Many-Help-One Problem for Gaussian Sources with a Tree Structure on their Correlation

Size: px
Start display at page:

Download "Many-Help-One Problem for Gaussian Sources with a Tree Structure on their Correlation"

Transcription

1 Many-Hep-One Probem for Gaussian Sources with a Tree Structure on their Correation Yasutada Oohama arxiv:090988v csit 4 Jan 009 Abstract In this paper we consider the separate coding probem for + correated Gaussian memoryess sources We dea with the case where separatey encoded data of sources work as side information at the decoder for the reconstruction of the remaining source The determination probem of the rate distortion region for this system is the so caed many-hep-one probem and has been known as a highy chaenging probem The author determined the rate distortion region in the case where the sources working as partia side information are conditionay independent if the remaining source we wish to reconstruct is given This condition on the correation is caed the CI condition In this paper we extend the author s previous resut to the case where + sources satisfy a kind of tree structure on their correation We ca this tree structure of information sources the TS condition which contains the CI condition as a specia case In this paper we derive an expicit outer bound of the rate distortion region when information sources satisfy the TS condition We further derive an expicit sufficient condtion for this outer bound to be tight In particuar we determine the rate sum part of the rate distortion region for the case where information sources satisfy the TS condition For some cass of Gaussian sources with the TS condition we derive an expicit recursive formua of this rate sum part Index Terms Mutitermina source coding many-hep-one probem Gaussian rate-distortion region CEO probem I INTROUCTION In muti-user source networks separate coding systems of correated information sources are significant from both theoretica and practica point of view The first fundamenta resut on those coding systems was obtained by Sepian and Wof They considered a separate source coding system of two correated information sources Those two sources are separatey encoded and sent to a singe destination where the decoder reconstruct the origina sources In the above source coding system we can consider the situation where the decoder wishes to reproduce one of two sources We ca this source the primary source In this case the remaining source that we ca the auxiiary source works as a partia side information at the decoder for the reconstruction of the primary source Wyner Ahswede and Körner 3 determined the admissibe rate region for this system the set that consists of a pair of transmission rates for which the primary source can be decoded with an arbitrary sma error probabiity We can naturay extend the system studied by Wyner Ahswede and Körner to the one where there are severa separatey encoded data of auxiiary sources serving as side Y Oohama is with the epartment of Information Science and Inteigent Systems University of Tokushima - Minami Josanjima-Cho Tokushima Japan informations at the decoder The determination of the admissibe rate region for this system is caed the many-hep-one probem In this sense Wyner Ahswede and Körner soved the so caed one-heps-one probem The many-hep-one probem has been known as a highy chaenging probem To date ony three soutions given by Körner and Marton 4 Gefand and Pinsker 5 and Oohama 8 are known Gefand and Pinsker 5 studied an interesting case of the many-hep-one probem They determined the admissibe rate region in the case where the auxiiary sources are conditionay independent if the primary source is given We hereafter say the above correation condition on the information sources the CI condition In Oohama 8 the author extended the many-hep-one probem studied by Gefand and Pinsker 5 to a continuous case He considered the many-hep-one probem for + correated memoryess Gaussian sources where auxiiary sources work as partia side information at the decoder for the reconstruction of the primary source The mean square error was adopted as a distortion criterion between the decoded output and the origina primary source output The rate distortion region was defined by the set of a transmission rates for which the average distortion can be upper bounded by a prescribed eve In 8 the author determined the rate distortion region when information sources satisfy the CI condition This resut contains the author s previous works for Gaussian one-heps-one probem 6 and Gaussian CEO probem 7 The probem sti remains open for Gaussian sources with genera correation Pandya et a 9 studied the genera case and derived an outer bound of the rate distortion region using some variant of bounding technique the author 6 used to prove the converse coding theorem for Gaussian one-heps-one probem However their bounding method was not sufficient to provide a tight resut Recenty in Oohama 0 the author extended the resut of 8 He considered a case of correation on Gaussian sources where + sources satisfy a kind of tree structure on their correation The author caed this tree structure of information sources the TS condition The TS condition contains the CI condition as a specia case In 0 the author derived an expicit outer bound of the rate distortion region for Gaussian sources satisfying TS condition Furthermore he had shown that for this outer bound coincides with the rate distortion region The author aso presented a sufficient condition for the outer bound to coincide with the rate distortion region Subsequenty Tavidar et a extended the TS condition to a binary Gauss Markov tree structure condition They studied a characterization of the rate distortion region for Gaussian

2 source with the binary tree structure In Oohama 0 the anaysis for matching condition of the inner and outer bounds was not sufficient In this paper we further investigate the matching condition of those two bounds and derive a condition much stronger than the matching condition in 0 We first derive an expicit outer bound of the rate distortion region when information sources satisfy the TS condition This outer bound is essentiay the same as the author s previous outer bound in 0 but it has a form more suitabe than the previous one for anaysis of a matching condition Using the derived outer bound we presented an expicit sufficient condition for the outer bound to coincide with the inner bound Furthermore we show that for the outer bound in this paper and that in 0 their rate sum parts coincide with the rate sum part of the inner bound Hence in the case where information sources satisfy the TS condition we estabish an expicit characterization of the rate sum part of the rate distortion region For some cass of Gaussian sources with the TS condition we derive an expicit recursive formua of the optima rate sum Our formua contains the resut of Oohama 7 for Gaussian CEO probem as a specia case II PROBEM STATEMENT AN PREVIOUS RESUTS In this section we state the probem formuation and previous resuts We first state some notations used throughout this paper et Φ Φ and A i i Φ be arbitrary sets Consider a random variabe A i i Φ taking vaues in A i We write n direct product of A i as A n i A i A i et a random vector consisting of n independent copies of the random variabe A i be denoted by A i A i A i A in We write an eement of A n i as a i a i a i a in et S be an arbitrary subset of Φ et A S and A S denote random vectors A i ) and A i ) respectivey Simiary et a S denote a vector a i ) When S k k + we aso use the notation A k for A S and use simiar notations for other vectors or random variabes When k we sometimes omit subscript Throughout this paper a ogarithms are taken to the natura A Forma Statement of the Probem et X i i 0 be correated zero mean Gaussian random variabes taking vaues in rea ines X i et Λ The CI condition Oohama 8 treated corresponds to the case where X X X are independent if X 0 is given In this paper we dea with the case where X X have some correation when X 0 is given et X 0t X t X t ) tbe a stationary memoryess mutipe Gaussian source For each t X 0t X t X t ) obeys the same distribution as X 0 X X ) The mutitermina source coding system treated in this paper is depicted in Fig For each i 0 the data sequence X i is separatey encoded to ϕ i X i ) by encoder function ϕ i The encoded data ϕ i X i ) i 0 are sent to the information processing center where the decoder observes them and outputs the estimation ˆX 0 of X 0 by n X 0 X X X 0 ϕ 0 X 0 ) ϕ 0 X ϕ X ) ϕ X ϕ X ) ϕ ψ ˆX0 Fig Communication system with side informations at the decoder using the decoder function ψ The encoder functions ϕ i i 0 are defined by and satisfy rate constraints ϕ i : X n i M i M i ) n og M i R i + δ ) where δ is an arbitrary prescribed positive number The decoder function ψ is defined by ψ : M 0 M M X n 0 3) enote by F n) δ R 0 R R ) the set that consists of a the + ) tupe of encoder and decoder functions ϕ 0 ϕ ϕ ψ) satisfying )-3) et dx ˆx) x ˆx) x ˆx) X0 be a square distortion measure For X 0 and its estimation ˆX 0 ψϕ 0 X 0 ) ϕ X ) ϕ X )) define the average distortion by X 0 ˆX 0 ) n EdX 0t n ˆX 0t ) t For a given > 0 the rate vector R 0 R R ) is admissibe if for any positive δ > 0 and any n with n n 0 δ) there exists ϕ 0 ϕ ϕ ψ) F n) δ R 0 R R ) such that X 0 ˆX 0 ) + δ et R ) denote the set of a the admissibe rate vector Our aim is to characterize R ) in an expicit form On the other hand we are often interested in a rate sum part of the rate distortion region R ) To examine this quantity define R sum R 0 ) R i min R 0R R ) R ) i To determine R sum R 0 ) in an expicit form is aso of our interest By the rate-distortion theory for singe Gaussian sources when R 0 og+ σ X 0 R R R 0 is admissibe Here og + a maxog a 0 Hence we have R ) R 0 R R ) : R 0 og+ σ X 0 R i 0 i Λ Throughout this paper we assume that σ X 0 and R 0 < og σ X 0

3 3 X 0 Z Y Y X N Z Y Y N Z3 Y 3 Y 3 X X 3 N 3 Fig TS condition in the case of 4 X 0 X 3 N 3 N4 X 4 Z 0 X 0 X 0 N Z 0 X 0 X 0 X N Z3 0 X 0 X 0 X Fig 3 CI condition in the case of 4 B Tree Structure of Gaussian Sources N4 X 4 In this subsection we expain the tree structure of Gaussian source which is an important cass of correation Consider the case where the + random variabes X 0 X X satisfy the foowing correations Y 0 X 0 Y Y + Z 4) X Y + N X Y N Z where Z i i Λ are independent Gaussian random variabes with mean 0 and variance σz i and N i i are independent Gaussian random variabes with mean 0 and variance σn i We assume that Z is independent of X 0 and that N is independent of X 0 and Z We can see that the above X 0 X X ) has a kind of tree structurets) We say that the source X 0 X X ) satisfies the TS condition when it satisfies 4) The TS condition contains the CI condition as a specia case by etting σ Zi i be zero et S be an arbitrary subset of Λ The TS condition is equivaent to the condition that for S Λ the random variabes X S X 0 Z ) X S c form Markov chains X S X 0 Z ) X S c in this order The TS and CI conditions in the case of 4 are shown in Fig and 3 respectivey C Previous Resuts In this subsection we state the previous resuts on the determination probem of R ) et U i i 0 be random variabes taking vaues in rea ines U i For S Λ define G) U 0 U ) : U 0 U ) is a Gaussian random vector that satisfies U X X 0 U 0 U S X S 0 X S c U S c for any S Λ and EX 0 ψu 0 U ) for some inear mapping ψ : U 0 U X 0 where S c Λ S et ) i π π) πi) π) be an arbitrary permutation on Λ and Π be a set of a permutations on Λ For S Λ we set πs) πi) efine subsets S i i of Λ by S i i i + Set R π ) R 0 R R ) : There exists a random vector U 0 U ) G) such that R 0 IX 0 ; U 0 U ) R πi) IX πi) ; U πi) U πs c i )) for i R in) ) conv R in) π ) π Π where conva denotes a convex hu of the set A Then we have the foowing Theorem Oohama 8): For Gaussian sources with genera correation R in) ) R ) For Gaussian sources with the CI condition the inner bound R in) ) is tight that is R in) ) R ) in) The above inner bound R ) can be regarded as a variant of the inner bound which is we known as the inner bound of Berger and Tung 3 Theorem contains the soution that Oohama 6 obtained to the one-heps-one probem for Gaussian sources as a specia case When R 0 0 the second resut of Theorem has some impications for the Gaussian CEO probem studied by Viswanathan and Berger 4 and Oohama 7 and source coding probem for mutitermina communication systems with a remote source investigated by Yamamoto and Itoh 5 and Fynn and Gray 6 The notion of TS condition for Gaussian sources was first introduced by Oohama 0 Tavidar et a extended the TS condition to a binary Gauss Markov tree structure condition They studied a characterization of the rate distortion region for Gaussian sources with a binary tree structure III RESUTS ON THE RATE ISTORTION REGION In this section we state our main resuts on inner and outer bounds of R ) in the case where X 0 X X ) satisfies the TS condition

