at University of California, San Diego and University of Washington. Connections to classes such as TC^0, NC^1, P/poly, and AC^0[p] inform cryptographic hardness assumptions, learning theory results from Carnegie Mellon University, and descriptive complexity analyses in collaborations with Institute for Advanced Study researchers. Understanding AC0's limits continues to influence research at institutions like Institute for Quantum Computing where comparisons with small-depth quantum circuits remain active.