A Hybrid Random-Structured Coding Scheme for the Gaussian Two-Terminal Source Coding Problem Under a Covariance Matrix Distortion Constraint
|
|
- Maude Willa Booker
- 5 years ago
- Views:
Transcription
1 A Hybrid Radom-Structured Codig Scheme for the Gaussia Two-Termial Source Codig Problem Uder a Covariace Matrix Distortio Costrait Yag Yag ad Zixiag Xiog Dept of Electrical ad Computer Egieerig Texas A&M Uiversity College Statio TX yagyag@tamu.edu ad zx@ece.tamu.edu Abstract This paper focuses o the Gaussia two-termial source codig problem uder a covariace matrix distortio costrait which subsumes the quadratic Gaussia two-termial source codig problem with idividual distortio costraits whose complete solutio is kow. Differet from existig schemes which are either radom or structured we propose a ew hybrid radom-structured scheme with a sum-rate strictly smaller tha the quatize-ad-bi (QB) upper boud i certai cases. The first layer of our scheme is a QB radom codig scheme attemptig to achieve a itermediate distortio matrix that is as symmetric as possible. The secod layer is a structured scheme that targets at recostructig a weighted differece of the observed sources coditioed o their quatized versios i the first layer. We prove that the gap betwee the sum-rate of our scheme ad its lower boud is o larger tha two bits per sample i particular this gap decreases to exactly oe bit per sample whe the source covariace matrix is symmetrifiable i the sese that the itermediate covariace matrix ca be made purely symmetric. I. INTRODUCTION Distributed source codig (DSC) of correlated sources has attracted a icreasig amout of attetio sice Slepia ad Wolf s celebrated work [ i 973. However lossy DSC is still a ope problem i geeral despite of some partial solutios provided for certai special cases. This paper cosiders a subclass of lossy DSC where two Gaussia sources are separately compressed ad joitly recostructed subject to a covariace matrix distortio costrait. This problem is of particular iterest because it is the most geeral Gaussia two-termial source codig problem i the sese that its achievable rate regio completely determies those of the followig problems (as poited out by Oohama i [). The quadratic Gaussia two-termial source codig problem with idividual MSE distortio costraits whose rate regio has bee characterized [3. The DSC problem of liear fuctios [4 [5 [6 with a sigle MSE distortio costrait o a liear fuctio of the two sources. The geeralized Gaussia CEO problem [7 with a sum MSE distortio costrait o some remote sources that are joitly Gaussia with the two observed sources. For a give source covariace matrix the target distortio matrix has three degrees of freedom. However the cove- Work supported by the NSF grat 0789 ad the Qatar Natioal Research Fud. tioal radom codig scheme with a quatize-ad-bi (QB) framework ca oly achieve distortio matrices o a twodimesioal surface parameterized by the two quatizatio oise variaces. Cosequetly amog those distortio matrices with uit diagoal elemets oly oe is achievable by a QB scheme while others require time-sharig betwee at least two QB schemes. Deote the off-diagoal elemet of the QB-achievable distortio matrix as θ. Two lower sum-rate bouds [3 [8 have bee give for the covariace matrix costraied two-termial problem. Specifically Wager [3 provided a composite sum-rate lower boud usig those for the µ-sum problem ad the cooperative twotermial problem respectively ad showed that the QB sumrate upper boud is tight for ay distortio matrix with uit diagoal elemets ad off-diagoal elemet θ that is o larger tha θ. O the other had Wager s composite lower boud degeerates to the cooperative boud which is relatively loose ad diverges from the QB upper boud as the off-diagoal elemet icreases beyod the threshold θ. Aother sumrate lower boud is provided i [8 which improves the cooperative boud by a ubouded amout whe θ > θ. However the maximum sum-rate gap betwee the improved lower boud ad the QB upper boud is also ubouded. The mai cotributio of this work is a ew hybrid radomstructured scheme that superimposes a distributed differeceformig structured codig compoet proposed by Krithivasa ad Pradha [4 o a QB radom codig scheme with a miimum sum-rate withi two b/s from the lower boud i [8. Our proposed scheme has a similar structure as Ahlswede ad Ha s scheme [9 Sectio VI which cocateates a QB radom codig scheme with Körer ad Marto s structured liear codig scheme [0 for ecodig the modulo- sum of two biary sources (that are asymmetric i geeral). The first layer of our hybrid scheme is a QB scheme where both ecoders quatize theirs sources usig radom quatizers with the resultig quatizatio idices further compressed by a pair of Slepia-Wolf ecoders. The at each ecoder with the availability of the quatized versio of its observed source a correlatio tuer forms a tued source by liearly combiig the observed source with its quatized versio i such a way that a certai weighted differece betwee the resultig tued sources is idepedet of the coarsely quatized versios. Next the secod layer comes i by applyig the Krithivasa
2 ad Pradha s structured scheme [4 o the tued sources to recostruct the weighted differece lossily at the decoder usig two fie lattice quatizers ad the same coarse lattice chael code. The joit decoder liearly combies the three pieces of recostructios i a MMSE maer to geerate the fial estimatios of the sources. Our scheme is parameterized by the four quatizatio oise variaces (two i each layer) ad the weight factor i the secod layer. By assumig the target distortio matrix to have uit diagoal elemet without loss of geerality we show that the optimal parameter set that achieves the miimum sum-rate is such that The first layer ad the correlatio tuers adjust the correlatio betwee the tued sources such that the itermediate covariace matrix of the sources give their quatized versios is as symmetric as possible. If the source covariace matrix is symmetrifiable i the sese that the itermediate covariace matrix has equal diagoal elemets the gap betwee the sum-rate of our hybrid scheme ad the lower boud i [8 is exactly oe b/s while this gap is capped at