4 4 A efinition of Functions and their Properties In this subsection we define severa functions which are necessary to describe our resuts and present their properties et r i i Λ be nonnegative numbers efine the sequence of nonnegative functions f r ) f 0r ) by the foowing recursion: f r ) e r σ N + e r σ N f r ) f + r + ) +σ Z + f + r + ) + e r σ N f 0 r ) f r ) +σ Z f r ) Next we define the sequence of nonnegative functions g r 0 ) 0 g r 0 r ) by the foowing recursion: g 0 r 0 ) e r 0 σx 0 g r 0 ) g 0r 0) σ Z g 0r 0) g + r 0 r ) + 6) g r 0r ) σ e r ) N + σ g Z r + 0r ) σ N e r ) where a + maxa 0 et B ) be the set of a nonnegative vectors r0 that satisfy f 0 r ) g 0 r 0 ) e r 0 σ X 0 et B ) be the boundary of B ) that is the set of a nonnegative vectors r 0 that satisfy f 0 r ) g 0 r 0 ) e r 0 σ X 0 We can easiy show that the functions we have defined satisfy the foowing property Property : a) For each i Λ f 0 r ) is a monotone increasing function of r i For each and for each i + f r ) is a monotone increasing function of r i b) For each and for each i 0 g r 0 r ) is a monotone decreasing function of r i c) If r 0 B ) then for 0 g r 0 r ) f r ) In the above inequaities the equaities simutaneousy hod if and ony if r 0 B ) efine Fr ) + σ Z f r ) G r 0 r ) + σ Z g r 0 r ) 5) For S Λ define f 0 r S ) f 0 r ) rs c0 Fr S) Fr ) rs c0 We can easiy show that the functions Fr ) and G r 0 r ) satisfy the foowing property Property : a) For each i S Fr S ) is a monotone increasing function of r i b) For each i 0 G r 0 r ) is a monotone decreasing function of r i c) If r 0 B ) then G r 0 r ) Fr ) The equaity hods if and ony if r 0 B ) For > 0 r i 0 i Λ and S Λ define J S r 0 r r S r S c) og+ Gr 0r ) Fr S c) K S r S r S c) og Fr ) Fr S c) +σ X 0 f 0r ) σ X 0 e r0 +σ f X 0 0r S c) +σ f X 0 0r S c) e ri e ri We can show that for S Λ K S r S r S c) and J S r 0 r r S r S c) satisfy the foowing two properties Property 3: a) If r 0 B ) then for any S Λ J S r 0 r r S r S c) K S r S r S c) The equaity hods when r0 B ) b) Suppose that r B ) If r rs0 sti beongs to B ) then J S r 0 r r S r S c) K rs0 Sr S r S c) rs0 0 Property 4: Fix r B ) For S Λ set ρ S ρ S r S r S c) J S r 0 r r S r S c) By definition it is obvious that ρ S S Λ are nonnegative We can show that ρ ρ S S Λ satisfies the foowings: a) ρ 0 b) ρ A ρ B for A B Λ c) ρ A + ρ B ρ A B + ρ A B In genera Λ ρ) is caed a co-poymatroid if the nonnegative function ρ on Λ satisfies the above three properties Simiary we set ρ S ρ S r S r S c) K S r S r S c) ρ ρ S S Λ Then Λ ρ) aso has the same three properties as those of Λ ρ) and becomes a co-poymatroid

5 5 B Resuts In this subsection we present our resuts on inner and outer bounds of R ) In the previous work 0 we derived an outer bound of R ) We denote this outer bound by ˆR out) Set ) According to 0 ˆRout) ) is given by ˆR out) ) R 0 R ) : There exists a nonnegative vector r 0 r ) such that R 0 r 0 og+ R i r i for any i Λ R 0 + R i og+ + i r i for any S Λ σx 0 +σ f X 0 0r ) Gr 0r )σ X 0 Fr S c) +σx f 0 0r S c) R out) r0 ) R 0 R R ) : R 0 r 0 R i J S r0 r r S r S c) for any S Λ R in) r 0 ) R 0 R R ) : R 0 r 0 R i K S r S r S c) R out) ) R in) ) for any S Λ r0 ) r 0 B) R out) r 0 B) R in) r 0 ) Our main resut is as foows Theorem : For Gaussian sources with the TS condition R in) in) ) R ) R ) ˆR out) ) R out) ) Proof of this theorem wi be given in Section V The out) incusion R ) ˆR ) and an outine of proof of this incusion was given in Oohama 0 Furthermore by in) Theorem we have R ) R ) Hence it suffices to out) show ˆR ) R out) ) and R in) in) ) R ) to prove Theorem Proofs of those two incusions wi be given in Section V We can directy prove R ) R out) ) in a manner simiar to that of Oohama 0 For the detai of the direct proof of R ) R out) ) see Appendix B An essentia difference between R out) ) and R in) ) is the difference between J S r 0 r r S r S c) in the definition of R out) of R in) R out) ) and K S r S r S c) in the definition ) By Property 3 part a) and the definitions of r0 ) and R in) r 0 ) if r0 B ) then R out) r0 ) R in) r 0 ) This gap suggests a possibiity that in some cases those two bounds match In the foowing we present a sufficient condition for R out) ) R in) ) We consider the foowing condition on G r 0 r ) Condition: For each e ri G r 0 r ) is a monotone increasing function of r We ca the above condition the MI condition The foowing is our main resut on a matching condition on inner and outer bounds emma : For Gaussian sources with the TS condition if G r 0 r ) satisfies the MI condition then R out) ) R in) ) Proof of this emma is given in Section V Note that when or σ Z 0 for 3 we have G r 0 r ) + σ Z g r 0 ) which satisfies the MI condition Combining emma and Theorem we estabish the foowing Theorem 3: For Gaussian sources with the TS condition R in) ) R ) ˆR out) ) R out) ) Furthermore if G r 0 r ) satisfies the MI condition then ˆR out) R in) ) R ) ) R out) ) out) In Oohama 0 the equaity R ) ˆR ) was stated without compete proof We can see that this equaity can be obtained by Theorem emma and the fact that the MI condition hods for Next we present a sufficient condition for G r 0 r ) to satisfy the MI condition et fj j positive numbers defined by the foowing recursion: f σ N + σ N f f + +σ Z + f + + σ N be a sequence of By definition it is obvious that f r ) f Then we have the foowing proposition Proposition : If k 7) σ ) Z k+ k ) + σ σ Z k+ fk+ + σz j fj 8) N j+ hod for then G r 0 r ) satisfies the MI condition Proof of this proposition wi be given in Appendix A It can be seen from this proposition that for 3 the MI condition hods for reativey sma vaues of σ Z In particuar when 3 the sufficient condition given by 8) is + σz σz σ N σ N + σ N 3 )

6 6 X 0 Z Y Y N Z Y Y X X N N3 X 3 Fig 4 TS condition in the case of 3 Soving the above inequaity with respect to σ Z we have σ Z + + 4σ N σ N + σ N 3 ) σ N The TS condition in the case of 3 is shown in Fig 4 IV RATE SUM PART OF THE RATE ISTORTION REGION In this section we state our resut on the rate sum part of R ) Set R ) sum R 0) R u) sum R 0) ˆR ) min J Λ R 0 r r ) r :f 0r ) g 0R 0) min K Λ r ) r :f 0r ) g 0R 0) ˆR out) et sum R 0) be the minimum rate sum for ) Then it immediatey foows from Theorem that we have the foowing coroary Coroary : For Gaussian sources with the TS condition R ) sum R 0) ˆR ) sum R 0) R sum R 0 ) R u) sum R 0) On the other hand we have the foowing emma emma : For Gaussian sources with the TS condition we have R ) sum R 0) R u) sum R 0) Proof of this emma wi be given in Section V Combining Coroary and emma we have the foowing Theorem 4: For Gaussian sources with the TS condition R sum R 0 ) R u) sum R 0) ˆR ) sum R 0) R ) sum R 0) min r + r :f 0r ) og Fr ) g 0R 0) R 0 + og σ X 0 The optima rate sum R sum R 0 ) has a form of optimization probem We dea with this optimization probem in a case where σ N σ for and σ Z ǫ σ for In this case we sove the optimization probem by deriving a parametric form of R sum R 0 ) In the above case of identica variances the recursion 5) is f + f + ǫ + σ + ) e r f + σ for 9) The optimization probem presenting R sum R 0 ) is R sum R 0 ) min r :f 0r ) g 0R 0) R 0 + og σ X 0 For we set r + og + ǫ σ f r )) α σ f r ) +ǫ σ f r ) for α e r for 0) By the above transformation we transform the variabe r into α From 0) we have f σ α for ) ǫ α Since f 0 α must satisfy 0 α < ǫ From 9) and ) we have e r α + α ǫ α Since r 0 for α must satisfy α α ǫ α < α ǫ α for α ǫ α < α < for ) et ǫ and et 0 ǫ ) be a direct product of the semi-open intervas 0 ǫ ) et A be a set of a dimensiona vectors α 0 ǫ ) that satisfy ) Using α R sum R 0 ) is rewritten as R sum R 0 ) min α ) Aα) α σ g 0R 0) R 0 + og σ X 0 og + og ǫ α ) ) α + α + ǫ α + og α ) where A α ) α : α α α ) A To sove the above optimization probem set ζ ζα ) ) α og + α + ǫ α + og ǫ α ) + og α ) Then we have the foowing emma

7 7 emma 3: For α Aα ) ζα ) is stricty concave with respect to α Proof of this emma wi be given in Appendix C It can be seen from this emma that if we can find θ satisfying ζ α θ 0 and θ A σ g 0 R 0 )) this θ is the unique vector which attains R sum R 0 ) We sha give such θ in an expicit form of recursion et ω 0 ) efine the sequence of functions θ ω) by the foowing recursion: θ ω) ω θ ω) θ ω) for θ ω) + ǫ θ ω) ω + ǫ ω θ ω) ǫ )θ + ω) ǫ +ǫ +θ + ω)) + ǫ θ ω) ǫ )θ + ω) ǫ +ǫ +θ + ω)) 3) Then we have the foowing emma emma 4: The sequence of functions θ ω) defined by 3) satisfies the foowings a) b) 0 θ ω) θ ω) ǫ θ ω) θ ω) ǫ θ ω) θ ω) ǫ < θ θ ω) ω) < 0 θ ω) θ ω) ǫ θ ω) ǫ +ǫ θ ω) ǫ θ ω) < θ ω) < ǫ For ǫ + +ǫ + +ǫ ǫ + θ ω) θ ω) ǫ θ ω) θ ω) ǫ θ ω) < θ ω) < ǫ ǫ +ǫ The above 4)-6) impy that θ ω) A θ ω)) ζ α θ ω) 0 4) 5) 6) c) For each θ ω) is differentiabe with respect to ω 0 ) and satisfies the foowing: dθ dω ) +ǫ ω j+ + ) +) +ǫ ω+) +ǫ j θ jω)+) j+ +ǫ j θ jω)+) > 0 This impies that for each the mapping ω 0 ) θ ω) is an injection Proof of this emma is given in Appendix From this emma we obtain the foowing theorem Theorem 5: et θ ω) be a sequence of functions defined by 3) Then we have the foowing parametric form of R sum R 0 ) with the parameter ω 0 ): σ g 0 R 0 ) σ e R 0 θ σ ω) X 0 R sum R 0 ) ) og R 0 + og σ X 0 ) θ ω) ǫ θ ω) + θ +ω) + og ǫ θ ω)) + og ω) When ǫ 0 for the recursion 3) becomes the foowing: θ ω) ω θ ω) ω θ ω) θ ω) θ + ω) 7) for Soving 7) we obtain θ ω) +)ω The parametric form of R sum R 0 ) becomes σ g 0 R 0 ) θ ω) ω R sum R 0 ) ) og ω) 8) R 0 + og σ X 0 From 8) we have R sum R 0 ) ) ) og σ g 0 R 0 ) R 0 + og σ X 0 9) In particuar by etting R 0 0 and in 9) we have im R sum 0) σ g 0 0) + og σ X 0 σ σ X0 σx 0 + og σ X 0 The above formua coincides with the rate distortion function for the quadratic Gaussian CEO probem obtained by Oohama 7 Hence our soution to R sum R 0 ) incudes the previous resut on the Gaussian CEO probem as a specia case V PROOFS OF THE RESUTS In this section we prove Theorem and emma stated in Section III and prove emma stated in Section IV A erivation of the Outer Bound out) In this subsection we prove ˆR ) R out) ) stated in Theorem Proof of ˆRout) ) R out) ): Set Ĵ S r 0 r r S r S c R 0 ) og+ Fr S c) Gr 0r )σ X 0 +σx f 0 0r S c) + + r i R 0 i

