International Conference on Algorithmic Learning Theory（ALT）是由算法学习理论协会（AALT），和其他相关活动一起来推广学习理论。官网链接：http://alt2019.algorithmiclearningtheory.org/

### 最新论文

$\newcommand{\F}{\mathbb{F}}$We study the Boolean function parameters sensitivity ($s$), block sensitivity ($bs$), and alternation ($alt$) under specially designed affine transforms. For a function $f:\F_2^n\to \{0,1\}$, and $A=Mx+b$ for $M \in \F_2^{n\times n}$ and $b\in \F_2^n$, the result of the transformation $g$ is defined as $\forall x\in\F_2^n, g(x)=f(Mx+b)$. We study alternation under linear shifts ($M$ is the identity matrix) called the shift invariant alternation (denoted by $salt(f)$). We exhibit an explicit family of functions for which $salt(f)$ is $2^{\Omega(s(f))}$. We show an affine transform $A$, such that the corresponding function $g$ satisfies $bs(f,0^n) \le s(g)$, using which we proving that for $F(x,y)=f(x\land y)$, the bounded error quantum communication complexity of $F$ with prior entanglement, $Q^*_{1/3}(F)=\Omega(\sqrt{bs(f,0^n)})$. Our proof builds on ideas from Sherstov (2010) where we use specific properties of the above affine transformation. We show, * For a prime $p$ and $0<\epsilon<1$, any $f$ with $deg_p(f)\le(1-\epsilon)\log n$ must satisfy $Q^*_{1/3}(F) = \Omega(\frac{n^{\epsilon/2}}{\log n})$. Here, $deg_p(f)$ denotes the degree of the multilinear polynomial of $f$ over $\F_p$. * For any $f$ such that there exists primes $p$ and $q$ with $deg_q(f) \ge \Omega(deg_p(f)^\delta)$ for $\delta > 2$, the deterministic communication complexity - $D(F)$ and $Q^*_{1/3}(F)$ are polynomially related. In particular, this holds when $deg_p(f) = O(1)$. Thus, for this class of functions, this answers an open question (see Buhrman and deWolf (2001)) about the relation between the two measures. We construct linear transformation $A$, such that $g$ satisfies, $alt(f) \le 2s(g)+1$. Using this, we exhibit a family of Boolean functions that rule out a potential approach to settle the XOR Log-Rank conjecture via a proof of Sensitivity conjecture [Hao Huang (2019)].

Top