two b/s i geeral. A ituitive reaso for the first layer to symmetrify the itermediate covariace matrix is that the secod structured layer is most efficiet i the symmetric case (withi oe b/s from optimality [4 [5). Compared to the QB sum-rate boud the sum-rate savig i our hybrid scheme is ubouded. However sice the secod layer of our scheme has a miimum of oe b/s trasmissio rate there are cases whe the QB sum-rate is strictly smaller. Moreover to get a covex composite sum-rate upper boud time-sharig betwee QB schemes ad our scheme is required with a maximum of three differet schemes ivolved due to Carathéodory s Theorem [. II. PROBLEM SETUP Let Y ad Y be joitly Gaussia with zero mea ad covariace matrix [ σ Σ Y ρσ σ ρσ σ σ ad Y (Y Y... Y ) Y (Y Y... Y ) be two legth- blocks of idepedet drawigs of Y ad Y respectively. Each of the two ecoders seds a fuctio of their ow source amely φ () i : R... M () i } for i. The decoder attempts to recostruct Y ad Y usig fuctio ψ () :... M () }... M () } R R. Deote the recostructed versios as Ŷ i i. For a target distortio matrix D [ θ θ () Note that without loss of geerality we assume i this paper that the target distortio matrix D have uit diagoal elemets ad a off-diagoal elemet θ sice otherwise oe ca set Σ Y µσ Y µ T ad D µdµ T with µ diag(([d ) ([D ) ) to obtai a equivalet problem with D takig the form of (). such that 0 D Σ Y we say a triple (R R D) is achievable if there exists a sequece of schemes (φ () φ() ψ() ) such that lim sup log M () i R i i lim sup i Cov(Y Y W W ) D ad deote the rate-distortio regio RD as the covex closure of the set of all achievable triple i.e. } RD cov (R R D) is achievable. The achievable rate regio is defied as R (D) (R R ) : (R R D) RD }. Ad the miimum sum-rate is defied as R (D) mi (R R ) R (D) R R. A. Existig lower ad upper bouds o R (D) For ay pair of Σ Y ad D oe ca always apply the covetioal QB scheme that employs idepedet Gaussia quatizatio followed by biig (Slepia-Wolf codig) to obtai the followig rate-distortio regio [ (R RD QB cov R Cov(Y Y G G ) ) : (G G ) U(Y Y ) R I(Y ; G G ) R I(Y ; G G ) } R R I(Y Y ; G G ). where cov( ) meas covex closure ad U(Y Y ) cotais all (G G ) pairs such that G Y N (0 σq ) G Y N (0 σq ) are idepedetly corrupted versios of Y ad Y. Cosequetly the QB ier rate regio ad the QB sum-rate upper boud are defied as R QB (D) (R R ) : (R R D) RD QB} R QB (D) mi R R (R R ) R QB (D) with R QB (D) R (D) R QB (D) R (D). O the other had Wager et al. [3 provided the followig composite sum-rate lower boud (which was used to prove the QB sum-rate tightess i the origial quadratic Gaussia two-termial problem) that combies a µ-sum boud with the cooperative boud } R (D) R W (D) max R µ (D) R coop (D) R µ (D) log (( ρ )σ σ ρσ σ ( θ)) ( θ) R coop (D) log ( ρ )σσ θ. This lower boud was partially improved i [8 as R (D) R(D) Rµ (D) θ θ R lb (D) θ > θ
3 where ( θ ρ ) σ σ 4ρ ( ρ )σ σ ρ R lb (D) log ( ρ ) σ 3 σ 3 ( θ) (( ρ )σ σ ρ( θ)). I additio R(D) is partially tight whe θ θ sice R(D) R W (D) R (D) R QB (D) whe θ θ. III. THE PROPOSED HYBRID RANDOM-STRUCTURED SCHEME We describe a hybrid radom-structured codig scheme ad give a ew ier rate-distortio regio RD for the covariace matrix costraied Gaussia two-termial problem. Let Λ Λ ad Λ C be three -dimesioal lattices with ormalized secod momets q q ad q C respectively such that Λ i Λ C i. Assume that (Λ Λ ) are good for source codig while Λ C good for chael codig. Also defie the dithered quatizer associated with a -dimesioal lattice Λ ad a legth- dither sequece T as Q Λ(T )(x ) arg mi q Λ x T q T. The block diagram of our proposed two-layer scheme is depicted i Fig.. I the first layer of our scheme the two ecoders quatize Y ad Y usig two radom quatizers with auxiliary radom variables U i Y i P i i () where P i N (0 p i ) are idepedet additive Gaussia oises. The idices of the codewords U ad U are compressed by a pair of Slepia-Wolf ecoders to trasmissio rates of R () ad R () respectively. Y Y Correlatio Tuer Correlatio Tuer Z Z Ecoder Radom Quatizer I Q (T ) U Ecoder c Fig.. Q (T ) Radom Quatizer II U V V Slepia-Wolf Ecoder I Slepia-Wolf Ecoder II R R R R Slepia-Wolf Decoder Decoder Block diagram of our proposed scheme. ^Z Liear Estimator ^U ; ^U The at the each ecoder a correlatio tuer computes where Z i Y i λ i U i i ( ρ )σ λ σ p (σ cρσ ) ( ρ )σ σ p σ p σ p p λ ( ρ )cσσ p σ (cσ ρσ ) ( ρ )σ σ p σ p σ p p ^Y ^Y with c > 0 beig a positive umber. The resultig tued sources (Z Z ) are scaled ad quatized usig aother two dithered lattices quatizers as Ṽ Q Λ(T ) (Z ) Ṽ Q Λ(T ) (cz ) with T ad T beig legth- radom dithers uiformly distributed over the Vorooi regios of Λ ad Λ respectively. The idices of Si Ṽ i mod Λ C i are set to the decoder usig trasmissio rates R () i log q C qi i. The decoder also cosists of two layers. The first layer recostructs Û ad Û usig the Slepia-Wolf Decoder while the secod layer receives S ad S ad computes Ẑ (S S ) mod Λ C. The last step at the decoder recostructs Y ad Y as Ŷ i α i Û α i Û α i3 Ẑ i (3) where α ij s are liear MMSE estimatio coefficiets of (Y Y ) give (U U ) i () ad Z V V with V Z Q V cz Q Z i Yi λ i U i (4) ad Q Q beig idepedet additive Gaussia oises with variaces q ad q respectively. Remarks: The correlatio tuers betwee the two layers of the ecoders select λ ad λ such that Z cz (Y cy ) E [ Y cy U U i.e. they completely remove the portio of Y cy that is correlated to (U U ) which ca be added back at the decoder oce (U U ) is recovered. There is o loss of geerality by usig a sigle weight factor c o Y sice recostructig Y cy as Ẑ is equivalet to restorig Y c Y as Ẑ c. A. Performace of our proposed scheme Our proposed scheme provides the followig ier ratedistortio regio with detailed proof give i Sectio IV. Theorem : It holds that RD RD where RD cotais all triples (R R D) satisfyig R I(Y ; U U ) I(Z; Y V U U ) R I(Y ; U U ) I(Z; V Y U U ) R R I(Y Y ; U U ) I(Z; Y V U U ) I(Z; V Y U U ) for some (U U Z V V ) defied i () ad (4) such that Cov(Y Y U U Z) D. A atural corollary of Theorem is that the covex closure of the QB rate-distortio regio RD QB ad the rate-distortio regio RD of our ew scheme is a subset of RD.