8 8 We first observe that Ĵ S r 0 r r S r S c r 0 ) og + Fr S c) Gr 0r og )σx 0 Fr S c) +σ f X 0 0r S c) Gr 0r )σ X 0 +σx f 0 0r S c) r i r 0 i + r i r 0 i J S r 0 r r S r S c) 0) Then we have the foowing a) ˆR out) ) R 0 R ) : There exists a nonnegative vector r 0 r ) such that R 0 r 0 og+ σ X 0 +σ f X 0 0r ) R i ĴS r 0 r r S r S c R 0 ) for any S Λ b) R 0 R ) : There exists a nonnegative vector r such that R 0 og+ σ X 0 +σx f 0 0r ) R i ĴS R 0 r r S r S c R 0 ) for any S Λ R 0 R ) : There exists a nonnegative vector r 0 r ) such that R 0 r 0 og σx 0 +σ f X 0 0r ) R i ĴS r 0 r r S r S c r 0 ) for any S Λ c) R 0 R ) : There exists a nonnegative vector r 0 r ) such that R 0 r 0 og σ X 0 +σ f X 0 0r ) R i J S r 0 r r S r S c) for any S Λ R out) ) Step a) foows from the definition of Ĵ S R 0 r r S r S c R 0 ) and the nonnegative property of R Step b) foows from that Ĵ S r 0 r r S r S c R 0 ) is a monotone decreasing function of r 0 Step c) foows from 0) Thus ˆR out) ) R out) ) is proved B erivation of the Inner Bound In this subsection we prove R in) ) R ) stated in Theorem We first derive a preiminary resut on a form of R in) ) Fix R 0 r 0 and set R in) r 0 R 0 ) R R ) : R 0 R R ) R in) r 0 ) et Λ ρ) ρ ρ S r S r S c) S Λ be a co-poymatroid defined in Property 4 Expression of R in) r 0 R 0 ) using Λ ρ) is R in) r 0 R 0) R R ) : R i ρ S r S r S c) for any S Λ The set R in) r 0 R 0 ) forms a kind of poytope which is caed a co-poymatroida poytope in the terminoogy of matroid theory It is we known as a property of this kind of poytope that the poytope R in) r 0 R 0) consists of! end-points whose components are given by R πi) ρ πi) π) r πi) π) r π) πi ) ) ρ πi+) π) r πi+) π) r π) πi) ) for i R π) ρ π) r π) r π) π ) ) ) where ) i π Π π) πi) π) is an arbitrary permutation on Λ For each π Π and r0 B ) et R in) π r 0 ) be the set of nonnegative vectors R 0 R R ) satisfying R 0 r 0 R πi) ρ πi) π) r πi) π) r π) πi ) ) ρ πi+) π) r πi+) π) r π) πi) ) for i R π) ρ π) r π) r π) π ) ) Then we have R in) r 0 ) conv π Π R in) π r 0 ) ) Proof of R in) in) ) R ): Fix π Π and r0 B ) arbitrary By ) it suffices to show that for r0 B ) R in) π r in) 0 ) R π ) to prove Rin) ) R in) ) et V i i 0 Λ be independent Gaussian random variabes with mean 0 and variance σv i Suppose that V0 is independent of X0 efine the Gaussian random variabes U i i 0 Λ by U i Xi + V i i 0 Λ From the above definition it is obvious that U X X 0 U 0 U S X S 0 X S c U S c for any S Λ 3)

9 9 For given r i 0 i S and > 0 set σ V i er i when σ N i r i > 0 When r i 0 we choose U i so that U i takes the constant vaue zero efine the sequence of random variabes Ω 0 by Ω e r σ N U + e r σ N U Ω +σ Z + f + r + ) Ω + + e r σ N U for Ω 0 +σz f r ) Ω Note that Ω 0 Ω 0 U ) is a inear function of U Then by an eementary computation we have X f σ σ 0 r ) X 0 V 0 σ V 0 U 0 + Ω 0 U ) 4) +Ñ0 5) where Ñ0 is a zero mean Gaussian random variabe with variance + + f σx σ 0 r ) 0 V 0 Ñ 0 is independent of U 0 U ) Since r 0 B ) we have We put Then from 6) and 7) we have e r 0 σ X 0 f 0 r ) 6) + σx 0 σ V 0 e r 0 7) σ V 0 + f 0 r ) σ X 0 + e r 0 + f 0 r ) 8) Based on 5) 7) and 8) define the inear function ψ of U 0 U ) by ψu 0 U ) + e r 0 σ + f 0 r ) X 0 e r 0 U 0 + Ω 0 U ) Then we obtain E X 0 ψu 0 U ) Var Ñ0 + e r 0 σ + f 0 r ) 9) X 0 From 3) and 9) we have U 0 U ) G) By simpe computations we can show that r 0 IX 0 ; U 0 U ) r i IX i ; U i X 0 Y ) for any i Λ og F S r S ) + σx 0 f 0 r S ) 30) IX 0 Y ; U S ) for any S Λ Using 3) and 30) the + inequaities of ) are rewritten as R 0 IX 0 ; U 0 U ) R πi) IX 0 Y ; U πsi) U πs c i )) +IX πi) ; U πi) X 0 Y ) IX 0 Y ; U πsi+) U πs c i+ )) IX 0 Y ; U πi) ; U πs c i )) +IX πi) ; U πi) X 0 Y U πs c i )) IX 0 Y X πi) ; U πi) U πs c i )) IX πi) ; U πi) U πs c i )) for i in) Thus we concude that R 0 R π) R π) ) R π ) C Proofs of emmas and In this subsection we prove emmas and We first present a preiminary observation on R out) ) Fix R 0 r 0 arbitrary and set R out) r 0 R 0 ) R R ) : R 0 R R ) R out) r 0 ) et Λ ρ) ρ ρ S r S r S c) S Λ be a co-poymatroid defined in Property 4 Expression of R out) 0 r0 R 0) using Λ ρ) is R out) r 0 R 0 ) R R ) : R i ρ S r S r S c) for any S Λ The set R out) r0 R 0) forms a co-poymatroida poytope The poytope R out) r0 R 0) consists of! end-points whose components are given by R πi) ρ πi) π) r πi) π) r π) πi ) ) ρ πi+) π) r πi+) π) r π) πi) ) for i R π) ρ π) r π) r π) π ) ) 3) where ) i π Π π) πi) π) For each π Π and set B π ) r0 : r0 B ) and r πi) 0 for i + B π ) r0 : r0 B ) and r πi) 0 for i + In particuar when π is the identity map we omit π to write B ) and B ) By Property 3 when r0 B π) the

10 0 end-point given by 3) becomes R πi) ρ πi) π) r πi) π) r π) πi ) ) ρ πi+) π) r πi+) π) r π) πi) ) for i R π) ρ π) r π) r π) π ) ) R πi) 0 for i + 3) Next we present a emma on a property of G r 0 r ) emma 5: For r0 B ) G r 0 r ) is computed as G r 0 r ) r σ Zk g k r 0 r k ) k Proof: By Property part c) for + k 0 g k r 0 r k ) fr k ) 0 Hence the resut of emma 5 foows Proof of emma : Fix π Π and r0 B ) arbitrary et R 0 R ) be a nonnegative rate vector such that R 0 r 0 and components of R satisfy 3) To prove emma it suffices to show that this nonnegative vector beongs to R in) ) For we prove the caim that under the MI condition if r0 B π ) then the rate vector R 0 R ) satisfying R 0 r 0 and 3) beongs to R in) ) We prove this caim by induction with respect to When from 3) we have R π) ρ π) r π) ) 33) R πi) 0 for i The function ρ π) r π) ) is computed as ρ π) r π) ) J π) r 0 r r π) r π) c) rπ) c0 Gr0r ) rπ) og+ c0 σ X 0 e r0e r π) 34) By the above form of ρ π) r π) ) and σ X 0 e r 0 σ X 0 e R0 > ρ π) r π) ) is positive Since r 0 B π ) we can decrease r π) keeping r 0 B π ) so that it arrives at r π) 0 or a positive r π) satisfying r 0 r π) r π) c) r 0 rπ) 0 0) B π ) 35) et R 0 Rπ) R π) ) be a rate vector corresponding to r 0 rπ) r π) c) If r π) 0 then by Property 3 part b) ρ π) r π) ) must be zero This contradicts the fact that ρ π) r π) ) is positive Therefore rπ) must be positive Then from 35) we have R 0 R π) R π) ) R 0 Rπ) 0 0) R in) ) On the other hand by emma 5 we have G r 0 r ) rπ) c0 G r 0 r ) r π)+ 0rπ) 0 π) k + σ Z g k r 0 r k ) r π) 0 36) From 34) and 36) we can see that G r 0 r ) rπ) c0 does not depend on r π) This impies that ρ π) r π) ) is a monotone increasing function of r π) Then we have R π) Rπ) Hence we have R 0 R π) R π) ) R 0 R π) 0 0) R in) ) Thus the caim hods for We assume that the caim hods for Since f 0 r0 ) is a monotone increasing function of r π) on B π ) we can decrease r π) keeping r0 B π ) so that it arrives at rπ) 0 or a positive r π) satisfying r 0 rπ) r π) c) B π) 37) et R 0 Rπ) R π) ) be a rate vector corresponding to r 0 rπ) r π) c) By Property 4 part b) and the MI condition the functions ρ πi) π) r πi) π) r π) πi ) ) ρ πi+) π) r πi+) π) r π) πi) ) for i ρ π) r π) r π) π ) ) appearing in the right members of 3) are monotone increasing functions of r π) Then from 3) we have R πi) Rπi) for i R πi) Rπi) 38) 0 for i + When r π) 0 we have r 0 r π) r π) c) B π ) Then by induction hypothesis we have R 0 Rπ) R π) ) Rin) ) When rπ) > 0 from 37) we have R 0 Rπ) R π) ) Rin) ) Hence by 38) we have R 0 R π) R π) ) R 0 R π) R π) 0 0) R in) ) Thus the caim hods for This competes the proof of emma Proof of emma : For R 0 > 0 and for set B R 0 ) r : R 0 r ) B ) B R 0 ) r : R 0 r ) B )

