Venturi flume

From formulasearchengine
Revision as of 17:20, 21 April 2013 by en>CommonsDelinker (Removing "Parshall_Flume.png", it has been deleted from Commons by JuTa because: No permission since 13 April 2013.)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

My name is Jestine (34 years old) and my hobbies are Origami and Microscopy.

Here is my web site; http://Www.hostgator1centcoupon.info/ (support.file1.com)

For a class of predicates H defined on a set X and a set of samples x=(x1,x2,,xm), where xiX, the empirical frequency of hH on x is Qx^(h)=1m|{i:1im,h(xi)=1}|. The Uniform Convergence Theorem states, roughly,that if H is "simple" and we draw samples independently (with replacement) from X according to a distribution P, then with high probability all the empirical frequency will be close to its expectation, where the expectation is given by QP(h)=P{yX:h(y)=1}. Here "simple" means that the Vapnik-Chernovenkis dimension of the class H is small relative to the size of the sample.
In other words, a sufficiently simple collection of functions behaves roughly the same on a small random sample as it does on the distribution as a whole.

Uniform convergence theorem statement[1]

If H is a set of {0,1}-valued functions defined on a set X and P is a probability distribution on X then for ϵ>0 and m a positive integer, we have,

Pm{|QP(h)Qx^(h)|ϵ for some hH}4ΠH(2m)eϵ2m8.

where, for any xXm,

QP(h)=P{(yX:h(y)=1},
Qx^(h)=1m|{i:1im,h(xi)=1}| and |x|=m. Pm indicates that the probability is taken over x consisting of m i.i.d. draws from the distribution P.

ΠH is defined as: For any {0,1}-valued functions H over X and DX,

ΠH(D)={hD:hH}.

And for any natural number m the shattering number ΠH(m) is defined as.

ΠH(m)=max|{hD:|D|=m,hH}|.

From the point of Learning Theory one can consider H to be the Concept/Hypothesis class defined over the instance set X. Before getting into the details of the proof of the theorem we will state Sauer's Lemma which we will need in our proof.

Sauer–Shelah lemma

The Sauer–Shelah lemma[2] relates the shattering number Πh(m) to the VC Dimension.

Lemma: ΠH(m)(emd)d, where d is the VC Dimension of the concept class H.

Corollary: ΠH(m)md.

Proof of uniform convergence theorem [1]

Before we get into the details of the proof of the Uniform Convergence Theorem we will present a high level overview of the proof.

  1. Symmetrization: We transform the problem of analyzing |QP(h)Q^x(h)|ϵ into the problem of analyzing |Q^r(h)Q^s(h)|ϵ/2, where r and s are i.i.d samples of size m drawn according to the distribution P. One can view r as the original randomly drawn sample of length m, while s may be thought as the testing sample which is used to estimate QP(h).
  2. Permutation: Since r and s are picked identically and independently, so swapping elements between them will not change the probability distribution on r and s. So, we will try to bound the probability of |Q^r(h)Q^s(h)|ϵ/2 for some hH by considering the effect of a specific collection of permutations of the joint sample x=r||s. Specifically, we consider permutations σ(x) which swap xi and xm+i in some subset of 1,2,...,m. The symbol r||s means the concatenation of r and s.
  3. Reduction to a finite class: We can now restrict the function class H to a fixed joint sample and hence, if H has finite VC Dimension, it reduces to the problem to one involving a finite function class.

We present the technical details of the proof.

Symmetrization

Lemma: Let V={xXm:|QP(h)Qx^(h)|ϵ for some hH} and

R={(r,s)Xm×Xm:|Qr^(h)Qs^(h)|ϵ/2 for some hH}.

Then for m2ϵ2, Pm(V)2P2m(R).

Proof: By the triangle inequality,
if |QP(h)Qr^(h)|ϵ and |QP(h)Qs^(h)|ϵ/2 then |Qr^(h)Qs^(h)|ϵ/2.
Therefore,

P2m(R)P2m{hH,|QP(h)Qr^(h)|ϵ and |QP(h)Qs^(h)|ϵ/2}
=VPm{s:hH,|QP(h)Qr^(h)|ϵ and |QP(h)Qs^(h)|ϵ/2}dPm(r)=A [since r and s are independent].

Now for rV fix an hH such that |QP(h)Qr^(h)|ϵ. For this h, we shall show that

Pm{|QP(h)Qs^(h)|ϵ2}12.

Thus for any rV, APm(V)2 and hence P2m(R)Pm(V)2. And hence we perform the first step of our high level idea.
Notice, mQs^(h) is a binomial random variable with expectation mQP(h) and variance mQP(h)(1QP(h)). By Chebyshev's inequality we get,

Pm{|QP(h)Qs(h)^|>ϵ2}mQP(h)(1QP(h))(ϵm/2)21ϵ2m12 for the mentioned bound on m. Here we use the fact that x(1x)1/4 for x.

Permutations

Let Γm be the set of all permutations of {1,2,3,,2m} that swaps i and m+i i in some subset of {1,2,3,...,2m}.

Lemma: Let R be any subset of X2m and P any probability distribution on X. Then,

P2m(R)=E[Pr[σ(x)R]]maxxX2m(Pr[σ(x)R]),

where the expectation is over x chosen according to P2m, and the probability is over σ chosen uniformly from Γm.

Proof: For any σΓm,

P2m(R)=P2m{x:σ(x)R} [since coordinate permutations preserve the product distribution P2m].
P2m(R)=X2m1R(x)dP2m(x)
=1|Γm|σΓmX2m1R(σ(x))dP2m(x)
=X2m1|Γm|σΓm1R(σ(x))dP2m(x) [Because |Γm| is finite]
=X2mPr[σ(x)R]dP2m(x) [The expectation]
maxxX2m(Pr[σ(x)R]).

The maximum is guaranteed to exist since there is only a finite set of values that probability under a random permutation can take.

Reduction to a finite class

Lemma: Basing on the previous lemma,

maxxX2m(Pr[σ(x)R])4ΠH(2m)eϵ2m8.

Proof: Let us define x=(x1,x2,...,x2m) and t=|H|x| which is atmost ΠH(2m). This means there are functions h1,h2,...,htH such that for any hH,i between 1 and t with hi(xk)=h(xk) for 1k2m.
We see that σ(x)R iff for some h in H satisfies, |1m|{1im:h(xσi)=1}|1m|{m+1i2m:h(xσi)=1}||ϵ2. Hence if we define wij=1 if hj(xi)=1 and wij=0 otherwise.
For 1im and 1jt, we have that σ(x)R iff for some j in 1,...,t satisfies |1m(iwσ(i)jiwσ(m+i)j)|ϵ2. By union bound we get,

Pr[σ(x)R]tmax(Pr[|1m(iwσijiwσm+ij)|ϵ2])
ΠH(2m)max(Pr[|1m(iwσijiwσm+ij)|ϵ2]).

Since, the distribution over the permutations σ is uniform for each i, so wσijwσm+ij equals ±|wijwm+ij|, with equal probability.
Thus,

Pr[|1m(i(wσijwσm+ij))|ϵ2]=Pr[|1m(i|wijwm+ij|βi)|ϵ2],

where the probability on the right is over βi and both the possibilities are equally likely. By Hoeffding's inequality, this is at most 2emϵ28.

Finally, combining all the three parts of the proof we get the Uniform Convergence Theorem.

References

43 year old Petroleum Engineer Harry from Deep River, usually spends time with hobbies and interests like renting movies, property developers in singapore new condominium and vehicle racing. Constantly enjoys going to destinations like Camino Real de Tierra Adentro.