4 Corollary : It holds that ( cov RD RD QB) RD. (5) O the other had the rate-distortio regio RD of our ew scheme gives a sum-rate upper boud stated i the followig theorem. The proof is also outlied i Sectio IV. Theorem : For a D i () with d d ad θ θ it holds that R(D) R(D) (6) log 4σ 3 σ3 ( ρ ) ( θ) (( ρ )σ σ ρ(θ)) δ δ 0 log 4σ 4 σ4 ( ρ ) (( ρ )σ ( θ )) ( θ ) (( ρ )σ σ σ ρσ θρσσ) δ <0 δ 0 log 4σ 4 σ4 ( ρ ) (( ρ )σ ( θ )) ( θ ) (( ρ )σ σ σ ρσ θρσσ) δ 0 δ <0 where for i j } i j δ i σ i σ j( ρ )( θ) ( ρ )σ i σ j (σ j ρσ i )( θ). Comparig our ew sum-rate upper boud with the lower boud R(D) we obtaied the followig theorem which states that the sum-rate gap betwee R(D) ad R(D) is o larger tha two b/s sadwichig the sum-rate-distortio fuctio R(D) i a costat ad relatively small rage. The detailed proof is omitted. Theorem 3: For a D i () with d d ad θ θ it holds that R(D) R(D). (7) Furthermore the followig theorem states that the gap betwee R(D) ad R(D) is exactly oe b/s i a special symmetrifiable case. Theorem 4: If D takes the form i () with d d ad θ θ ad Σ Y is symmetrifiable i the sese that there exists p p 0 such that [Σ Z (p p ) [Σ Z (p p ) (8) [ with Σ Z (p p ) Σ Y diag( p p ) the R(D) R(D). (9) IV. PROOF OF MAIN RESULTS This sectio cotais proof outlies of Theorems ad. A. Proof of Theorem Proof: Before provig Theorem we states two lemmas with detailed proofs omitted. Lemma : There exist sequeces of lattices Λ i ()} N i j that are good for source codig i the sese that Ṽ i Ẑ Ṽ Ṽ Z V i i with (Z V V ) defied i (4) ad X Y meas X coverges to Y i the Kullback-Leibler divergece sese as. Lemma : There exist a sequece of lattices Λ C ()} N that is good for chael codig i the sese that q C Var(Z) ad Pr(Û i Ui ) 0 i Pr(Ẑ Ṽ Ṽ ) 0. Cosequetly for Ŷ i s defied i (3) Cov(Y Ŷ Y Ŷ ) Cov(Y Y U U Z). (0) To prove Theorem we ca simply apply Lemmas ad to show achievability of (R R D) with R I(Y ; U ) I(Z; Y V U U ) R I(Y ; U U ) I(Z; V Y U U ). The usig the same techique we ca prove that the dual scheme ca achieve (R R D) with R I(Y ; U U ) I(Z; Y V U U ) R I(Y ; U ) I(Z; V Y U U ) ad Theorem follows by takig the covex closure of (R R D) ad (R R D) triples. B. Proof of Theorem Proof: With Theorem we oly eed to provide proper (c p p q q ) parameters for the three differet cases. Whe δ δ 0 let c () σ σ (p () p() ) (δ δ ) q () σ σ ( ρ )( θ ) i (θσ σ ( ρ ) ρ( θ i. )) Whe δ < 0 ad δ 0 let p () c () θσ σ ( ρ ) ρ( θ ) σ ( ρ ) ( θ ) p () σ (( ρ )σ ( θ )) (σ )(σ ) ρ σ σ θ(ρσ σ θ) q () σ i ( ρ )( θ ) (( ρ )σ ( i. θ )) Whe δ < 0 ad δ 0 let p (3) c (3) ( ρ )σ ρ( θ ) θσ σ ( ρ ) ρ( θ ) p (3) σ (( ρ )σ ( θ )) (σ )(σ ) ρ σ σ θ(ρσ σ θ) q (3) i σ ( ρ )( θ )(( ρ )σ ( θ )) (θσ σ ( ρ ) ρ( θ )) i. The verificatios of (6) is straightforward.
5 ρ V. NUMERICAL EXAMPLES Fig. compares the sum-rate lower ad upper bouds R(D) R W (D) R QB (D) R(D) ad R T S (D) respectively with σ 0 σ 0 ρ 0.9 d d ad ( R T S (D) mi R R :(R R D) cov RD RD QB)}. We observe that for this case R(D) does ot directly improves R QB (D). However by time-sharig betwee our proposed two-layer scheme ad the QB scheme R T S (D) is strictly smaller tha R QB (D) for all θ > θ. I additio due to Theorem 3 the gap betwee R(D) ad R(D) is o larger tha two b/s. sum-rate (b/s) BT sum-rate upper boud R BT (D) New sum-rate upper boud R(D) Sum-rate upper boud R TS (D) w/ time-sharig Sum-rate lower boud R(D) Wager et al. s composite lower boud R W (D) θ θ b/s θ Fig.. A compariso amog differet lower ad upper sum-rate bouds whe σ 0 σ 0 ρ 0.9 ad d d. Aother compariso whe σ 00 9 σ 0 ρ 0.99 ad d d. We observe that for this case R(D) itself already improves R QB (D). Ad oe ca show that (ρ D) is always symmetrifiable for ay θ > θ hece due to Theorem 4 the gap betwee R(D) ad R(D) is exact oe b/s. sum-rate (b/s) BT upper boud R BT (D) New sum-rate upper boud R(D) Sum-rate upper boud R TS (D) w/ time-sharig Sum-rate lower boud R(D) Wager et al. s composite sum-rate lower boud θ θ b/s symmetrifiable θ Fig. 3. A compariso amog differet lower ad upper sum-rate bouds whe σ 00 9 σ 0 ρ 0.99 ad d d. Fig. 4 shows the ( σ σ ) regio where R(D) < R QB (D) for some θ whe ρ is fixed. Aalytical forms of the boudaries are very hard to compute hece the plot is obtaied through umerical comparisos betwee R(D) ad R QB (D). We observe that i order for R(D) < R QB (D) to hold ρ must be at least σ Fig. 4. ( fixed. 0.4 σ ρ 0.6 ρ 0.65 ρ 0.65 ρ ρ 0.7 ρ 0.75 ρ 0.75 ρ ρ 0.8 ρ 0.85 ρ 0.85 ρ ρ 0.9 ρ 0.95 ρ 0.95 ρ ρ σ σ ) regio where R(D) < R QB (D) for some θ whe ρ is Oe ca show that as ρ goes to oe the maximum sumrate savig betwee R(D) ad R QB (D) is ubouded with detailed proof omitted. It is worth otig that R T S (D) ca be achieved by at most three differet schemes due to Carathéodory s Theorem [. VI. CONCLUSIONS We have proposed a two-layer structured scheme for the covariace matrix costraied Gaussia two-termial source codig problem. The first layer is a QB compoet that adjusts the source correlatio while the secod layer directly aims at recostructig a weighted differece of the sources give the iformatio set from the first layer to the decoder. By choosig appropriate parameters the proposed scheme ca strictly improve the QB rate-distortio regio with a sumrate that is withi two b/s from optimality. Coceptually our proposed hybrid scheme brigs together the two differet worlds of radom QB codig ad structured lattice codig. REFERENCES [ D. Slepia ad J. Wolf Noiseless codig of correlated iformatio sources IEEE Tras. If. Theory vol. 9 pp July 973. [ Y. Oohama Distributed source codig of correlated Gaussia sources preprit available at cache/arxiv/pdf/007/ v.pdf. [3 A. Wager S. Tavildar ad P. Viswaath The rate regio of the quadratic Gaussia