11 We first observe that R ) sum R 0) min R u) min J Λ R 0 r r ) r r B R 0) + 0 sum R 0) min min K Λ r ) r B R 0) We compute J Λ R 0 r r ) r By emma 5 for 0 + r B R 0 ) G R 0 r ) r σ Z g k R 0 r k ) k From the above formua we can see that for r B R 0 ) G R 0 r ) r + 0 is a function of r We denote this function by G R 0 r ) that is G R 0 r ) Then for r B R 0 ) + σ Z g k R 0 r k ) k J Λ R 0 r r ) r + 0 og+ G R 0 r ) σ X 0 e R0 e ri 39) i We denote the right member of 39) by J Λ R 0 r r ) Using this function R ) sum R 0) can be written as R ) sum R 0) min min J Λ R 0 r r ) r B R 0) Note here that J Λ R 0 r r ) is a monotone increasing function of r To prove R ) sum R 0) R u) sum R 0) it suffices to show that for min J Λ R 0 r r ) min K Λ r ) r B R 0) r B R 0) We prove this caim by induction with respect to When the function J Λ R 0 r ) is computed as J Λ R 0 r ) +σ og+ Z g R 0)σ X 0 e R 0e r og+ σ X 0 e R 0e r σ Z g 0R 0) Since σ X 0 e R 0 > J Λ R 0 r ) is positive Since J Λ R 0 r ) is a monotone increasing function of r the minimum of this function is attained by r 0 or a positive r satisfying r B R 0 ) If r 0 then by Property 3 part b) J Λ R 0 r ) must be zero This contradicts that J Λ R 0 r ) is positive Therefore r must be positive Then by r B R 0 ) we have J Λ R 0 r ) J Λ R 0 r ) K Λ r ) min r B R 0) K Λr ) Thus the caim hods for We assume that the caim hods for Since J Λ R 0 r r ) is a monotone increasing function of r the minimum of this function is attained by r 0 or a positive r satisfying r r ) B R 0 ) When r 0 we have r B R 0 ) and J Λ R 0 r r ) J Λ R 0 r r r ) 40) Computing J Λ R 0 r r r ) we obtain J Λ R 0 r r r ) J Λ R 0 r r ) r 0 og+ G R 0 r ) σ X 0 e R0 e ri i J Λ R 0 r r ) 4) Combining 40) and 4) we have J Λ R 0 r r ) J Λ R 0 r r ) 4) On the other hand by induction hypothesis we have J Λ R 0 r r ) Combining 4) and 43) we have J Λ R 0 r r ) When r > 0 we have min K Λ r ) 43) r B R 0) min K Λ r ) r B R 0) min K Λ r ) r B R 0) J Λ R 0 r r ) J Λ R 0 r r r ) K Λ r r ) min K Λ r ) r B R 0) where the second equaity foows from r r ) B R 0 ) Thus the caim hods for competing the proof VI CONCUSIONS We have considered the Gaussian many-hep-one probem and given a partia soution to this probem by deriving expicit outer bound of the rate distortion region for the case where information sources satisfy the TS condition Furthermore we estabished a sufficient condition under which this outer bound is tight We have determined the rate sum part of the rate distortion region for the case where information sources satisfy the TS condition For the case that information sources do not satisfy the TS condition we can not derive an outer bound having a simiar form of R out) ) since the proof of the converse coding theorem depends heaviy on this property of information sources Hence the compete soution is sti acking for Gaussian information sources with genera correation

12 og G r 0 r ) can be rewritten as A Proof of Proposition APPENIX In this appendix we prove Proposition To prove this proposition we give some preparations For 0 we set η η r 0 r ) g0 r 0 ) for 0 g r 0 r ) σ N e r ) for For and a < σ Z define τ a) Then η 0 Fix a < σ Z k+ and set ) a+ σ a Z + σ e r N satisfies the foowing: η r 0 r ) τ η r 0 r ) ) for 44) p k a) sup p :og σ Zk+ a + pb a) σ b Z + k+ for any b < σ Z k+ By a simpe computation we have σz k+ p k a) σ a for 0 a < σ Z k+ Z k+ 0 for a < 0 σ Z k+ σ Z k+ a for a < σ Z k+ 45) Fix a < and set σz j q j a) sup q :τ j b) τ j a) qb a) for any b < σ Z j By a simpe computation we have for 0 a < σ q j a) a) Z σ j Z j 0 for a < 0 σ Z j a) for a < σ Z j 46) Proof of Proposition : et be a set of integers such that η r 0 r ) takes a positive vaue for some r0 B ) From 44) there exists a unique integer such that 0 Using η and og G r 0 r ) og + σz k g r 0 r k ) k k k0 k0 og og og σz η k r k 0r k ) + σ η Z k r k+ 0r k ) + σ η Z k r k+ 0r 47) k ) + Fix nonnegative vector r For each s r et Gs ) be a function obtained by repacing r in G r 0 r ) with s that is Gs ) G r 0 r s r + ) It is obvious that when s r Gr ) G r 0 r r r + ) G r 0 r ) By Property part b) we have Gs ) Gr ) for For each s k r k k et η k s ) be a function obtained by repacing r in η k r 0 r k ) with s that is η k s ) η k r 0 r s r k + ) It is obvious that when s r η k r ) η k r 0 r r r k + ) η k r 0 r k ) By Property part b) we have η k s ) η k r ) for k For each we evauate an upper bound of og Gs ) og Gr ) Using 47) we have og Gs ) Gr ) k0 k σ Z η k r ) + og k+ σz η k s ) + k+ σ Z η k r ) + og k+ σ η Z k s 48) ) + k+ By definition of p k ) we have σ Z η k r ) + og k+ σ η Z k s p ) + k η k r ))η k s ) η k r ) k+ σ Z k+ σ Z k+ η k r ) η ks ) η k r ) 49) where the ast inequaity foows from η k s ) η k r ) and 45) From 48) and 49) we have og Gs ) Gr ) σz k+ σ η Z k r ) η ks ) η k r )) 50) k+ k By definition of q j ) and 44) for + j k we have η j s ) η j r ) q j η j r ))η j s ) η j r ) ) η j s ) η j r ) 5) σ η Z j j r )

13 3 where the ast inequaity foows from η j s ) η j r ) and 46) Using 5) iterativey for + j k we obtain η k s ) η k r ) η s ) η r )) k j+ σ Zj η j ) 5) Observe that η s ) η r ) σ e s e r N From 5) and 53) we have η k s ) η k r ) e r σ N s r ) σ N s r ) k j+ From 50) and 54) we have og Gs ) Gr ) s r ) k e r σ N s r ) 53) k j+ σ Zj η j ) σ Zj η j ) 54) σ Z k+ σ N σ Z k+ η k k j+ σ Zj η j ) 55) By Property part b) and the definition of η j we have η j f j σ N j e rj ) from which we have f j+ +σ Z j+ f j+ σ Z j+ η j + σ Z j+ f j+ + σ Z j+ f j+ 56) From 55) and 56) we have If og Gs ) Gr ) s r ) k k σ ) Z k+ + σ σ Z k+ fk N k j+ + σ Z j f j ) 57) σ ) Z k+ k ) + σ σ Z k+ fk+ + σz j fj 58) N j+ hod for then by 57) we have or equivaent to og Gs ) Gr ) s r ) s + og Gs ) r + og Gr ) for Hence 58) is a sufficient condition for the MI condition B Proof of R ) R out) ) In this appendix we prove R ) R out) ) stated in Theorem We first present a emma necessary for the proof of this incusion emma 6: IX 0 ; ˆX 0 ) n ) og σ X 0 X 0 ˆX 0) Proof: See the proof of emma in Oohama 7 Next we present an important emma which is a mathematica core of the converse coding theorem et the encoded outputs of X i i 0 by encoder functions ϕ i be denoted by ϕ i X i ) W i Set r 0 n IX 0; W 0 W ) r i n IX i; W i Y ) n IX i; W i Y i ) for i n IX ; W Y ) for i ξ σ X 0 e n IX0;W0W ) Then we have the foowing emma emma 7: IX 0 ; W ) n og + σ X 0 f 0 r ) For we have n og + σz g r 0 r ξ) IY ; W Y ) n og + σz f r ) From the above emma we immediatey obtain the foowing emma 8: IX 0 ; W S ) n og + σ X 0 f 0 r S ) IY ; W S X 0 ) n og Fr S) S Λ IY ; W X 0 ) n og Gξ r 0 r ) We prove R ) R out) ) by emmas 6 and 8 and standard arguments for the proof of the converse coding theorem Proof of R ) R out) ): We first observe that by virtue of the TS condition W S X S X 0 Z ) X S c W S c 59) hod for any subset S of Λ Assume R 0 R R ) R ) Then for any δ > 0 there exists an integer n 0 δ) such that for n n 0 δ) and for i Λ we obtain the foowing chain of inequaities: nr 0 + δ) og M 0 HW 0 ) HW 0 W ) IX 0 ; W 0 W ) nr 0 60)

14 4 Furthermore for any subset S Λ we obtain nr 0 + nr i + δ) IX 0 ; W 0 W ) + HW i ) HW 0 W ) + HW i ) HW 0 W S W S c) + HW S W S c) HW 0 W S W S c) IX 0 Z ; W 0 W S W S c) +HW 0 W S W S cx 0 Z ) a) IX 0 Z ; W 0 W S W S c) + HW i X 0 Z ) IX 0 Y ; W 0 W S W S c) + IX i ; W i Y ) 6) Step a) foows from 59) On the other hand by emma 6 we have for n n 0 δ) IX 0 ; W 0 W ) n ) og σ X 0 ξ IX 0 ; ˆX 0 ) n ) og σ X 0 +δ which together with 60) 6) and emma 8 yieds the foowing ower bounds of IX 0 ; W 0 W ) and IX 0 Y ; W 0 W S W S c): IX 0 ; W 0 W ) IX 0 ; W 0 W ) IX 0 ; W ) n og σ X 0 +σx f 0 0r ) ξ n og σx 0 6) +σ X 0 f 0r ) +δ) IX 0 Y ; W 0 W S W S c) IX 0 Y ; W 0 W S W S c) IX 0 Y ; W S c) IX 0 ; W 0 W ) + IY ; W X 0 ) IX 0 ; W S c) IY ; W S c X 0 ) n og σ X 0 Gξr 0r ) Fr S c) +σ f X 0 0r S c) ξ n og σ X 0 G+δr 0r ) 63) Fr S c) +σ X 0 f 0r S c) From 6) and 63) we have R i + δ) og + +δ) σ X 0 G+δr 0r ) Fr S c)+δ) +σ X 0 f 0r S c) r i r 0 64) Note here that R i + δ) are nonnegative Hence from 60) 6) and 64) we obtain R 0 + δ r 0 og σx 0 65) +σ X 0 f 0r ) +δ) and for S Λ R i + δ) J S + δ r 0 r r S r S c) The inequaity 65) impies that r0 B + δ) Thus by etting δ 0 we obtain R 0 R R ) R out) ) Finay we prove emma 7 For n dimensiona random vector U with density et hu) be a differentia entropy of U The foowing two emmas are some variants of the entropy power inequaity emma 9: et U i i 3 be n dimensiona random vectors with densities and et T be a random variabe taking vaues in a finite set We assume that U 3 is independent of U U and T Then we have πe e n hu+u3 UT) πe e n hu UT) + πe e n hu3) emma 0: et U i i 3 be n random vectors with densities et T T be random variabes taking vaues in finite sets We assume that those five random variabes form a Markov chain T U U 3 U T in this order Then we have πe e n hu+u U3TT) πe e n hu U3T) + πe e n hu U3T) Proof of emma 7: efine the sequence of n dimensiona random vectors S by S σ N X + σ Z + Y + 66) By an eementary computation we obtain X 0 σˆn 0 Y σ + ˆN 0 Z Y σˆn Y σz + σ ˆN S + ˆN 67) where ˆN 0 is an n dimensiona random vector whose components are n independent copies of a Gaussian random variabe with mean 0 and variance σ ˆN Nˆ 0 is independent of Y For each ˆN is independent of Y and S The variance σ ˆN 0 have the foowing form: Set + σ ˆN σ σ X 0 0 Z + + σ ˆN σ σ Z N σ Z + λ 0 πe e n hx0 W ) µ 0 πe e n hy W ) µ 0 πe e n hy X0W ) λ πe e n hy Y W ) µ πe e n hs Y W ) µ πe e n hs Y W ) We can easiy verify that 68) µ 0 µ 0 λ 0 σ ˆN 0 µ µ λ σ ˆN 69) Appying emma 9 to 67) we obtain λ 0 σ4ˆn 0 µ σz σ ˆN0 λ σ 4ˆN µ + σ ˆN 70)

