AND testing and robust judgement aggregation

AND testing and robust judgement aggregation
复制标题

AND 测试和稳健的判断聚合

DOI:
10.1145/3357713.3384254
复制
发表时间:
2020
期刊:
STOC 2020: Proceedings of the 52nd Annual ACM SIGACT Symposium on Theory of Computing
影响因子:
--
通讯作者:
Mossel, E.
Mossel, E.
中科院分区:
--
文献类型:
--
作者:
Filmus, Y;Lifshitz, N;Minzer, D;Mossel, E.

文献摘要

相似文献

称函数f ∶{0,1}n→ {0,1}是一个近似与-同态,如果随机一致地逼近x,y∈ n,我们有f(x <$y)=f(x)<$f(y),概率至少为1-ε,其中x <$y=(x1 <$y1,...,xn <$yn).证明了当ε→ 0时,当δ(ε)→ 0时,当f → {0,1}n→ {0,1}是一个这改进了Nehama的一个结果,他证明了一个类似的陈述,其中δ依赖于n。我们的定理暗示了计算社会选择中判断聚集的一个强结果。在社会选择的语言中,我们的结果表明,如果ffi是ε-接近于满足的判断聚集,那么它是δ(ε)-接近于寡头(社会选择理论中AND函数的名称)。这改进了Nehama的结果,其中δ随n多项式衰减.我们的结果是从一个更一般的结果得到的,其中我们刻画了特征值方程f = λg的近似解,其中是向下噪声算子f(x)=y[f(x <$y)],f是[0,1]-值,gis是{0,1}-值.我们确定了该方程的所有精确解,并证明了f和λ g接近的任何近似解都接近于精确解。
A functionf∶{0,1}n→ {0,1} is called an approximate AND-homomorphism if choosingx,y∈nuniformly at random, we have thatf(x∧y) =f(x)∧f(y) with probability at least 1−ε, wherex∧y= (x1∧y1,…,xn∧yn). We prove that iff∶ {0,1}n→ {0,1} is an approximate AND-homomorphism, thenfis δ-close to either a constant function or an AND function, where δ(ε) → 0 as ε→ 0. This improves on a result of Nehama, who proved a similar statement in which δ depends onn.Our theorem implies a strong result on judgement aggregation in computational social choice. In the language of social choice, our result shows that iffis ε-close to satisfying judgement aggregation, then it is δ(ε)-close to an oligarchy (the name for the AND function in social choice theory). This improves on Nehama’s result, in which δ decays polynomially withn.Our result follows from a more general one, in which we characterize approximate solutions to the eigenvalue equationf= λg, where is the downwards noise operatorf(x) =y[f(x∧y)],fis [0,1]-valued, andgis {0,1}-valued. We identify all exact solutions to this equation, and show that any approximate solution in whichfand λgare close is close to an exact solution.