two-termial source-codig problem IEEE Tras. If. Theory vol. 54 pp May 008. [4 D. Krithivasa S. S. Pradha Lattices for distributed source codig: joitly Gaussia sources ad recostructio of a liear fuctio IEEE Tras. If. Theory vol. 55 pp Dec [5 A. B. Wager O distributed compressio of liear fuctios IEEE Tras. If. Theory vol. 57 pp Jauary 0. [6 M. A. Maddah-Ali ad D. Tse Iterferece eutralizatio i distributed lossy source codig Proc. ISIT 0 Ja. 00. [7 Y. Yag ad Z. Xiog O the geeralized Gaussia CEO problem IEEE Tras. If. Theory to appear. [8 Y. Yag Y. Zhag ad Z. Xiog A ew sufficiet coditio for sum-rate tightess i quadratic Gaussia multitermial source codig submitted to IEEE Tras. If. Theory Jue 00. [9 R. Ahlswede ad T. S. Ha O source codig with side iformatio via a multiple-access chael ad related problems i multi-user iformatio theory IEEE Tras. If. Theory vol. 9 pp May 983. [0 J. Körer ad K. Marto How to ecode the modulo-two sum of biary sources IEEE Tras. Iform. Theory vol. 5 pp 9- March 979. [ T. Cover ad J. Thomas Elemets of Iformatio Theory Wiley 99.
Cooperative Communication Fundamentals & Coding Techniques
3 th ICACT Tutorial Cooperative commuicatio fudametals & codig techiques Cooperative Commuicatio Fudametals & Codig Techiques 0..4 Electroics ad Telecommuicatio Research Istitute Kiug Jug 3 th ICACT Tutorial
More informationAre Slepian-Wolf Rates Necessary for Distributed Parameter Estimation?
Are Slepia-Wolf Rates Necessary for Distributed Parameter Estimatio? Mostafa El Gamal ad Lifeg Lai Departmet of Electrical ad Computer Egieerig Worcester Polytechic Istitute {melgamal, llai}@wpi.edu arxiv:1508.02765v2
More informationFinite Block-Length Gains in Distributed Source Coding
Decoder Fiite Block-Legth Gais i Distributed Source Codig Farhad Shirai EECS Departmet Uiversity of Michiga A Arbor,USA Email: fshirai@umichedu S Sadeep Pradha EECS Departmet Uiversity of Michiga A Arbor,USA
More informationSymmetric Two-User Gaussian Interference Channel with Common Messages
Symmetric Two-User Gaussia Iterferece Chael with Commo Messages Qua Geg CSL ad Dept. of ECE UIUC, IL 680 Email: geg5@illiois.edu Tie Liu Dept. of Electrical ad Computer Egieerig Texas A&M Uiversity, TX
More informationThe Maximum-Likelihood Decoding Performance of Error-Correcting Codes
The Maximum-Lielihood Decodig Performace of Error-Correctig Codes Hery D. Pfister ECE Departmet Texas A&M Uiversity August 27th, 2007 (rev. 0) November 2st, 203 (rev. ) Performace of Codes. Notatio X,
More informationRun-length & Entropy Coding. Redundancy Removal. Sampling. Quantization. Perform inverse operations at the receiver EEE
Geeral e Image Coder Structure Motio Video (s 1,s 2,t) or (s 1,s 2 ) Natural Image Samplig A form of data compressio; usually lossless, but ca be lossy Redudacy Removal Lossless compressio: predictive
More informationLecture 27. Capacity of additive Gaussian noise channel and the sphere packing bound
Lecture 7 Ageda for the lecture Gaussia chael with average power costraits Capacity of additive Gaussia oise chael ad the sphere packig boud 7. Additive Gaussia oise chael Up to this poit, we have bee
More informationCEE 522 Autumn Uncertainty Concepts for Geotechnical Engineering
CEE 5 Autum 005 Ucertaity Cocepts for Geotechical Egieerig Basic Termiology Set A set is a collectio of (mutually exclusive) objects or evets. The sample space is the (collectively exhaustive) collectio
More informationHybrid Coding for Gaussian Broadcast Channels with Gaussian Sources
Hybrid Codig for Gaussia Broadcast Chaels with Gaussia Sources Rajiv Soudararaja Departmet of Electrical & Computer Egieerig Uiversity of Texas at Austi Austi, TX 7871, USA Email: rajivs@mailutexasedu
More informationECE 8527: Introduction to Machine Learning and Pattern Recognition Midterm # 1. Vaishali Amin Fall, 2015
ECE 8527: Itroductio to Machie Learig ad Patter Recogitio Midterm # 1 Vaishali Ami Fall, 2015 tue39624@temple.edu Problem No. 1: Cosider a two-class discrete distributio problem: ω 1 :{[0,0], [2,0], [2,2],
More information(A sequence also can be thought of as the list of function values attained for a function f :ℵ X, where f (n) = x n for n 1.) x 1 x N +k x N +4 x 3
MATH 337 Sequeces Dr. Neal, WKU Let X be a metric space with distace fuctio d. We shall defie the geeral cocept of sequece ad limit i a metric space, the apply the results i particular to some special
More informationAsymptotic Coupling and Its Applications in Information Theory
Asymptotic Couplig ad Its Applicatios i Iformatio Theory Vicet Y. F. Ta Joit Work with Lei Yu Departmet of Electrical ad Computer Egieerig, Departmet of Mathematics, Natioal Uiversity of Sigapore IMS-APRM
More informationInequalities for Entropies of Sets of Subsets of Random Variables
Iequalities for Etropies of Sets of Subsets of Radom Variables Chao Tia AT&T Labs-Research Florham Par, NJ 0792, USA. tia@research.att.com Abstract Ha s iequality o the etropy rates of subsets of radom
More informationVector Quantization: a Limiting Case of EM
. Itroductio & defiitios Assume that you are give a data set X = { x j }, j { 2,,, }, of d -dimesioal vectors. The vector quatizatio (VQ) problem requires that we fid a set of prototype vectors Z = { z
More informationInformation Theory and Coding
Sol. Iformatio Theory ad Codig. The capacity of a bad-limited additive white Gaussia (AWGN) chael is give by C = Wlog 2 ( + σ 2 W ) bits per secod(bps), where W is the chael badwidth, is the average power
More informationOn Random Line Segments in the Unit Square
O Radom Lie Segmets i the Uit Square Thomas A. Courtade Departmet of Electrical Egieerig Uiversity of Califoria Los Ageles, Califoria 90095 Email: tacourta@ee.ucla.edu I. INTRODUCTION Let Q = [0, 1] [0,