15 5 From 69) and 70) we obtain ) λ 0 + λ σ σ σ X 0 Z Z λ + + µ σ σ σ Z N Z + 7) On the other hand we note that for each the five random variabes W X Y Y + and W+ form a Markov chain W X Y Y + W+ in this order Then appying emma 0 to 66) we obtain µ σ N e r + σ 4 Z + λ + 7) Combining 7) and 7) we obtain for ) λ + e r ) + λ + 73) σ σ σ σ Z N Z + Z + Set ν 0 λ 0 σ X 0 ν λ σ Z Then we have IX 0 W ) n og + σ X 0 ν 0 ) IY W Y ) n og + σ Z ν ) IY W Y ) n og + σ Z ν ) nr Note that ν 0 are nonnegative From 7) and 73) ν 0 satisfies the foowing recursion: ν σ Z e r ) 74) ν ν ν 0 ν +σ Z ν + e r σ N e r σ N + e r σ N 75) ν + +σ Z + ν + + e r σ N 76) ν +σ Z ν ν 0 e r 0 ξ σ X 0 77) From 74)-77) we obtain the upper bounds of IX 0 ; W ) and IY ; W Y ) in emma 7 On the other hand from 76) 77) and the nonnegative property of ν 0 we have ν 0 ν + e r 0 ξ σ X 0 + ν ν σ N e r ) + ν 0 σ Z ν 0 78) σ Z + ν σ N e r ) + 79) From 78) and 79) we obtain the ower bound of IY ; W Y ) in emma 7 C Proof of emma 3 et α β A α ) Then we have the foowing chain of inequaities: tζα ) + t)ζβ ) t og α ) + α + ǫα + t)og β ) + β + ǫβ t og ǫα ) + t)og ǫβ ) + +t og α ) + t)og β ) α og t + tα + ǫα β t) + t)β + ǫβ a) og ǫtα + t)β + + og tα t)β og b) + tα + t)β ǫtα + t)β +tα + + t)β + og ǫtα + t)β + og tα + t)β ζ tα + ) t)β Step a) foows from the strict concavity of the ogarithm a function Step b) foows from the strict concavity of ǫa for a > 0 Proof of emma 4 Proof of emma 4 part a): For the proof we use the foowing inequaity + a + ǫ + a) a + ǫa + ǫ 80) The recursion of 3) is equivaent to θ ω) ǫ θ ω) θ + θ + ω) + ǫ + θ + ω) + 8) for Appying 80) to the second term in the right members of 8) we have θ ω) ǫ θ ω) θ θ + ω) ω) + ǫ θ + ω) + ǫ + ǫ or equivaent to θ ω) + ǫ θ ω) θ θ + ω) ω) θ ω) ǫ θ + ω) + ǫ + ǫ 8)

16 6 for The vaidity of 4) is obvious From 4) we have ǫ θ > θ ω) ω) +ǫ 0 θ ω) 83) θ ω) < +θ ω) +ǫ +θ ω)) We first prove 5) Using 8) we have θ ω) ǫ θ ω) θ ω) θ ω) + θ ω) + ǫ + θ ω)) + a) < Step a) foows from the second inequaity of 83) Using 8) we have θ ω) ǫ θ ω) θ ω) θ ω) θ ω) + ǫ θ ω) + ǫ a) ǫ + ǫ + ǫ Step a) foows from the first inequaity of 83) Thus 5) is proved Next we prove 6) by induction with respect to From 5) we obtain and ǫ > θ ω) θ ω) < θ ω) +ǫ > 0 θ ω) +θ ω) +ǫ +θ ω)) 84) θ ω) ǫ θ ω) ǫ + ǫ 85) Soving 85) with respect to θ ω) we obtain θ ω) Using 8) we have ǫ + ǫ + ǫ ǫ θ 3 ω) ǫ 3 θ 3 ω) θ ω) θ ω) + θ ω) + ǫ + θ ω)) + a) < Step a) foows from the second inequaity of 84) Using 8) we have θ 3 ω) ǫ 3 θ 3 ω) θ ω) θ ω) θ ω) + ǫ θ ω) + ǫ a) ǫ + ǫ + ǫ Step a) foows from the first inequaity of 84) Thus 6) hods for We assume that 6) hods for some + with + that is ǫ + +ǫ + +ǫ + ǫ + θ + ω) θ ω) ǫ θ ω) ǫ + θ ω) ǫ θ ω) < θ +ω) < ǫ + From 86) we obtain ǫ > θ ω) θ +ω) +ǫ θ + ω) > 0 θ ω) < +θ +ω) +ǫ +θ + ω)) +ǫ + 86) 87) and θ ω) ǫ θ ω) ǫ + + ǫ + 88) Soving 88) with respect to θ ω) we obtain ǫ θ ω) + ǫ + ǫ ǫ + Using 8) we have θ ω) ǫ θ ω) θ ω) θ ω) + θ + ω) + ǫ + θ + ω)) + a) < Step a) foows from the second inequaity of 87) Using 8) we have θ ω) ǫ θ ω) θ ω) θ ω) θ + ω) + ǫ θ + ω) + ǫ a) ǫ + ǫ + ǫ Step a) foows from the first inequaity of 87) Thus 6) hods for competing the proof Proof of emma 4 part b): We first observe that ζα ) og ) α + α + + og ǫ α ) ǫ α + og α ) og ǫ α α + ǫ α )α + og α ) og + α + + ǫ + α + ) α Computing + og α ) α ζα ) we obtain α ζα ) α α ǫ α + α α ζα ) α α +α ǫ α + + +ǫ +α + α for From 89) when ζα ) 0 α must satisfy α α + ǫ α 0 +α + +ǫ +α + α α + ǫ α 0 for From 90) we obtain α α + ǫ α α ǫ )α + ǫ +ǫ +α + ) α + ǫ α ǫ )α + ǫ +ǫ +α + ) for 89) 90) 9)

17 7 The reation 9) impies that ζ α θ ω) 0 Proof of emma 4 part c): For the proof we use the foowing recursion: θ ω) ǫ θ ω) θ + θ ω) ω) + ǫ + θ + ω) + 9) Taking the derivative of both sides of 9) with respect to ω we obtain dθ ǫ θ ω)) dω dθ dω + ǫ θ + ω) + ) dθ + dω 93) Since θ ω) A θ ω)) we have θ ω) ǫ θ ω) < θ ω) The above inequaity is equivaent to + ǫ θ ω) + ) > From 93) and 94) we have For set ǫ θ ω) 94) + ǫ θ ω) + ) dθ dω dθ dω dθ + + ǫ θ + ω) + ) dω 95) Φ ω) dθ dω Then by 95) we have j +ǫ j θ jω)+) Φ ω) Φ ω) Φ + ω) for 96) From 96) we have Set Φ ω) Φ ω) Φ ω) Φ + ω) Φ ω) Φ ω) dθ dω +ǫ ω+) j +ǫ ω +ǫ ω+) +ǫ j θ jω)+) 97) +ǫ j θ jω)+) j Aω) +ǫ ω +ǫ ω+) Then by 97) we have j +ǫ j θ jω)+) Φ ω) Φ ω) + )Aω) ) +ǫ ω j +) +ǫ ω+) +ǫ θ jω)+) from which we obtain dθ dω ) +ǫ ω competing the proof j+ +) +ǫ ω+) +ǫ j θ jω)+) REFERENCES Sepian and J K Wof Noiseess coding of correated information sources IEEE Trans Inform Theory vo IT-9 pp Juy 973 A Wyner On source coding with side information at the decoder IEEE Trans Inform Theory vo IT- pp May R F Ahswede and J Körner Source coding with side information and a converse for degraded broadcast channes IEEE Trans Inform Theory vo IT- pp Nov J Körner and K Marton How to encode the modue-two sum of binary sources IEEE Trans Inform Theory vo IT-5 pp 9- Mar S I Gefand and M S Pinsker Source coding with incompete side information in Russian) Prob Pered Inform vo 5 no pp Y Oohama Gaussian mutitermina source coding IEEE Trans Inform Theory vo 43 pp 9-93 Nov Y Oohama The rate-distortion function for the quadratic Gaussian CEO probem IEEE Trans Inform Theory vo 44 pp May Y Oohama Rate-distortion theory for Gaussian mutitermina source coding systems with severa side informations at the decoder IEEE Trans Inform Theory vo 5 pp Juy A Pandya A Kansa G Pottie and M Srivastava ossy source coding of mutipe Gaussian sources: m-heper probem Proceedings of IEEE Information Theory Workshop San AntonioTX pp Oct Y Oohama Gaussian mutitermina source coding with severa side informations at the decoder Proceedings of IEEE Internationa Symposium on Information Theory Seatte USA Juy 9-4 pp Juy 006 S Tavidar P Viswanath and A B Wagner The Gaussian manyhep-one distributed source coding probem Proceedings of IEEE Information Theory Workshop pp Oct 006 T Berger Mutitermina source coding in the Information Theory Approach to Communications CISM Courses and ectures no 9) G ongo Ed Vienna and New York : Springer-Verag 978 pp S Y Tung Mutitermina source coding Ph dissertation Schoo of Eectrica Engineering Corne University Ithaca NY May H Viswanathan and T Berger The quadratic Gaussian CEO probem IEEE Trans Inform Theory vo 43 pp Sept H Yamamoto and K Itoh Source coding theory for mutitermina communication systems with a remote source Trans of the IECE of Japan vo E63 no0 pp Oct T J Fynn and R M Gray Encoding of correated observations IEEE Trans Inform Theory vo IT-33 pp Nov 987

Pricing Multiple Products with the Multinomial Logit and Nested Logit Models: Concavity and Implications

Pricing Multiple Products with the Multinomial Logit and Nested Logit Models: Concavity and Implications Pricing Mutipe Products with the Mutinomia Logit and Nested Logit Modes: Concavity and Impications Hongmin Li Woonghee Tim Huh WP Carey Schoo of Business Arizona State University Tempe Arizona 85287 USA

More information

An Infeasibility Result for the Multiterminal Source-Coding Problem

An Infeasibility Result for the Multiterminal Source-Coding Problem An Infeasibiity Resut for the Mutitermina Source-Coding Probem Aaron B. Wagner, Venkat Anantharam, November 22, 2005 Abstract We prove a new outer bound on the rate-distortion region for the mutitermina

More information

CS229 Lecture notes. Andrew Ng

CS229 Lecture notes. Andrew Ng CS229 Lecture notes Andrew Ng Part IX The EM agorithm In the previous set of notes, we taked about the EM agorithm as appied to fitting a mixture of Gaussians. In this set of notes, we give a broader view

More information

Iterative Decoding Performance Bounds for LDPC Codes on Noisy Channels

Iterative Decoding Performance Bounds for LDPC Codes on Noisy Channels Iterative Decoding Performance Bounds for LDPC Codes on Noisy Channes arxiv:cs/060700v1 [cs.it] 6 Ju 006 Chun-Hao Hsu and Achieas Anastasopouos Eectrica Engineering and Computer Science Department University

More information

(f) is called a nearly holomorphic modular form of weight k + 2r as in [5].

