Generation of Quasi-Uniform Distributions and Verification of The Tightness of The Inner Bound
Generation of Quasi-Uniform Distributions and Verification of The Tightness of The Inner Bound
Dauood Saleem
1 / 18
Outline
• Introduction
• Background
• Research Problem
• References
2 / 18
Introduction
3 / 18
Introduction
1
R. W. Yeung, “A framework for linear information inequalities,” IEEE
Trans. Inform. Theory, vol. 43, pp. 1924–1934, Nov. 1997.
2
Z. Zhang and R. W. Yeung, “A non-Shannon-type conditional inequality of
information quantities,” IEEE Trans. Inform. Theory, vol. 43, pp. 1982–1986,
Nov. 1997.
4 / 18
Background
• X [n] is a quasi-uniform random vector if for all ∅ =
6 α ⊆ [n],
the random vectors X α are uniformly distributed over their
respective supports Sα where,
Sα , {xα : Pr {X α = xα } > 0}
Ψn , {h ∈ Hn : h is quasi-uniform.}
5 / 18
Research Problem
6 / 18
Work Done and Target Set for Last Semester
Require: d, X[n]
1: i ← 1, p ← 0|X
[n] |×1
, m ← |X[n] |
2: function MakeQUD2 (i, d, m, p, X[n] )
3: if m = 1 then
P|X[n] |−1
4: p(|X[n] |) ← 1 − x=1 p(x)
5:
P
if p(xα ) ∈ {0, cα }for some constant cα ∈ (0, 1), ∀xα ∈ Xα , xα p(xα ) = 1, ∀∅ 6= α ( [n]
then
6: return p
7: end if
8: else h i
9: v ← 0, d 1
10: for j = 1 : 1 : 2 do
11: p(i) ← v (j)
P|X[n] | |X[n] |−i
12: if 0 ≤ 1 − a=1 p(a) ≤ d
, then
13: return MakeQUD2(i + 1, d, m − 1, p, X[n] )
14: end if
15: end for
16: end if
17: end function
18: return p
7 / 18
Work Done and Target Set for Last Semester
8 / 18
Work Done and Target Set for Last Semester
9 / 18
Work Done and Target Set for Last Semester
The total number of elemental inequalities (m) for n random
variables are given by
n n−2
m=n+ 2
2
For n = 2, the number of elemental inequalities are m = 2 + 1 = 3.
H(X1 |X2 ) ≥ 0 (1)
H(X2 |X1 ) ≥ 0 (2)
I (X1 ; X2 ) ≥ 0 (3)
For 2 random variable h vector have the elements
{h : h = [h1 h2 h12 ]T } (4)
The half-space corresponding to equation (1)-(3) are given as.
{h : h12 − h2 ≥ 0} (5)
{h : h12 − h1 ≥ 0} (6)
{h : h12 − h1 − h2 ≥ 0} (7)
10 / 18
Work Done and Target Set for Last Semester
The supporting hyperplane corresponding to equation (5)-(7) are
as follows.
{h : h12 − h2 = 0} (8)
{h : h12 − h1 = 0} (9)
{h : h12 − h1 − h2 = 0} (10)
{h : h = [h1 h2 h2 ]T , h1 , h2 ∈ R+ } (11)
{h : h = [h1 h2 h1 ]T , h1 , h2 ∈ R+ } (12)
{h : h = [h1 h2 h1 + h2 ]T , h1 , h2 ∈ R+ } (13)
11 / 18
Work Done and Target Set for Last Semester
Taking intersection of (11) and (12), we get,
{h : h = [0 0 0]T } (15)
{h : h = [h1 h2 h2 ]T , h1 , h2 ∈ R+ } (16)
{h : h = h2 [0 1 1]T , h2 ∈ R+ } (17)
{h : h = [h1 h2 h1 ]T , h1 , h2 ∈ R+ } (18)
{h : h = h1 [1 0 1]T , h1 ∈ R+ } (19)
{h : h = [h1 h1 h1 ]T , h1 ∈ R+ } (20)
T +
{h : h = h1 [1 1 1] , h1 ∈ R } (21)
Here we can conclude that the random variable X1 and X2 are the
same random variable.
13 / 18
Work Done and Target Set for Last Semester
For n = 3, the total number of elemental inequalities are m = 9
and the total extreme rays are 8. For 3 random variable h vector
have the elements
{h : h = [h1 h2 h3 h12 h13 h23 h123 ]T }
One extreme ray out of the 8 extreme rays have the nature.
R1230 = r1230 [1 1 1 2 2 2 2]T , r1230 ∈ R+
Using the basic properties, we can find the value of r1230 . In this
case, X1 = X2 = X3 and each follow uniform distribution. Let
support size of each ranrom variable is n.
H(X1 ) = H(X2 ) = H(X3 ) = log n = r1230
R1230 = log n[1 1 1 2 2 2 2]T , n ∈ R+
Hence, R1230 contain non-entropic points. A face formed by convex
combination of R1230 contain non-entropic points as well as
entropic points.
14 / 18
Work Done and Target Set for Last Semester
15 / 18
Planned work for the Next Semester
16 / 18
References I
R. W. Yeung, Information theory and network coding. Springer Science & Business Media, 2008.
R. Dougherty, C. Freiling, and K. Zeger, “Networks, matroids, and non-shannon information inequalities,”
IEEE Transactions on Information Theory, vol. 53, no. 6, pp. 1949–1969, 2007.
B. Hassibi and S. Shadbakht, “Normalized entropy vectors, network information theory and convex
optimization,” in 2007 IEEE Information Theory Workshop on Information Theory for Wireless Networks.
IEEE, 2007, pp. 1–5.
X. Yan, R. W. Yeung, and Z. Zhang, “The capacity region for multi-source multi-sink network coding,” in
2007 IEEE International Symposium on Information Theory. IEEE, 2007, pp. 116–120.
Q. Chen and R. W. Yeung, “Characterizing the entropy function region via extreme rays,” in 2012 IEEE
Information Theory Workshop. IEEE, 2012, pp. 272–276.
D. Fong, S. Shadbakht, and B. Hassibi, “On the entropy region and the ingleton inequality,” Mathematical
Theory of Networks and Systems (MTNS), 2008.
17 / 18
Thank you
18 / 18