More information62. Power series Definition 16. (Power series) Given a sequence {c n }, the series. c n x n = c 0 + c 1 x + c 2 x 2 + c 3 x 3 +
62. Power series Defiitio 16. (Power series) Give a sequece {c }, the series c x = c 0 + c 1 x + c 2 x 2 + c 3 x 3 + is called a power series i the variable x. The umbers c are called the coefficiets of
More informationMultiterminal source coding with complementary delivery
Iteratioal Symposium o Iformatio Theory ad its Applicatios, ISITA2006 Seoul, Korea, October 29 November 1, 2006 Multitermial source codig with complemetary delivery Akisato Kimura ad Tomohiko Uyematsu
More informationEfficient GMM LECTURE 12 GMM II
DECEMBER 1 010 LECTURE 1 II Efficiet The estimator depeds o the choice of the weight matrix A. The efficiet estimator is the oe that has the smallest asymptotic variace amog all estimators defied by differet
More informationInfinite Sequences and Series
Chapter 6 Ifiite Sequeces ad Series 6.1 Ifiite Sequeces 6.1.1 Elemetary Cocepts Simply speakig, a sequece is a ordered list of umbers writte: {a 1, a 2, a 3,...a, a +1,...} where the elemets a i represet
More informationSequences A sequence of numbers is a function whose domain is the positive integers. We can see that the sequence
Sequeces A sequece of umbers is a fuctio whose domai is the positive itegers. We ca see that the sequece 1, 1, 2, 2, 3, 3,... is a fuctio from the positive itegers whe we write the first sequece elemet
More informationLecture 7: October 18, 2017
Iformatio ad Codig Theory Autum 207 Lecturer: Madhur Tulsiai Lecture 7: October 8, 207 Biary hypothesis testig I this lecture, we apply the tools developed i the past few lectures to uderstad the problem
More informationProbability 2 - Notes 10. Lemma. If X is a random variable and g(x) 0 for all x in the support of f X, then P(g(X) 1) E[g(X)].
Probability 2 - Notes 0 Some Useful Iequalities. Lemma. If X is a radom variable ad g(x 0 for all x i the support of f X, the P(g(X E[g(X]. Proof. (cotiuous case P(g(X Corollaries x:g(x f X (xdx x:g(x
More informationTopic 9: Sampling Distributions of Estimators
Topic 9: Samplig Distributios of Estimators Course 003, 2018 Page 0 Samplig distributios of estimators Sice our estimators are statistics (particular fuctios of radom variables), their distributio ca be
More informationChapter 6 Infinite Series
Chapter 6 Ifiite Series I the previous chapter we cosidered itegrals which were improper i the sese that the iterval of itegratio was ubouded. I this chapter we are goig to discuss a topic which is somewhat
More informationSolutions to Math 347 Practice Problems for the final
Solutios to Math 347 Practice Problems for the fial 1) True or False: a) There exist itegers x,y such that 50x + 76y = 6. True: the gcd of 50 ad 76 is, ad 6 is a multiple of. b) The ifiimum of a set is
More informationUniversal source coding for complementary delivery
SITA2006 i Hakodate 2005.2. p. Uiversal source codig for complemetary delivery Akisato Kimura, 2, Tomohiko Uyematsu 2, Shigeaki Kuzuoka 2 Media Iformatio Laboratory, NTT Commuicatio Sciece Laboratories,
More informationTopic 9: Sampling Distributions of Estimators
Topic 9: Samplig Distributios of Estimators Course 003, 2016 Page 0 Samplig distributios of estimators Sice our estimators are statistics (particular fuctios of radom variables), their distributio ca be
More informationTopic 9: Sampling Distributions of Estimators
Topic 9: Samplig Distributios of Estimators Course 003, 2018 Page 0 Samplig distributios of estimators Sice our estimators are statistics (particular fuctios of radom variables), their distributio ca be
More informationLinear Regression Demystified
Liear Regressio Demystified Liear regressio is a importat subject i statistics. I elemetary statistics courses, formulae related to liear regressio are ofte stated without derivatio. This ote iteds to
More informationTHE interference channel problem describes a setup where multiple pairs of transmitters and receivers share a communication
Trade-off betwee Commuicatio ad Cooperatio i the Iterferece Chael Farhad Shirai EECS Departmet Uiversity of Michiga A Arbor,USA Email: fshirai@umich.edu S. Sadeep Pradha EECS Departmet Uiversity of Michiga
More informationInformation-based Feature Selection
Iformatio-based Feature Selectio Farza Faria, Abbas Kazeroui, Afshi Babveyh Email: {faria,abbask,afshib}@staford.edu 1 Itroductio Feature selectio is a topic of great iterest i applicatios dealig with
More informationAxioms of Measure Theory
MATH 532 Axioms of Measure Theory Dr. Neal, WKU I. The Space Throughout the course, we shall let X deote a geeric o-empty set. I geeral, we shall ot assume that ay algebraic structure exists o X so that
More informationPRELIM PROBLEM SOLUTIONS
PRELIM PROBLEM SOLUTIONS THE GRAD STUDENTS + KEN Cotets. Complex Aalysis Practice Problems 2. 2. Real Aalysis Practice Problems 2. 4 3. Algebra Practice Problems 2. 8. Complex Aalysis Practice Problems
More informationOutput Analysis (2, Chapters 10 &11 Law)
B. Maddah ENMG 6 Simulatio Output Aalysis (, Chapters 10 &11 Law) Comparig alterative system cofiguratio Sice the output of a simulatio is radom, the comparig differet systems via simulatio should be doe
More informationMultiterminal Source Coding with an Entropy-Based Distortion Measure
20 IEEE Iteratioal Symposium o Iformatio Theory Proceedigs Multitermial Source Codig with a Etropy-Based Distortio Measure Thomas A. Courtade ad Richard D. Wesel Departmet of Electrical Egieerig Uiversity
More informationCov(aX, cy ) Var(X) Var(Y ) It is completely invariant to affine transformations: for any a, b, c, d R, ρ(ax + b, cy + d) = a.s. X i. as n.