(f) is called a nearly holomorphic modular form of weight k + 2r as in [5]. PRODUCTS OF NEARLY HOLOMORPHIC EIGENFORMS JEFFREY BEYERL, KEVIN JAMES, CATHERINE TRENTACOSTE, AND HUI XUE Abstract. We prove that the product of two neary hoomorphic Hece eigenforms is again a Hece eigenform

More information

A NOTE ON QUASI-STATIONARY DISTRIBUTIONS OF BIRTH-DEATH PROCESSES AND THE SIS LOGISTIC EPIDEMIC

A NOTE ON QUASI-STATIONARY DISTRIBUTIONS OF BIRTH-DEATH PROCESSES AND THE SIS LOGISTIC EPIDEMIC (January 8, 2003) A NOTE ON QUASI-STATIONARY DISTRIBUTIONS OF BIRTH-DEATH PROCESSES AND THE SIS LOGISTIC EPIDEMIC DAMIAN CLANCY, University of Liverpoo PHILIP K. POLLETT, University of Queensand Abstract

More information

Mat 1501 lecture notes, penultimate installment

Mat 1501 lecture notes, penultimate installment Mat 1501 ecture notes, penutimate instament 1. bounded variation: functions of a singe variabe optiona) I beieve that we wi not actuay use the materia in this section the point is mainy to motivate the

More information

Math 124B January 31, 2012

Math 124B January 31, 2012 Math 124B January 31, 212 Viktor Grigoryan 7 Inhomogeneous boundary vaue probems Having studied the theory of Fourier series, with which we successfuy soved boundary vaue probems for the homogeneous heat

More information

Lecture Note 3: Stationary Iterative Methods

Lecture Note 3: Stationary Iterative Methods MATH 5330: Computationa Methods of Linear Agebra Lecture Note 3: Stationary Iterative Methods Xianyi Zeng Department of Mathematica Sciences, UTEP Stationary Iterative Methods The Gaussian eimination (or

More information

Explicit overall risk minimization transductive bound

Explicit overall risk minimization transductive bound 1 Expicit overa risk minimization transductive bound Sergio Decherchi, Paoo Gastado, Sandro Ridea, Rodofo Zunino Dept. of Biophysica and Eectronic Engineering (DIBE), Genoa University Via Opera Pia 11a,

More information

ORTHOGONAL MULTI-WAVELETS FROM MATRIX FACTORIZATION

ORTHOGONAL MULTI-WAVELETS FROM MATRIX FACTORIZATION J. Korean Math. Soc. 46 2009, No. 2, pp. 281 294 ORHOGONAL MLI-WAVELES FROM MARIX FACORIZAION Hongying Xiao Abstract. Accuracy of the scaing function is very crucia in waveet theory, or correspondingy,

More information

Homogeneity properties of subadditive functions

Homogeneity properties of subadditive functions Annaes Mathematicae et Informaticae 32 2005 pp. 89 20. Homogeneity properties of subadditive functions Pá Burai and Árpád Száz Institute of Mathematics, University of Debrecen e-mai: buraip@math.kte.hu

More information

Integrating Factor Methods as Exponential Integrators

Integrating Factor Methods as Exponential Integrators Integrating Factor Methods as Exponentia Integrators Borisav V. Minchev Department of Mathematica Science, NTNU, 7491 Trondheim, Norway Borko.Minchev@ii.uib.no Abstract. Recenty a ot of effort has been

More information

The Group Structure on a Smooth Tropical Cubic

The Group Structure on a Smooth Tropical Cubic The Group Structure on a Smooth Tropica Cubic Ethan Lake Apri 20, 2015 Abstract Just as in in cassica agebraic geometry, it is possibe to define a group aw on a smooth tropica cubic curve. In this note,

More information

A Brief Introduction to Markov Chains and Hidden Markov Models

A Brief Introduction to Markov Chains and Hidden Markov Models A Brief Introduction to Markov Chains and Hidden Markov Modes Aen B MacKenzie Notes for December 1, 3, &8, 2015 Discrete-Time Markov Chains You may reca that when we first introduced random processes,

More information

SUPPLEMENTARY MATERIAL TO INNOVATED SCALABLE EFFICIENT ESTIMATION IN ULTRA-LARGE GAUSSIAN GRAPHICAL MODELS

SUPPLEMENTARY MATERIAL TO INNOVATED SCALABLE EFFICIENT ESTIMATION IN ULTRA-LARGE GAUSSIAN GRAPHICAL MODELS ISEE 1 SUPPLEMENTARY MATERIAL TO INNOVATED SCALABLE EFFICIENT ESTIMATION IN ULTRA-LARGE GAUSSIAN GRAPHICAL MODELS By Yingying Fan and Jinchi Lv University of Southern Caifornia This Suppementary Materia

More information

THE REACHABILITY CONES OF ESSENTIALLY NONNEGATIVE MATRICES

THE REACHABILITY CONES OF ESSENTIALLY NONNEGATIVE MATRICES THE REACHABILITY CONES OF ESSENTIALLY NONNEGATIVE MATRICES by Michae Neumann Department of Mathematics, University of Connecticut, Storrs, CT 06269 3009 and Ronad J. Stern Department of Mathematics, Concordia

More information

Reichenbachian Common Cause Systems

Reichenbachian Common Cause Systems Reichenbachian Common Cause Systems G. Hofer-Szabó Department of Phiosophy Technica University of Budapest e-mai: gszabo@hps.ete.hu Mikós Rédei Department of History and Phiosophy of Science Eötvös University,

More information

Week 6 Lectures, Math 6451, Tanveer

Week 6 Lectures, Math 6451, Tanveer Fourier Series Week 6 Lectures, Math 645, Tanveer In the context of separation of variabe to find soutions of PDEs, we encountered or and in other cases f(x = f(x = a 0 + f(x = a 0 + b n sin nπx { a n

More information

Research Article On the Lower Bound for the Number of Real Roots of a Random Algebraic Equation

Research Article On the Lower Bound for the Number of Real Roots of a Random Algebraic Equation Appied Mathematics and Stochastic Anaysis Voume 007, Artice ID 74191, 8 pages doi:10.1155/007/74191 Research Artice On the Lower Bound for the Number of Rea Roots of a Random Agebraic Equation Takashi

More information

Volume 13, MAIN ARTICLES

Volume 13, MAIN ARTICLES Voume 13, 2009 1 MAIN ARTICLES THE BASIC BVPs OF THE THEORY OF ELASTIC BINARY MIXTURES FOR A HALF-PLANE WITH CURVILINEAR CUTS Bitsadze L. I. Vekua Institute of Appied Mathematics of Iv. Javakhishvii Tbiisi

More information

On colorings of the Boolean lattice avoiding a rainbow copy of a poset arxiv: v1 [math.co] 21 Dec 2018

On colorings of the Boolean lattice avoiding a rainbow copy of a poset arxiv: v1 [math.co] 21 Dec 2018 On coorings of the Booean attice avoiding a rainbow copy of a poset arxiv:1812.09058v1 [math.co] 21 Dec 2018 Baázs Patkós Afréd Rényi Institute of Mathematics, Hungarian Academy of Scinces H-1053, Budapest,

More information

Akaike Information Criterion for ANOVA Model with a Simple Order Restriction

Akaike Information Criterion for ANOVA Model with a Simple Order Restriction Akaike Information Criterion for ANOVA Mode with a Simpe Order Restriction Yu Inatsu * Department of Mathematics, Graduate Schoo of Science, Hiroshima University ABSTRACT In this paper, we consider Akaike

More information

Convergence Property of the Iri-Imai Algorithm for Some Smooth Convex Programming Problems

Convergence Property of the Iri-Imai Algorithm for Some Smooth Convex Programming Problems Convergence Property of the Iri-Imai Agorithm for Some Smooth Convex Programming Probems S. Zhang Communicated by Z.Q. Luo Assistant Professor, Department of Econometrics, University of Groningen, Groningen,

More information

NOISE-INDUCED STABILIZATION OF STOCHASTIC DIFFERENTIAL EQUATIONS

NOISE-INDUCED STABILIZATION OF STOCHASTIC DIFFERENTIAL EQUATIONS NOISE-INDUCED STABILIZATION OF STOCHASTIC DIFFERENTIAL EQUATIONS TONY ALLEN, EMILY GEBHARDT, AND ADAM KLUBALL 3 ADVISOR: DR. TIFFANY KOLBA 4 Abstract. The phenomenon of noise-induced stabiization occurs

More information

Uniprocessor Feasibility of Sporadic Tasks with Constrained Deadlines is Strongly conp-complete

Uniprocessor Feasibility of Sporadic Tasks with Constrained Deadlines is Strongly conp-complete Uniprocessor Feasibiity of Sporadic Tasks with Constrained Deadines is Strongy conp-compete Pontus Ekberg and Wang Yi Uppsaa University, Sweden Emai: {pontus.ekberg yi}@it.uu.se Abstract Deciding the feasibiity

More information

The Relationship Between Discrete and Continuous Entropy in EPR-Steering Inequalities. Abstract

The Relationship Between Discrete and Continuous Entropy in EPR-Steering Inequalities. Abstract The Reationship Between Discrete and Continuous Entropy in EPR-Steering Inequaities James Schneeoch 1 1 Department of Physics and Astronomy, University of Rochester, Rochester, NY 14627 arxiv:1312.2604v1

More information

Problem set 6 The Perron Frobenius theorem.

Problem set 6 The Perron Frobenius theorem. Probem set 6 The Perron Frobenius theorem. Math 22a4 Oct 2 204, Due Oct.28 In a future probem set I want to discuss some criteria which aow us to concude that that the ground state of a sef-adjoint operator

More information

The Construction of a Pfaff System with Arbitrary Piecewise Continuous Characteristic Power-Law Functions

The Construction of a Pfaff System with Arbitrary Piecewise Continuous Characteristic Power-Law Functions Differentia Equations, Vo. 41, No. 2, 2005, pp. 184 194. Transated from Differentsia nye Uravneniya, Vo. 41, No. 2, 2005, pp. 177 185. Origina Russian Text Copyright c 2005 by Izobov, Krupchik. ORDINARY

More information

Homework 5 Solutions

Homework 5 Solutions Stat 310B/Math 230B Theory of Probabiity Homework 5 Soutions Andrea Montanari Due on 2/19/2014 Exercise [5.3.20] 1. We caim that n 2 [ E[h F n ] = 2 n i=1 A i,n h(u)du ] I Ai,n (t). (1) Indeed, integrabiity

More information

Completion. is dense in H. If V is complete, then U(V) = H.

Completion. is dense in H. If V is complete, then U(V) = H. Competion Theorem 1 (Competion) If ( V V ) is any inner product space then there exists a Hibert space ( H H ) and a map U : V H such that (i) U is 1 1 (ii) U is inear (iii) UxUy H xy V for a xy V (iv)

More information

Efficiently Generating Random Bits from Finite State Markov Chains

Efficiently Generating Random Bits from Finite State Markov Chains 1 Efficienty Generating Random Bits from Finite State Markov Chains Hongchao Zhou and Jehoshua Bruck, Feow, IEEE Abstract The probem of random number generation from an uncorreated random source (of unknown

More information

Secure Information Flow Based on Data Flow Analysis

Secure Information Flow Based on Data Flow Analysis SSN 746-7659, Engand, UK Journa of nformation and Computing Science Vo., No. 4, 007, pp. 5-60 Secure nformation Fow Based on Data Fow Anaysis Jianbo Yao Center of nformation and computer, Zunyi Norma Coege,

More information

Efficient Generation of Random Bits from Finite State Markov Chains

Efficient Generation of Random Bits from Finite State Markov Chains Efficient Generation of Random Bits from Finite State Markov Chains Hongchao Zhou and Jehoshua Bruck, Feow, IEEE Abstract The probem of random number generation from an uncorreated random source (of unknown

More information

Some Measures for Asymmetry of Distributions

Some Measures for Asymmetry of Distributions Some Measures for Asymmetry of Distributions Georgi N. Boshnakov First version: 31 January 2006 Research Report No. 5, 2006, Probabiity and Statistics Group Schoo of Mathematics, The University of Manchester

More information

Small generators of function fields

Small generators of function fields Journa de Théorie des Nombres de Bordeaux 00 (XXXX), 000 000 Sma generators of function fieds par Martin Widmer Résumé. Soit K/k une extension finie d un corps goba, donc K contient un éément primitif

More information

Another Class of Admissible Perturbations of Special Expressions

Another Class of Admissible Perturbations of Special Expressions Int. Journa of Math. Anaysis, Vo. 8, 014, no. 1, 1-8 HIKARI Ltd, www.m-hikari.com http://dx.doi.org/10.1988/ijma.014.31187 Another Cass of Admissibe Perturbations of Specia Expressions Jerico B. Bacani

More information

Theory of Generalized k-difference Operator and Its Application in Number Theory

Theory of Generalized k-difference Operator and Its Application in Number Theory Internationa Journa of Mathematica Anaysis Vo. 9, 2015, no. 19, 955-964 HIKARI Ltd, www.m-hiari.com http://dx.doi.org/10.12988/ijma.2015.5389 Theory of Generaized -Difference Operator and Its Appication

More information

MARKOV CHAINS AND MARKOV DECISION THEORY. Contents

MARKOV CHAINS AND MARKOV DECISION THEORY. Contents MARKOV CHAINS AND MARKOV DECISION THEORY ARINDRIMA DATTA Abstract. In this paper, we begin with a forma introduction to probabiity and expain the concept of random variabes and stochastic processes. After

More information

An approximate method for solving the inverse scattering problem with fixed-energy data

An approximate method for solving the inverse scattering problem with fixed-energy data J. Inv. I-Posed Probems, Vo. 7, No. 6, pp. 561 571 (1999) c VSP 1999 An approximate method for soving the inverse scattering probem with fixed-energy data A. G. Ramm and W. Scheid Received May 12, 1999

