Refuting approaches to the log-rank conjecture for XOR functions

Hamed Hatami,Kaave Hosseini,Shachar Lovett,Anthony Ostuni

2024-05-03

Abstract:The log-rank conjecture, a longstanding problem in communication complexity, has persistently eluded resolution for decades. Consequently, some recent efforts have focused on potential approaches for establishing the conjecture in the special case of XOR functions, where the communication matrix is lifted from a boolean function, and the rank of the matrix equals the Fourier sparsity of the function, which is the number of its nonzero Fourier coefficients. In this note, we refute two conjectures. The first has origins in Montanaro and Osborne (arXiv'09) and is considered in Tsang et al. (FOCS'13), and the second one is due to Mande and Sanyal (FSTTCS'20). These conjectures were proposed in order to improve the best-known bound of Lovett (STOC'14) regarding the log-rank conjecture in the special case of XOR functions. Both conjectures speculate that the set of nonzero Fourier coefficients of the boolean function has some strong additive structure. We refute these conjectures by constructing two specific boolean functions tailored to each.

Computational Complexity

What problem does this paper attempt to address?

The problem that this paper attempts to solve concerns two conjectures regarding the log - rank conjecture in communication complexity in the special case of XOR functions. Specifically, the main objective of the paper is to refute two conjectures proposed by Montanaro and Osborne, and Mande and Sanyal. These conjectures aim to improve Lovett's best - known bound on the log - rank conjecture in the case of XOR functions.

### Background
The log - rank conjecture is a long - standing unsolved problem in communication complexity. It states that the deterministic communication complexity of a two - party Boolean function is upper - bounded by a polylogarithm of the matrix rank of the function. That is:
\[ \text{CC}(f) \leq \text{polylog}(\text{rank}(f)) \]

Here, \(\text{CC}(f)\) is the communication complexity of \(f\), and \(\text{rank}(f)\) is the rank over the real numbers of the Boolean matrix corresponding to \(f\).

### XOR functions
In the case of XOR functions, the communication matrix is lifted from a Boolean function, and the rank of the matrix is equal to the Fourier sparsity of the function, that is, the number of non - zero Fourier coefficients. Therefore, the log - rank conjecture in the case of XOR functions can be stated as:
\[ \text{PDT}(f) \leq \text{polylog}(|\text{supp}(\hat{f})|) \]

Here, \(\text{PDT}(f)\) is the parity decision tree complexity of \(f\), and \(\text{supp}(\hat{f})\) is the Fourier support set of \(f\).

### Known results
The best - known bound currently is:
\[ \text{PDT}(f) \leq O(\sqrt{|\text{supp}(\hat{f})|}) \]

### Main contributions of the paper
1. **Refuting the conjecture of Montanaro and Osborne**:
- They proposed that there exists a constant \(K>1\) such that for any distinct \(\gamma_1,\gamma_2\), we have:
\[ |(S + \gamma_1)\cap(S + \gamma_2)|\geq\frac{|S|}{K} \]
- The paper proves that this conjecture is not valid by constructing a specific Boolean function and gives a more precise bound:
\[ |(S + \gamma_1)\cap(S + \gamma_2)|\leq O(|S|^{5/6}) \]

2. **Refuting the conjecture of Mande and Sanyal**:
- They proposed that there exist constants \(\alpha,\beta\in(0,1)\) such that for any Boolean function \(f\), we have:
\[ \Pr_{\gamma_1,\gamma_2\in S}\left[|(S + \gamma_1)\cap(S + \gamma_2)| > |S|^\beta\right]\geq\alpha \]
- The paper proves that this conjecture is also not valid by constructing a specific Boolean function and gives a more precise bound:
\[ \Pr_{\gamma_1,\gamma_2\in S}\left[|(S + \gamma_1)\cap(S + \gamma_2)| > k\right]=O\left(\frac{1}{k}\right) \]

### Conclusion
Although these specific constructed functions rule out some specific methods, they are not counterexamples to the log - rank conjecture. The results of the paper show that even in the special case of XOR functions, the log - rank conjecture is still not completely solved, and further research is needed to find new methods or stronger counterexamples.