分割军政府的鲁棒性


16

我们说,一个布尔函数˚F { 0 1 } ñ{ 0 1 }f:{0,1}n{0,1}ķk -junta如果˚Ff最多有ķk干扰因素。

˚F { 0 1 } Ñ{ 0 1 }f:{0,1}n{0,1}是一个2 ķ2k -junta。用x 1x 2x n表示f的变量。修正 S 1 = { x 1x 2x nfx1,x2,,xn2 }S 2 = { x n2 +1xn2 +2xn}

S1={x1,x2,,xn2},S2={xn2+1,xn2+2,,xn}.
显然,存在小号{小号1s ^2}S{S1,S2},使得小号S含有至少ķk的作用变量的˚Ff

现在让ε > 0ϵ>0,并假定˚F { 0 1 } Ñ{ 0 1 }f:{0,1}n{0,1}εϵ -far从每2 ķ2k -junta(即,一个具有改变至少的一小部分εϵ的值的ff使其成为2 k2k -junta)。我们可以对上述陈述做一个“健壮”的版本吗?即是,是否有一个普遍的恒定Çc,和一组小号{ 小号1s ^ 2 }S{S1,S2}使得ffϵc-ϵc远离S中最多包含k个k影响变量的每个函数。S

注意:在问题的原始表述中,cc固定为22。尼尔(Neal)的例子表明,这样的c值是c不够的。但是,由于在属性测试中我们通常不太关心常量,因此我稍微放松了条件。


您可以澄清您的条款吗?除非f的值始终独立于变量,否则变量是否“影响”?“更改f的值”是否意味着针对某个特定x更改值f x 之一?ff(x)x
Neal Young

当然,变量x i影响是否存在n位字符串y,以使f y f y ',其中xinyf(y)f(y) y '是字符串 y,其第 i个坐标被翻转。更改 f的值意味着更改其真值表。yyif

Answers:


17

答案是“是”。证明是矛盾的。

为了符号上的方便,让我们用x表示第一个n / 2变量,第二个n /n/2x 2由变量 ÿ。假设 ˚F X ÿ δ -close到功能 ˚F 1X ÿ 其中仅在取决于n/2yf(x,y)δf1(x,y) x的k坐标的。用 T 1表示其影响坐标。同样,假设 f x y kxT1f(x,y)δ-接近函数 f 2x y ,后者仅取决于 y的k个坐标-接近 2 ķ -junta˚FX ÿ δf2(x,y)ky。用T 2表示其影响坐标。我们需要证明˚F4 δT2f4δ2kf~(x,y)

假设x 1y 1x 2y 2如果x 1(x1,y1)(x2,y2)x1 x 2 T 1的所有坐标上一致,而 y 1 y 2 T 2的所有坐标上一致。我们从每个等效类中随机选择一个代表。让ˉ Xˉ ÿ是用于类的代表X x2T1y1y2T2(x¯,y¯)y 。限定˚F如下: ˚FX ÿ = ˚F ˉ Xˉ ÿ(x,y)f~

f~(x,y)=f(x¯,y¯).

很明显,˚F2 ķf~2k -junta(它仅在变量取决于在Ť 1Ť 2。我们将证明它是在距离4 δ˚F于预期。T1T2)4δf

我们要证明的是 ˚FX Y ^˚FX Y ^ ˚F X Y ^ = ˚F ˉ Xˉ ÿ˚F X Y ^ 4 δ 其中Xÿ是随机统一选择。考虑随机向量

Prf~(Prx,y(f~(x,y)f(x,y)))=Pr(f(x¯,y¯)f(x,y))4δ,
xyX从得到的X通过保持所有位在Ť1和随机翻转所有位不Ť1,和一个矢量ý类似地定义。注意, ˚FXÿ˚FXÿ=˚F ˉ X ˉ ÿ˚FXÿ=x~xT1T1y~˚F Xÿ˚FXÿ
Pr(f~(x,y)f(x,y))=Pr(f(x¯,y¯)f(x,y))=Pr(f(x~,y~)f(x,y)).

