Abstract:We solve the derandomized direct product testing question in the low acceptance regime, by constructing new high dimensional expanders that have no small connected covers. We show that our complexes have swap cocycle expansion, which allows us to deduce the agreement theorem by relying on previous work.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the effectiveness of derandomized direct product testing in the low acceptance regime. Specifically, the researchers hope to construct a new type of high - dimensional expanders (HDXs) that can support agreement testing under the condition of low acceptance rate. Agreement testing is a method to verify the effectiveness of function encoding. It is achieved by selecting two random sets \(s_1\) and \(s_2\) and checking whether the local functions on their intersection are consistent. If the probability of consistency of these local functions on the intersection is greater than a small positive number \(\varepsilon\), it is hoped to find a global function \(G\) such that most local functions \(f_s\) are approximately equal to \(G\) on their domains.
The main contributions of the paper are as follows:
1. **Constructing new high - dimensional expanders**: The author proposes a new construction method based on bounded - degree symplectic HDXs. These expanders do not have small connected covers, so they can support agreement testing with a low acceptance rate.
2. **Proving the agreement theorem**: By using previous work, the author proves that their complexes have swap cocycle expansion, which enables them to derive the agreement theorem, that is, under the condition of low acceptance rate, there exists a global function \(G\) such that most local functions \(f_s\) are approximately equal to \(G\).
3. **Algorithm implementation**: The author also provides a polynomial - time algorithm to construct such symplectic high - dimensional expanders.
### Specific problem description
In low - acceptance - rate agreement testing, given a set of local functions \(\{f_s:s\rightarrow\Sigma\mid s\in X\}\), where \(X\) is a \(k\)-subset family and each \(f_s\) is defined on a subset \(s\subset [N]\). The process of agreement testing is to randomly select a pair of sets \(s_1, s_2\in X\), these two sets intersect on \(\sqrt{k}\) elements, and check whether the values of \(f_{s_1}\) and \(f_{s_2}\) on the intersection are consistent. If the consistency probability \(\text{Agree}(\{f_s\})>\varepsilon\), it is hoped to find a global function \(G:[N]\rightarrow\Sigma\) such that for most \(s\in X\), \(f_s\) is approximately equal to \(G\) on \(s\).
### Main results of the paper
The main results of the paper are as follows:
- **Theorem 1.1**: For any \(\varepsilon > 0\), there exist a constant \(c>0\) and a large enough integer \(k < g\) and a prime number \(p\), such that there exists an infinite family of constant - degree connected \(g + 1\)-partite simplicial complexes \(\{X_N\}_N\), these complexes are the finite quotient spaces of \(\tilde{C}_g\), where \(\tilde{C}_g\) is the affine Bruhat - Tits building related to the symplectic group \(Sp(2g,\mathbb{Q}_p)\). These complexes have \(N\) vertices and \(O_g(N)\) faces, and satisfy the following property:
\[
\text{Agree}(\{f_s\})>\varepsilon\implies\exists G:X^{(0)}\to\Sigma,\quad P_s[f_s\approx G|_s]\geq\varepsilon^c.
\]
- **Theorem 1**