CS 189 Itroductio to Machie Learig Sprig 218 Note 11 1 Caoical Correlatio Aalysis The Pearso Correlatio Coefficiet ρ(x, Y ) is a way to measure how liearly related (i other words, how well a liear model
More informationMath 61CM - Solutions to homework 3
Math 6CM - Solutios to homework 3 Cédric De Groote October 2 th, 208 Problem : Let F be a field, m 0 a fixed oegative iteger ad let V = {a 0 + a x + + a m x m a 0,, a m F} be the vector space cosistig
More informationSequences and Series of Functions
Chapter 6 Sequeces ad Series of Fuctios 6.1. Covergece of a Sequece of Fuctios Poitwise Covergece. Defiitio 6.1. Let, for each N, fuctio f : A R be defied. If, for each x A, the sequece (f (x)) coverges
More informationRademacher Complexity
EECS 598: Statistical Learig Theory, Witer 204 Topic 0 Rademacher Complexity Lecturer: Clayto Scott Scribe: Ya Deg, Kevi Moo Disclaimer: These otes have ot bee subjected to the usual scrutiy reserved for
More informationSOLUTIONS TO EXAM 3. Solution: Note that this defines two convergent geometric series with respective radii r 1 = 2/5 < 1 and r 2 = 1/5 < 1.
SOLUTIONS TO EXAM 3 Problem Fid the sum of the followig series 2 + ( ) 5 5 2 5 3 25 2 2 This series diverges Solutio: Note that this defies two coverget geometric series with respective radii r 2/5 < ad
More informationA sequence of numbers is a function whose domain is the positive integers. We can see that the sequence
Sequeces A sequece of umbers is a fuctio whose domai is the positive itegers. We ca see that the sequece,, 2, 2, 3, 3,... is a fuctio from the positive itegers whe we write the first sequece elemet as
More information1 of 7 7/16/2009 6:06 AM Virtual Laboratories > 6. Radom Samples > 1 2 3 4 5 6 7 6. Order Statistics Defiitios Suppose agai that we have a basic radom experimet, ad that X is a real-valued radom variable
More informationRandom Variables, Sampling and Estimation
Chapter 1 Radom Variables, Samplig ad Estimatio 1.1 Itroductio This chapter will cover the most importat basic statistical theory you eed i order to uderstad the ecoometric material that will be comig
More informationAn Introduction to Randomized Algorithms
A Itroductio to Radomized Algorithms The focus of this lecture is to study a radomized algorithm for quick sort, aalyze it usig probabilistic recurrece relatios, ad also provide more geeral tools for aalysis
More informationEECS564 Estimation, Filtering, and Detection Hwk 2 Solns. Winter p θ (z) = (2θz + 1 θ), 0 z 1
EECS564 Estimatio, Filterig, ad Detectio Hwk 2 Sols. Witer 25 4. Let Z be a sigle observatio havig desity fuctio where. p (z) = (2z + ), z (a) Assumig that is a oradom parameter, fid ad plot the maximum
More informationTHE KALMAN FILTER RAUL ROJAS
THE KALMAN FILTER RAUL ROJAS Abstract. This paper provides a getle itroductio to the Kalma filter, a umerical method that ca be used for sesor fusio or for calculatio of trajectories. First, we cosider
More informationFixed-Threshold Polar Codes
Fixed-Threshold Polar Codes Jig Guo Uiversity of Cambridge jg582@cam.ac.uk Albert Guillé i Fàbregas ICREA & Uiversitat Pompeu Fabra Uiversity of Cambridge guille@ieee.org Jossy Sayir Uiversity of Cambridge
More informationUC Berkeley CS 170: Efficient Algorithms and Intractable Problems Handout 17 Lecturer: David Wagner April 3, Notes 17 for CS 170
UC Berkeley CS 170: Efficiet Algorithms ad Itractable Problems Hadout 17 Lecturer: David Wager April 3, 2003 Notes 17 for CS 170 1 The Lempel-Ziv algorithm There is a sese i which the Huffma codig was
More informationINFINITE SEQUENCES AND SERIES
11 INFINITE SEQUENCES AND SERIES INFINITE SEQUENCES AND SERIES 11.4 The Compariso Tests I this sectio, we will lear: How to fid the value of a series by comparig it with a kow series. COMPARISON TESTS
More informationSession 5. (1) Principal component analysis and Karhunen-Loève transformation
200 Autum semester Patter Iformatio Processig Topic 2 Image compressio by orthogoal trasformatio Sessio 5 () Pricipal compoet aalysis ad Karhue-Loève trasformatio Topic 2 of this course explais the image
More informationProperties and Hypothesis Testing
Chapter 3 Properties ad Hypothesis Testig 3.1 Types of data The regressio techiques developed i previous chapters ca be applied to three differet kids of data. 1. Cross-sectioal data. 2. Time series data.