我们有, ˚F X Y ^ ˚F XY ^ ˚F X Y ^ ˚F 1X ÿ + ˚F 1X ÿ ˚F 1Xÿ + ˚F1Xÿ ˚F Xÿ δ + 0 + δ = 2 δ

Pr(f(x,y)f(x~,y))Pr(f(x,y)f1(x,y))+Pr(f1(x,y)f1(x~,y))+Pr(f1(x~,y)f(x~,y))δ+0+δ=2δ.

同样,˚F Xÿ ˚F Xÿ2 δ。我们有 ˚F ˉ Xˉ ÿ˚F X Y ^ 4 δ QEDPr(f(x~,y)f(x~,y~))2δ

Pr(f(x¯,y¯)f(x,y))4δ.

容易将这个证明“去随机化”。对于每一个X ÿ ,让˚FX ÿ = 1如果˚F X ÿ = 1对于最(x,y)f~(x,y)=1f(x,y)=1X 'ÿ '在等价类X ÿ ,和˚Fx y = 0,否则。(x,y)(x,y)f~(x,y)=0


12

最小的 界限所保持 c c = 1c2 - 12.41c=1212.41

引理1和2表明,该c的界成立。引理3表明这个界限是紧密的。c

(相比之下,Juri的优雅概率论得出c = 4。c=4

c = 12 -1c=121。引理1给出k = 0的上限。k=0

引理1: 如果fϵ g-S 2中没有影响变量的函数g附近,而fϵ h-S 1中没有影响变量的函数h附近。fϵggS2fϵhhS1,则fϵ-在常数函数附近。 ,其中ε ε + εfϵ ħ/ 2cϵ(ϵg+ϵh)/2c

证明。ϵ为从f到常数函数的距离。矛盾的是,ϵ不满足要求的不等式。令y = x 1x 2x n / 2z = x n / 2ϵfϵy=(x1,x2,,xn/2) + 1 x n 并将 f g h记 f yz=(xn/2+1,,xn)fghz g独立于 yf(y,z)y z h y z ,因此 g y z 独立于 z h y zg(y,z)h(y,z)g(y,z)zh(y,z)y

(我发现将f可视化为带有顶点集{ y }的完整二部图的边缘标签和f{y} { z },其中 g给出了 { y } h的顶点标签{z}g{y}h给出的顶点标记{ z }。){z}

g 0y z 对的分数,使得g y z = 0。令g 1 = 1 - g 0是对的分数,使得g y z =g0(y,z)g(y,z)=0g1=1g0 1。同样,令 h 0为对的分数,使得 h y z = 0,令 h 1g(y,z)=1h0h(y,z)=0h1为对的分数,使hy z = 1h(y,z)=1

在不失一般性的前提下,假定对于g y z = h y z )的任何对,它还认为f y z = g y z = h y z)。(否则,切换 f y z )的值可使我们将ϵ gϵ h都减少1。g(y,z)=h(y,z)f(y,z)=g(y,z)=h(y,z)f(y,z)ϵgϵh / 2 n1/2n,同时减小ε由至多1 / 2 Ñ,因此所得到的功能仍然是一个反例。)说任何这样的一对是``在协议“”。ϵ1/2n

fg的距离加上从fh的距离 是x y 对不一致的分数。即,ϵ g + ϵ h = g 0 h 1 + g 1 h 0fgfh(x,y)ϵg+ϵh=g0h1+g1h0

f到全零函数的距离最大为1 g 0f h 01g0h0

f到all-ones函数的距离最大为1 g 1f ħ 11g1h1

此外,从距离˚F到最近的常数函数为至多1 / 2f1/2

因此,该比率ε /ε + ε ħ为至多 分钟1 / 2 1 - 0 ħ 01 - 1 ħ 1ϵ/(ϵg+ϵh)0 ħ 1个 + 1 ħ 0 其中0ħ0[01]1=1-0ħ1=1-H ^0

min(1/2,1g0h0,1g1h1)g0h1+g1h0,
g0,h0[0,1]g1=1g0h1=1h0

经计算,该比率最多为 12 2 -1=c ^/2。QED12(21)=c/2

Lemma 2 extends Lemma 1 to general kk by arguing pointwise, over every possible setting of the 2k2k influencing variables. Recall that c=121c=121.