More information

Intuitionistic Fuzzy Optimization Technique for Nash Equilibrium Solution of Multi-objective Bi-Matrix Games

Intuitionistic Fuzzy Optimization Technique for Nash Equilibrium Solution of Multi-objective Bi-Matrix Games Journa of Uncertain Systems Vo.5, No.4, pp.27-285, 20 Onine at: www.jus.org.u Intuitionistic Fuzzy Optimization Technique for Nash Equiibrium Soution of Muti-objective Bi-Matri Games Prasun Kumar Naya,,

More information

u(x) s.t. px w x 0 Denote the solution to this problem by ˆx(p, x). In order to obtain ˆx we may simply solve the standard problem max x 0

u(x) s.t. px w x 0 Denote the solution to this problem by ˆx(p, x). In order to obtain ˆx we may simply solve the standard problem max x 0 Bocconi University PhD in Economics - Microeconomics I Prof M Messner Probem Set 4 - Soution Probem : If an individua has an endowment instead of a monetary income his weath depends on price eves In particuar,

More information

Coupling of LWR and phase transition models at boundary

Coupling of LWR and phase transition models at boundary Couping of LW and phase transition modes at boundary Mauro Garaveo Dipartimento di Matematica e Appicazioni, Università di Miano Bicocca, via. Cozzi 53, 20125 Miano Itay. Benedetto Piccoi Department of

More information

XSAT of linear CNF formulas

XSAT of linear CNF formulas XSAT of inear CN formuas Bernd R. Schuh Dr. Bernd Schuh, D-50968 Kön, Germany; bernd.schuh@netcoogne.de eywords: compexity, XSAT, exact inear formua, -reguarity, -uniformity, NPcompeteness Abstract. Open

More information

ON THE POSITIVITY OF SOLUTIONS OF SYSTEMS OF STOCHASTIC PDES

ON THE POSITIVITY OF SOLUTIONS OF SYSTEMS OF STOCHASTIC PDES ON THE POSITIVITY OF SOLUTIONS OF SYSTEMS OF STOCHASTIC PDES JACKY CRESSON 1,2, MESSOUD EFENDIEV 3, AND STEFANIE SONNER 3,4 On the occasion of the 75 th birthday of Prof. Dr. Dr.h.c. Wofgang L. Wendand

More information

First-Order Corrections to Gutzwiller s Trace Formula for Systems with Discrete Symmetries

First-Order Corrections to Gutzwiller s Trace Formula for Systems with Discrete Symmetries c 26 Noninear Phenomena in Compex Systems First-Order Corrections to Gutzwier s Trace Formua for Systems with Discrete Symmetries Hoger Cartarius, Jörg Main, and Günter Wunner Institut für Theoretische

More information

Age of Information: The Gamma Awakening

Age of Information: The Gamma Awakening Age of Information: The Gamma Awakening Eie Najm and Rajai Nasser LTHI, EPFL, Lausanne, Switzerand Emai: {eie.najm, rajai.nasser}@epf.ch arxiv:604.086v [cs.it] 5 Apr 06 Abstract Status update systems is

More information

6 Wave Equation on an Interval: Separation of Variables

6 Wave Equation on an Interval: Separation of Variables 6 Wave Equation on an Interva: Separation of Variabes 6.1 Dirichet Boundary Conditions Ref: Strauss, Chapter 4 We now use the separation of variabes technique to study the wave equation on a finite interva.

More information

Componentwise Determination of the Interval Hull Solution for Linear Interval Parameter Systems

Componentwise Determination of the Interval Hull Solution for Linear Interval Parameter Systems Componentwise Determination of the Interva Hu Soution for Linear Interva Parameter Systems L. V. Koev Dept. of Theoretica Eectrotechnics, Facuty of Automatics, Technica University of Sofia, 1000 Sofia,

More information

ProblemsWeCanSolveWithaHelper

ProblemsWeCanSolveWithaHelper ITW 2009, Volos, Greece, June 10-12, 2009 ProblemsWeCanSolveWitha Haim Permuter Ben-Gurion University of the Negev haimp@bgu.ac.il Yossef Steinberg Technion - IIT ysteinbe@ee.technion.ac.il Tsachy Weissman

More information

arxiv: v1 [math.fa] 23 Aug 2018

arxiv: v1 [math.fa] 23 Aug 2018 An Exact Upper Bound on the L p Lebesgue Constant and The -Rényi Entropy Power Inequaity for Integer Vaued Random Variabes arxiv:808.0773v [math.fa] 3 Aug 08 Peng Xu, Mokshay Madiman, James Mebourne Abstract

More information

Limited magnitude error detecting codes over Z q

Limited magnitude error detecting codes over Z q Limited magnitude error detecting codes over Z q Noha Earief choo of Eectrica Engineering and Computer cience Oregon tate University Corvais, OR 97331, UA Emai: earief@eecsorstedu Bea Bose choo of Eectrica

More information

MIXING AUTOMORPHISMS OF COMPACT GROUPS AND A THEOREM OF SCHLICKEWEI

MIXING AUTOMORPHISMS OF COMPACT GROUPS AND A THEOREM OF SCHLICKEWEI MIXING AUTOMORPHISMS OF COMPACT GROUPS AND A THEOREM OF SCHLICKEWEI KLAUS SCHMIDT AND TOM WARD Abstract. We prove that every mixing Z d -action by automorphisms of a compact, connected, abeian group is

More information

On the Goal Value of a Boolean Function

On the Goal Value of a Boolean Function On the Goa Vaue of a Booean Function Eric Bach Dept. of CS University of Wisconsin 1210 W. Dayton St. Madison, WI 53706 Lisa Heerstein Dept of CSE NYU Schoo of Engineering 2 Metrotech Center, 10th Foor

More information

UNIFORM LIPSCHITZ CONTINUITY OF SOLUTIONS TO DISCRETE MORSE FLOWS

UNIFORM LIPSCHITZ CONTINUITY OF SOLUTIONS TO DISCRETE MORSE FLOWS Internationa Journa of Pure and Appied Mathematics Voume 67 No., 93-3 UNIFORM LIPSCHITZ CONTINUITY OF SOLUTIONS TO DISCRETE MORSE FLOWS Yoshihiko Yamaura Department of Mathematics Coege of Humanities and

More information

MATH 172: MOTIVATION FOR FOURIER SERIES: SEPARATION OF VARIABLES

MATH 172: MOTIVATION FOR FOURIER SERIES: SEPARATION OF VARIABLES MATH 172: MOTIVATION FOR FOURIER SERIES: SEPARATION OF VARIABLES Separation of variabes is a method to sove certain PDEs which have a warped product structure. First, on R n, a inear PDE of order m is

More information

A. Distribution of the test statistic

A. Distribution of the test statistic A. Distribution of the test statistic In the sequentia test, we first compute the test statistic from a mini-batch of size m. If a decision cannot be made with this statistic, we keep increasing the mini-batch

More information

4 Separation of Variables

4 Separation of Variables 4 Separation of Variabes In this chapter we describe a cassica technique for constructing forma soutions to inear boundary vaue probems. The soution of three cassica (paraboic, hyperboic and eiptic) PDE

More information

Optimality of Gaussian Fronthaul Compression for Uplink MIMO Cloud Radio Access Networks

Optimality of Gaussian Fronthaul Compression for Uplink MIMO Cloud Radio Access Networks Optimaity of Gaussian Fronthau Compression for Upink MMO Coud Radio Access etworks Yuhan Zhou, Yinfei Xu, Jun Chen, and Wei Yu Department of Eectrica and Computer Engineering, University of oronto, Canada

More information

JENSEN S OPERATOR INEQUALITY FOR FUNCTIONS OF SEVERAL VARIABLES

JENSEN S OPERATOR INEQUALITY FOR FUNCTIONS OF SEVERAL VARIABLES PROCEEDINGS OF THE AMERICAN MATHEMATICAL SOCIETY Voume 128, Number 7, Pages 2075 2084 S 0002-99390005371-5 Artice eectronicay pubished on February 16, 2000 JENSEN S OPERATOR INEQUALITY FOR FUNCTIONS OF

More information

C. Fourier Sine Series Overview

C. Fourier Sine Series Overview 12 PHILIP D. LOEWEN C. Fourier Sine Series Overview Let some constant > be given. The symboic form of the FSS Eigenvaue probem combines an ordinary differentia equation (ODE) on the interva (, ) with a

More information

The Symmetric and Antipersymmetric Solutions of the Matrix Equation A 1 X 1 B 1 + A 2 X 2 B A l X l B l = C and Its Optimal Approximation

The Symmetric and Antipersymmetric Solutions of the Matrix Equation A 1 X 1 B 1 + A 2 X 2 B A l X l B l = C and Its Optimal Approximation The Symmetric Antipersymmetric Soutions of the Matrix Equation A 1 X 1 B 1 + A 2 X 2 B 2 + + A X B C Its Optima Approximation Ying Zhang Member IAENG Abstract A matrix A (a ij) R n n is said to be symmetric

More information

Establishment of Weak Conditions for Darboux- Goursat-Beudon Theorem

Establishment of Weak Conditions for Darboux- Goursat-Beudon Theorem Georgia Southern University Digita Commons@Georgia Southern Mathematica Sciences Facuty Pubications Department of Mathematica Sciences 2009 Estabishment of Weak Conditions for Darboux- Goursat-Beudon Theorem

More information

Tikhonov Regularization for Nonlinear Complementarity Problems with Approximative Data

Tikhonov Regularization for Nonlinear Complementarity Problems with Approximative Data Internationa Mathematica Forum, 5, 2010, no. 56, 2787-2794 Tihonov Reguarization for Noninear Compementarity Probems with Approximative Data Nguyen Buong Vietnamese Academy of Science and Technoogy Institute

More information

Symbolic models for nonlinear control systems using approximate bisimulation

Symbolic models for nonlinear control systems using approximate bisimulation Symboic modes for noninear contro systems using approximate bisimuation Giordano Poa, Antoine Girard and Pauo Tabuada Abstract Contro systems are usuay modeed by differentia equations describing how physica

More information

Improving the Reliability of a Series-Parallel System Using Modified Weibull Distribution

Improving the Reliability of a Series-Parallel System Using Modified Weibull Distribution Internationa Mathematica Forum, Vo. 12, 217, no. 6, 257-269 HIKARI Ltd, www.m-hikari.com https://doi.org/1.12988/imf.217.611155 Improving the Reiabiity of a Series-Parae System Using Modified Weibu Distribution

More information

Numerical methods for elliptic partial differential equations Arnold Reusken

Numerical methods for elliptic partial differential equations Arnold Reusken Numerica methods for eiptic partia differentia equations Arnod Reusken Preface This is a book on the numerica approximation of partia differentia equations. On the next page we give an overview of the

More information

Stochastic Variational Inference with Gradient Linearization

Stochastic Variational Inference with Gradient Linearization Stochastic Variationa Inference with Gradient Linearization Suppementa Materia Tobias Pötz * Anne S Wannenwetsch Stefan Roth Department of Computer Science, TU Darmstadt Preface In this suppementa materia,

More information

Supplementary Appendix (not for publication) for: The Value of Network Information

Supplementary Appendix (not for publication) for: The Value of Network Information Suppementary Appendix not for pubication for: The Vaue of Network Information Itay P. Fainmesser and Andrea Gaeotti September 6, 03 This appendix incudes the proof of Proposition from the paper "The Vaue

More information

TitleCryptanalysis of the Quaternion Rai. IEICE Transactions on Fundamentals.