More informationLecture 22: Review for Exam 2. 1 Basic Model Assumptions (without Gaussian Noise)
Lecture 22: Review for Exam 2 Basic Model Assumptios (without Gaussia Noise) We model oe cotiuous respose variable Y, as a liear fuctio of p umerical predictors, plus oise: Y = β 0 + β X +... β p X p +
More information1 Review of Probability & Statistics
1 Review of Probability & Statistics a. I a group of 000 people, it has bee reported that there are: 61 smokers 670 over 5 960 people who imbibe (drik alcohol) 86 smokers who imbibe 90 imbibers over 5
More informationA statistical method to determine sample size to estimate characteristic value of soil parameters
A statistical method to determie sample size to estimate characteristic value of soil parameters Y. Hojo, B. Setiawa 2 ad M. Suzuki 3 Abstract Sample size is a importat factor to be cosidered i determiig
More informationOn the Capacity of Symmetric Gaussian Interference Channels with Feedback
O the Capacity of Symmetric Gaussia Iterferece Chaels with Feedback La V Truog Iformatio Techology Specializatio Departmet ITS FPT Uiversity, Haoi, Vietam E-mail: latv@fpteduv Hirosuke Yamamoto Dept of
More informationElement sampling: Part 2
Chapter 4 Elemet samplig: Part 2 4.1 Itroductio We ow cosider uequal probability samplig desigs which is very popular i practice. I the uequal probability samplig, we ca improve the efficiecy of the resultig
More informationA New Achievability Scheme for the Relay Channel
A New Achievability Scheme for the Relay Chael Wei Kag Seur Ulukus Departmet of Electrical ad Computer Egieerig Uiversity of Marylad, College Park, MD 20742 wkag@umd.edu ulukus@umd.edu October 4, 2007
More informationECE 901 Lecture 12: Complexity Regularization and the Squared Loss
ECE 90 Lecture : Complexity Regularizatio ad the Squared Loss R. Nowak 5/7/009 I the previous lectures we made use of the Cheroff/Hoeffdig bouds for our aalysis of classifier errors. Hoeffdig s iequality
More informationInformation Theory Tutorial Communication over Channels with memory. Chi Zhang Department of Electrical Engineering University of Notre Dame
Iformatio Theory Tutorial Commuicatio over Chaels with memory Chi Zhag Departmet of Electrical Egieerig Uiversity of Notre Dame Abstract A geeral capacity formula C = sup I(; Y ), which is correct for
More information6.3 Testing Series With Positive Terms
6.3. TESTING SERIES WITH POSITIVE TERMS 307 6.3 Testig Series With Positive Terms 6.3. Review of what is kow up to ow I theory, testig a series a i for covergece amouts to fidig the i= sequece of partial
More informationSensitivity Analysis of Daubechies 4 Wavelet Coefficients for Reduction of Reconstructed Image Error
Proceedigs of the 6th WSEAS Iteratioal Coferece o SIGNAL PROCESSING, Dallas, Texas, USA, March -4, 7 67 Sesitivity Aalysis of Daubechies 4 Wavelet Coefficiets for Reductio of Recostructed Image Error DEVINDER
More information1 Introduction to reducing variance in Monte Carlo simulations
Copyright c 010 by Karl Sigma 1 Itroductio to reducig variace i Mote Carlo simulatios 11 Review of cofidece itervals for estimatig a mea I statistics, we estimate a ukow mea µ = E(X) of a distributio by
More informationProblem Set 4 Due Oct, 12
EE226: Radom Processes i Systems Lecturer: Jea C. Walrad Problem Set 4 Due Oct, 12 Fall 06 GSI: Assae Gueye This problem set essetially reviews detectio theory ad hypothesis testig ad some basic otios
More informationChapter 3. Strong convergence. 3.1 Definition of almost sure convergence
Chapter 3 Strog covergece As poited out i the Chapter 2, there are multiple ways to defie the otio of covergece of a sequece of radom variables. That chapter defied covergece i probability, covergece i
More informationStatistical Properties of OLS estimators
1 Statistical Properties of OLS estimators Liear Model: Y i = β 0 + β 1 X i + u i OLS estimators: β 0 = Y β 1X β 1 = Best Liear Ubiased Estimator (BLUE) Liear Estimator: β 0 ad β 1 are liear fuctio of
More informationOn Routing-Optimal Network for Multiple Unicasts
O Routig-Optimal Network for Multiple Uicasts Chu Meg, Athia Markopoulou Abstract I this paper, we cosider etworks with multiple uicast sessios. Geerally, o-liear etwork codig is eeded to achieve the whole
More informationDefinitions and Theorems. where x are the decision variables. c, b, and a are constant coefficients.
Defiitios ad Theorems Remember the scalar form of the liear programmig problem, Miimize, Subject to, f(x) = c i x i a 1i x i = b 1 a mi x i = b m x i 0 i = 1,2,, where x are the decisio variables. c, b,
More informationMath 451: Euclidean and Non-Euclidean Geometry MWF 3pm, Gasson 204 Homework 3 Solutions
Math 451: Euclidea ad No-Euclidea Geometry MWF 3pm, Gasso 204 Homework 3 Solutios Exercises from 1.4 ad 1.5 of the otes: 4.3, 4.10, 4.12, 4.14, 4.15, 5.3, 5.4, 5.5 Exercise 4.3. Explai why Hp, q) = {x
More informationThe Three-Terminal Interactive. Lossy Source Coding Problem
The Three-Termial Iteractive 1 Lossy Source Codig Problem Leoardo Rey Vega, Pablo Piataida ad Alfred O. Hero III arxiv:1502.01359v3 cs.it] 18 Ja 2016 Abstract The three-ode multitermial lossy source codig
More informationBinary codes from graphs on triples and permutation decoding
Biary codes from graphs o triples ad permutatio decodig J. D. Key Departmet of Mathematical Scieces Clemso Uiversity Clemso SC 29634 U.S.A. J. Moori ad B. G. Rodrigues School of Mathematics Statistics
More informationBeurling Integers: Part 2
Beurlig Itegers: Part 2 Isomorphisms Devi Platt July 11, 2015 1 Prime Factorizatio Sequeces I the last article we itroduced the Beurlig geeralized itegers, which ca be represeted as a sequece of real umbers
More informationLecture 7: Channel coding theorem for discrete-time continuous memoryless channel
Lecture 7: Chael codig theorem for discrete-time cotiuous memoryless chael Lectured by Dr. Saif K. Mohammed Scribed by Mirsad Čirkić Iformatio Theory for Wireless Commuicatio ITWC Sprig 202 Let us first
More informationOn Rate Requirements for Achieving the Centralized Performance in Distributed Estimation
O Rate Requiremets for Achievig the Cetralized Performace i Distributed Estimatio Mostafa El Gamal ad Lifeg Lai Abstract We cosider a distributed parameter estimatio problem, i which multiple termials
More informationLossy Compression via Sparse Linear Regression: Computationally Efficient Encoding and Decoding
ossy Compressio via Sparse iear Regressio: Computatioally Efficiet Ecodig ad Decodig Ramji Vekataramaa Dept of Egieerig, Uiv of Cambridge ramjiv@egcamacuk Tuhi Sarkar Dept of EE, IIT Bombay tuhi91@gmailcom
More informationOn Evaluating the Rate-Distortion Function of Sources with Feed-Forward and the Capacity of Channels with Feedback.
O Evaluatig the Rate-Distortio Fuctio of Sources with Feed-Forward ad the Capacity of Chaels with Feedback. Ramji Vekataramaa ad S. Sadeep Pradha Departmet of EECS, Uiversity of Michiga, A Arbor, MI 4805
More informationApplication to Random Graphs
A Applicatio to Radom Graphs Brachig processes have a umber of iterestig ad importat applicatios. We shall cosider oe of the most famous of them, the Erdős-Réyi radom graph theory. 1 Defiitio A.1. Let
More informationMASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.436J/15.085J Fall 2008 Lecture 19 11/17/2008 LAWS OF LARGE NUMBERS II THE STRONG LAW OF LARGE NUMBERS
MASSACHUSTTS INSTITUT OF TCHNOLOGY 6.436J/5.085J Fall 2008 Lecture 9 /7/2008 LAWS OF LARG NUMBRS II Cotets. The strog law of large umbers 2. The Cheroff boud TH STRONG LAW OF LARG NUMBRS While the weak
More informationThe Method of Least Squares. To understand least squares fitting of data.