Lemma 2: Fix any kk. If ff is ϵgϵg-near a function gg that has kk influencing variables in S2S2, and ff is ϵhϵh-near a function hh that has kk influencing variables in S1S1, then ff is ϵϵ-near a function ˆff^ that has at most 2k2k influencing variables, where ϵ(ϵg+ϵh)/2cϵ(ϵg+ϵh)/2c.

Proof. Express ff as f(a,y,b,z)f(a,y,b,z) where (a,y)(a,y) contains the variables in S1S1 with aa containing those that influence hh, while (b,z)(b,z) contains the variables in S2S2 with bb containing those influencing gg. So g(a,y,b,z)g(a,y,b,z) is independent of zz, and h(a,y,b,z)h(a,y,b,z) is independent of yy.

For each fixed value of aa and bb, define Fab(y,z)=f(a,y,b,z)Fab(y,z)=f(a,y,b,z), and define GabGab and HabHab similarly from gg and hh respectively. Let ϵgabϵgab be the distance from FabFab to GabGab (restricted to (y,z)(y,z) pairs). Likewise let ϵhabϵhab be the distance from FabFab to HabHab.

By Lemma 1, there exists a constant cabcab such that the distance (call it ϵabϵab) from FabFab to the constant function cabcab is at most (ϵhab+ϵgab)/(2c)(ϵhab+ϵgab)/(2c). Define ˆf(a,y,b,z)=cabf^(a,y,b,z)=cab.

Clearly ˆff^ depends only on aa and bb (and thus at most kk variables).

Let ϵˆfϵf^ be the average, over the (a,b)(a,b) pairs, of the ϵabϵab's, so that the distance from ff to ˆff^ is ϵˆfϵf^.

Likewise, the distances from ff to gg and from ff to hh (that is, ϵgϵg and ϵh)ϵh) are the averages, over the (a,b)(a,b) pairs, of, respectively, ϵgabϵgab and ϵhabϵhab.

Since ϵab(ϵhab+ϵgab)/(2c)ϵab(ϵhab+ϵgab)/(2c) for all a,ba,b, it follows that ϵˆf(ϵg+ϵh)/(2c)ϵf^(ϵg+ϵh)/(2c). QED

Lemma 3 shows that the constant cc above is the best you can hope for (even for k=0k=0 and ϵ=0.5ϵ=0.5).

Lemma 3: There exists ff such that ff is (0.5/c)(0.5/c)-near two functions gg and hh, where gg has no influencing variables in S2S2 and hh has no influencing variables in S1S1, and ff is 0.50.5-far from every constant function.

Proof. Let y and z be x restricted to, respectively, S1 and S2. That is, y=(x1,,xn/2) and z=(xn/2+1,,xn).

Identify each possible y with a unique element of [N], where N=2n/2. Likewise, identify each possible z with a unique element of [N]. Thus, we think of f as a function from [N]×[N] to {0,1}.

Define f(y,z) to be 1 iff max(y,z)12N.

By calculation, the fraction of f's values that are zero is (12)2=12, so both constant functions have distance 12 to f.

Define g(y,z) to be 1 iff y12N. Then g has no influencing variables in S2. The distance from f to g is the fraction of pairs (y,z) such that y<12N and z12N. By calculation, this is at most 12(112)=0.5/c

Similarly, the distance from f to h, where h(y,z)=1 iff z12N, is at most 0.5/c.

QED


First of all, thanks Neal! This indeed sums it up for k=0, and sheds some light on the general problem. However in the case of k=0 the problem is a bit degenerate (as 2k=k), so I'm more curious regarding the case of k1. I didn't manage to extend this claim for k>0, so if you have an idea on how to do it - I'd appreciate it. If it simplifies the problem, then the exact constants are not crucial; that is, ϵ/2-far can be replaced by ϵ/c-far, for some universal constant c.

2
I've edited it to add the extension to general k. And Yuri's argument below gives a slightly looser factor with an elegant probabilistic argument.
Neal Young

Sincere thanks Neal! This line of reasoning is quite enlightening.
By using our site, you acknowledge that you have read and understand our Cookie Policy and Privacy Policy.
Licensed under cc by-sa 3.0 with attribution required.