TitleCryptanalysis of the Quaternion Rai. IEICE Transactions on Fundamentals. TiteCryptanaysis of the Quaternion Rai Author(s Hashimoto, Yasufumi Citation IEICE Transactions on Fundamentas Communications and Computer Science Issue Date 205-0-0 URL http://hd.hande.net/20.500.2000/

More information

Partial permutation decoding for MacDonald codes

Partial permutation decoding for MacDonald codes Partia permutation decoding for MacDonad codes J.D. Key Department of Mathematics and Appied Mathematics University of the Western Cape 7535 Bevie, South Africa P. Seneviratne Department of Mathematics

More information

An Extension of Almost Sure Central Limit Theorem for Order Statistics

An Extension of Almost Sure Central Limit Theorem for Order Statistics An Extension of Amost Sure Centra Limit Theorem for Order Statistics T. Bin, P. Zuoxiang & S. Nadarajah First version: 6 December 2007 Research Report No. 9, 2007, Probabiity Statistics Group Schoo of

More information

PHYS 110B - HW #1 Fall 2005, Solutions by David Pace Equations referenced as Eq. # are from Griffiths Problem statements are paraphrased

PHYS 110B - HW #1 Fall 2005, Solutions by David Pace Equations referenced as Eq. # are from Griffiths Problem statements are paraphrased PHYS 110B - HW #1 Fa 2005, Soutions by David Pace Equations referenced as Eq. # are from Griffiths Probem statements are paraphrased [1.] Probem 6.8 from Griffiths A ong cyinder has radius R and a magnetization

More information

The ordered set of principal congruences of a countable lattice

The ordered set of principal congruences of a countable lattice The ordered set of principa congruences of a countabe attice Gábor Czédi To the memory of András P. Huhn Abstract. For a attice L, et Princ(L) denote the ordered set of principa congruences of L. In a

More information

Uniformly Reweighted Belief Propagation: A Factor Graph Approach

Uniformly Reweighted Belief Propagation: A Factor Graph Approach Uniformy Reweighted Beief Propagation: A Factor Graph Approach Henk Wymeersch Chamers University of Technoogy Gothenburg, Sweden henkw@chamers.se Federico Penna Poitecnico di Torino Torino, Itay federico.penna@poito.it

More information

Math-Net.Ru All Russian mathematical portal

Math-Net.Ru All Russian mathematical portal Math-Net.Ru A Russian mathematica porta D. Zaora, On properties of root eements in the probem on sma motions of viscous reaxing fuid, Zh. Mat. Fiz. Ana. Geom., 217, Voume 13, Number 4, 42 413 DOI: https://doi.org/1.1547/mag13.4.42

More information

T.C. Banwell, S. Galli. {bct, Telcordia Technologies, Inc., 445 South Street, Morristown, NJ 07960, USA

T.C. Banwell, S. Galli. {bct, Telcordia Technologies, Inc., 445 South Street, Morristown, NJ 07960, USA ON THE SYMMETRY OF THE POWER INE CHANNE T.C. Banwe, S. Gai {bct, sgai}@research.tecordia.com Tecordia Technoogies, Inc., 445 South Street, Morristown, NJ 07960, USA Abstract The indoor power ine network

More information

Rate-Distortion Theory of Finite Point Processes

Rate-Distortion Theory of Finite Point Processes Rate-Distortion Theory of Finite Point Processes Günther Koiander, Dominic Schuhmacher, and Franz Hawatsch, Feow, IEEE Abstract We study the compression of data in the case where the usefu information

More information

Alberto Maydeu Olivares Instituto de Empresa Marketing Dept. C/Maria de Molina Madrid Spain

Alberto Maydeu Olivares Instituto de Empresa Marketing Dept. C/Maria de Molina Madrid Spain CORRECTIONS TO CLASSICAL PROCEDURES FOR ESTIMATING THURSTONE S CASE V MODEL FOR RANKING DATA Aberto Maydeu Oivares Instituto de Empresa Marketing Dept. C/Maria de Moina -5 28006 Madrid Spain Aberto.Maydeu@ie.edu

More information

Higher dimensional PDEs and multidimensional eigenvalue problems

Higher dimensional PDEs and multidimensional eigenvalue problems Higher dimensiona PEs and mutidimensiona eigenvaue probems 1 Probems with three independent variabes Consider the prototypica equations u t = u (iffusion) u tt = u (W ave) u zz = u (Lapace) where u = u

More information

UNIFORM CONVERGENCE OF MULTIPLIER CONVERGENT SERIES

UNIFORM CONVERGENCE OF MULTIPLIER CONVERGENT SERIES royecciones Vo. 26, N o 1, pp. 27-35, May 2007. Universidad Catóica de Norte Antofagasta - Chie UNIFORM CONVERGENCE OF MULTILIER CONVERGENT SERIES CHARLES SWARTZ NEW MEXICO STATE UNIVERSITY Received :

More information

#A48 INTEGERS 12 (2012) ON A COMBINATORIAL CONJECTURE OF TU AND DENG

#A48 INTEGERS 12 (2012) ON A COMBINATORIAL CONJECTURE OF TU AND DENG #A48 INTEGERS 12 (2012) ON A COMBINATORIAL CONJECTURE OF TU AND DENG Guixin Deng Schoo of Mathematica Sciences, Guangxi Teachers Education University, Nanning, P.R.China dengguixin@ive.com Pingzhi Yuan

More information

Research Article Numerical Range of Two Operators in Semi-Inner Product Spaces

Research Article Numerical Range of Two Operators in Semi-Inner Product Spaces Abstract and Appied Anaysis Voume 01, Artice ID 846396, 13 pages doi:10.1155/01/846396 Research Artice Numerica Range of Two Operators in Semi-Inner Product Spaces N. K. Sahu, 1 C. Nahak, 1 and S. Nanda

More information

Continued fractions with low complexity: Transcendence measures and quadratic approximation. Yann BUGEAUD

Continued fractions with low complexity: Transcendence measures and quadratic approximation. Yann BUGEAUD Continued fractions with ow compexity: Transcendence measures and quadratic approximation Yann BUGEAUD Abstract. We estabish measures of non-quadraticity and transcendence measures for rea numbers whose

More information

Minimum Enclosing Circle of a Set of Fixed Points and a Mobile Point

Minimum Enclosing Circle of a Set of Fixed Points and a Mobile Point Minimum Encosing Circe of a Set of Fixed Points and a Mobie Point Aritra Banik 1, Bhaswar B. Bhattacharya 2, and Sandip Das 1 1 Advanced Computing and Microeectronics Unit, Indian Statistica Institute,

More information

Asymptotic Properties of a Generalized Cross Entropy Optimization Algorithm

Asymptotic Properties of a Generalized Cross Entropy Optimization Algorithm 1 Asymptotic Properties of a Generaized Cross Entropy Optimization Agorithm Zijun Wu, Michae Koonko, Institute for Appied Stochastics and Operations Research, Caustha Technica University Abstract The discrete

More information

Contribution to a formulation of integral type of the mechanics of Cosserat continua ( * ).

Contribution to a formulation of integral type of the mechanics of Cosserat continua ( * ). ontributo per una formuazione di tipo integrae dea meccanica dei continui di osserat Ann Mat pura ed App (976) 75-83 ontribution to a formuation of integra type of the mechanics of osserat continua ( *

More information

Math 124B January 17, 2012

Math 124B January 17, 2012 Math 124B January 17, 212 Viktor Grigoryan 3 Fu Fourier series We saw in previous ectures how the Dirichet and Neumann boundary conditions ead to respectivey sine and cosine Fourier series of the initia

More information

MONOCHROMATIC LOOSE PATHS IN MULTICOLORED k-uniform CLIQUES

MONOCHROMATIC LOOSE PATHS IN MULTICOLORED k-uniform CLIQUES MONOCHROMATIC LOOSE PATHS IN MULTICOLORED k-uniform CLIQUES ANDRZEJ DUDEK AND ANDRZEJ RUCIŃSKI Abstract. For positive integers k and, a k-uniform hypergraph is caed a oose path of ength, and denoted by

More information

arxiv: v1 [math.co] 17 Dec 2018

arxiv: v1 [math.co] 17 Dec 2018 On the Extrema Maximum Agreement Subtree Probem arxiv:1812.06951v1 [math.o] 17 Dec 2018 Aexey Markin Department of omputer Science, Iowa State University, USA amarkin@iastate.edu Abstract Given two phyogenetic

More information

Steepest Descent Adaptation of Min-Max Fuzzy If-Then Rules 1

Steepest Descent Adaptation of Min-Max Fuzzy If-Then Rules 1 Steepest Descent Adaptation of Min-Max Fuzzy If-Then Rues 1 R.J. Marks II, S. Oh, P. Arabshahi Λ, T.P. Caude, J.J. Choi, B.G. Song Λ Λ Dept. of Eectrica Engineering Boeing Computer Services University

More information

A NOTE ON INFINITE DIVISIBILITY OF ZETA DISTRIBUTIONS

A NOTE ON INFINITE DIVISIBILITY OF ZETA DISTRIBUTIONS A NOTE ON INFINITE DIVISIBILITY OF ZETA DISTRIBUTIONS SHINGO SAITO AND TATSUSHI TANAKA Abstract. The Riemann zeta istribution, efine as the one whose characteristic function is the normaise Riemann zeta

More information

The arc is the only chainable continuum admitting a mean

The arc is the only chainable continuum admitting a mean The arc is the ony chainabe continuum admitting a mean Aejandro Ianes and Hugo Vianueva September 4, 26 Abstract Let X be a metric continuum. A mean on X is a continuous function : X X! X such that for

More information

Primal and dual active-set methods for convex quadratic programming

Primal and dual active-set methods for convex quadratic programming Math. Program., Ser. A 216) 159:469 58 DOI 1.17/s117-15-966-2 FULL LENGTH PAPER Prima and dua active-set methods for convex quadratic programming Anders Forsgren 1 Phiip E. Gi 2 Eizabeth Wong 2 Received:

More information

SYMMETRICAL MULTILEVEL DIVERSITY CODING AND SUBSET ENTROPY INEQUALITIES. A Dissertation JINJING JIANG

SYMMETRICAL MULTILEVEL DIVERSITY CODING AND SUBSET ENTROPY INEQUALITIES. A Dissertation JINJING JIANG SYMMETRICA MUTIEVE DIVERSITY CODING AND SUBSET ENTROPY INEQUAITIES A Dissertation by JINJING JIANG Submitted to the Office of Graduate Studies of Texas A&M University in partia fufiment of the requirements

More information

Integrality ratio for Group Steiner Trees and Directed Steiner Trees

Integrality ratio for Group Steiner Trees and Directed Steiner Trees Integraity ratio for Group Steiner Trees and Directed Steiner Trees Eran Haperin Guy Kortsarz Robert Krauthgamer Aravind Srinivasan Nan Wang Abstract The natura reaxation for the Group Steiner Tree probem,

More information

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 15, NO. 2, FEBRUARY

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 15, NO. 2, FEBRUARY IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 5, NO. 2, FEBRUARY 206 857 Optima Energy and Data Routing in Networks With Energy Cooperation Berk Gurakan, Student Member, IEEE, OmurOze,Member, IEEE,

More information

On The Binary Lossless Many-Help-One Problem with Independently Degraded Helpers

On The Binary Lossless Many-Help-One Problem with Independently Degraded Helpers On The Binary Lossless Many-Help-One Problem with Independently Degraded Helpers Albrecht Wolf, Diana Cristina González, Meik Dörpinghaus, José Cândido Silveira Santos Filho, and Gerhard Fettweis Vodafone

More information

Appendix of the Paper The Role of No-Arbitrage on Forecasting: Lessons from a Parametric Term Structure Model

Appendix of the Paper The Role of No-Arbitrage on Forecasting: Lessons from a Parametric Term Structure Model Appendix of the Paper The Roe of No-Arbitrage on Forecasting: Lessons from a Parametric Term Structure Mode Caio Ameida cameida@fgv.br José Vicente jose.vaentim@bcb.gov.br June 008 1 Introduction In this

More information

An explicit Jordan Decomposition of Companion matrices

An explicit Jordan Decomposition of Companion matrices An expicit Jordan Decomposition of Companion matrices Fermín S V Bazán Departamento de Matemática CFM UFSC 88040-900 Forianópois SC E-mai: fermin@mtmufscbr S Gratton CERFACS 42 Av Gaspard Coriois 31057

More information