Explicit depth-three circuit lower bounds beyond the square-root exponent
Theorem 1.1: there is a language L decidable by a deterministic Turing machine in time C(n+1)a whose n-bit membership functions satisfy log2S3(fn)/n→∞, counting all gates of OR-AND-OR circuits with no fan-in restriction. The language and machine are fixed before the constant A. It does not give a fixed exponent n1/2+ε (Hastad-Jukna-Pudlak's problem stays open), says nothing about AND-OR-AND circuits beyond what duality gives for the complement, and nothing about depth above three or Valiant's 2ω(n/loglogn) threshold.
Theorem 1.1: there is a language L decidable by a deterministic Turing machine in time C(n+1)a whose n-bit membership functions satisfy log2S3(fn)/n→∞, counting all gates of OR-AND-OR circuits with no fan-in restriction. The language and machine are fixed before the constant A. It does not give a fixed exponent n1/2+ε (Hastad-Jukna-Pudlak's problem stays open), says nothin…
Theorem 1.1: there is a language L decidable by a deterministic Turing machine in time C(n+1)a whose n-bit membership functions satisfy log2S3(fn)/n→∞, counting all gates of OR-AND-OR circuits with no fan-in restriction. The language and machine are fixed before the constant A. It does not give a fixed exponent n1/2+ε (Hastad-Jukna-Pudlak's problem stays open), says nothing about AND-OR-AND circuits beyond what duality gives for the complement, and nothing about depth above three or Valiant's 2ω(n/loglogn) threshold.
Theorem 1.1: there is a language L decidable by a deterministic Turing machine in time C(n+1)a whose n-bit membership functions satisfy log2S3(fn)/n→∞, counting all gates of OR-AND-OR circuits with no fan-in restriction. The language and machine are fixed before the constant A. It does not give a fixed exponent n1/2+ε (Hastad-Jukna-Pudlak's problem stays open), says nothing about AND-OR-AND circuits beyond what duality gives for the complement, and nothing about depth above three or Valiant's 2ω(n/loglogn) threshold.