A Review of Some Recent Lower Bounds Against Low-Depth Threshold Circuits
summary
Presents concrete lower bounds: an explicit Andreev-style function needs superlinear gates and roughly n^{1.5}/polylog n size against depth‑3 TC0, proven by a random‑restriction lemma that makes an LTF constant with probability 1 − O(K/√n) when only K inputs remain, and a counting lower bound on the restricted function. Develops probabilistic polynomials and probabilistic PTFs (degree ~√(D log S) or improved cube‑root tradeoffs), yields faster SAT for ACC-of-thresholds with a sub‑quadratic bottom layer and an E^NP circuit lower bound.
AI-written summary · about
chapters (13)
- 0:00Start
- 0:50Outline
- 2:34Introduction
- 3:32Depth-Two LTF Circuits
- 6:21Lower Bounds for Explicit Functions
- 7:59Thm 1: Andreev's Function A'87
- 10:01Thm 1: Andreev's Function [A'87]
- 11:28THR-THR Lower Bound for Andreev?
- 14:21Random Restrictions to LTFs
- 15:57Let's take a moment to adjust our eyes...
- 16:17Probabilistic Polynomials for Functions
- 21:41Applications for Circuit Complexity
- 29:09Open Problems Tight upper and lower bounds for Andreev?