The Method of Least Squares KEY WORDS Curve fittig, least square GOAL To uderstad least squares fittig of data To uderstad the least squares solutio of icosistet systems of liear equatios 1 Motivatio Curve
More information( θ. sup θ Θ f X (x θ) = L. sup Pr (Λ (X) < c) = α. x : Λ (x) = sup θ H 0. sup θ Θ f X (x θ) = ) < c. NH : θ 1 = θ 2 against AH : θ 1 θ 2
82 CHAPTER 4. MAXIMUM IKEIHOOD ESTIMATION Defiitio: et X be a radom sample with joit p.m/d.f. f X x θ. The geeralised likelihood ratio test g.l.r.t. of the NH : θ H 0 agaist the alterative AH : θ H 1,
More informationLecture 23: Minimal sufficiency
Lecture 23: Miimal sufficiecy Maximal reductio without loss of iformatio There are may sufficiet statistics for a give problem. I fact, X (the whole data set) is sufficiet. If T is a sufficiet statistic
More informationOptimally Sparse SVMs
A. Proof of Lemma 3. We here prove a lower boud o the umber of support vectors to achieve geeralizatio bouds of the form which we cosider. Importatly, this result holds ot oly for liear classifiers, but
More informationResampling Methods. X (1/2), i.e., Pr (X i m) = 1/2. We order the data: X (1) X (2) X (n). Define the sample median: ( n.
Jauary 1, 2019 Resamplig Methods Motivatio We have so may estimators with the property θ θ d N 0, σ 2 We ca also write θ a N θ, σ 2 /, where a meas approximately distributed as Oce we have a cosistet estimator
More informationSummary and Discussion on Simultaneous Analysis of Lasso and Dantzig Selector
Summary ad Discussio o Simultaeous Aalysis of Lasso ad Datzig Selector STAT732, Sprig 28 Duzhe Wag May 4, 28 Abstract This is a discussio o the work i Bickel, Ritov ad Tsybakov (29). We begi with a short
More informationIP Reference guide for integer programming formulations.
IP Referece guide for iteger programmig formulatios. by James B. Orli for 15.053 ad 15.058 This documet is iteded as a compact (or relatively compact) guide to the formulatio of iteger programs. For more
More informationTEACHER CERTIFICATION STUDY GUIDE
COMPETENCY 1. ALGEBRA SKILL 1.1 1.1a. ALGEBRAIC STRUCTURES Kow why the real ad complex umbers are each a field, ad that particular rigs are ot fields (e.g., itegers, polyomial rigs, matrix rigs) Algebra
More informationStat 421-SP2012 Interval Estimation Section
Stat 41-SP01 Iterval Estimatio Sectio 11.1-11. We ow uderstad (Chapter 10) how to fid poit estimators of a ukow parameter. o However, a poit estimate does ot provide ay iformatio about the ucertaity (possible
More informationMASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 3 9/11/2013. Large deviations Theory. Cramér s Theorem
MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/5.070J Fall 203 Lecture 3 9//203 Large deviatios Theory. Cramér s Theorem Cotet.. Cramér s Theorem. 2. Rate fuctio ad properties. 3. Chage of measure techique.
More informationCHAPTER 5 SOME MINIMAX AND SADDLE POINT THEOREMS
CHAPTR 5 SOM MINIMA AND SADDL POINT THORMS 5. INTRODUCTION Fied poit theorems provide importat tools i game theory which are used to prove the equilibrium ad eistece theorems. For istace, the fied poit
More informationPrinciple Of Superposition
ecture 5: PREIMINRY CONCEP O RUCUR NYI Priciple Of uperpositio Mathematically, the priciple of superpositio is stated as ( a ) G( a ) G( ) G a a or for a liear structural system, the respose at a give
More informationLecture 14: Graph Entropy
15-859: Iformatio Theory ad Applicatios i TCS Sprig 2013 Lecture 14: Graph Etropy March 19, 2013 Lecturer: Mahdi Cheraghchi Scribe: Euiwoog Lee 1 Recap Bergma s boud o the permaet Shearer s Lemma Number
More informationThe variance of a sum of independent variables is the sum of their variances, since covariances are zero. Therefore. V (xi )= n n 2 σ2 = σ2.
SAMPLE STATISTICS A radom sample x 1,x,,x from a distributio f(x) is a set of idepedetly ad idetically variables with x i f(x) for all i Their joit pdf is f(x 1,x,,x )=f(x 1 )f(x ) f(x )= f(x i ) The sample
More informationLinear regression. Daniel Hsu (COMS 4771) (y i x T i β)2 2πσ. 2 2σ 2. 1 n. (x T i β y i ) 2. 1 ˆβ arg min. β R n d
Liear regressio Daiel Hsu (COMS 477) Maximum likelihood estimatio Oe of the simplest liear regressio models is the followig: (X, Y ),..., (X, Y ), (X, Y ) are iid radom pairs takig values i R d R, ad Y
More informationAdvanced Analysis. Min Yan Department of Mathematics Hong Kong University of Science and Technology
Advaced Aalysis Mi Ya Departmet of Mathematics Hog Kog Uiversity of Sciece ad Techology September 3, 009 Cotets Limit ad Cotiuity 7 Limit of Sequece 8 Defiitio 8 Property 3 3 Ifiity ad Ifiitesimal 8 4
More informationChapter 6 Principles of Data Reduction
Chapter 6 for BST 695: Special Topics i Statistical Theory. Kui Zhag, 0 Chapter 6 Priciples of Data Reductio Sectio 6. Itroductio Goal: To summarize or reduce the data X, X,, X to get iformatio about a
More informationEntropy and Ergodic Theory Lecture 5: Joint typicality and conditional AEP
Etropy ad Ergodic Theory Lecture 5: Joit typicality ad coditioal AEP 1 Notatio: from RVs back to distributios Let (Ω, F, P) be a probability space, ad let X ad Y be A- ad B-valued discrete RVs, respectively.
More informationThe Boolean Ring of Intervals
MATH 532 Lebesgue Measure Dr. Neal, WKU We ow shall apply the results obtaied about outer measure to the legth measure o the real lie. Throughout, our space X will be the set of real umbers R. Whe ecessary,
More informationBounds for the Extreme Eigenvalues Using the Trace and Determinant
ISSN 746-7659, Eglad, UK Joural of Iformatio ad Computig Sciece Vol 4, No, 9, pp 49-55 Bouds for the Etreme Eigevalues Usig the Trace ad Determiat Qi Zhog, +, Tig-Zhu Huag School of pplied Mathematics,
More informationIt should be unbiased, or approximately unbiased. Variance of the variance estimator should be small. That is, the variance estimator is stable.
Chapter 10 Variace Estimatio 10.1 Itroductio Variace estimatio is a importat practical problem i survey samplig. Variace estimates are used i two purposes. Oe is the aalytic purpose such as costructig
More information3. Z Transform. Recall that the Fourier transform (FT) of a DT signal xn [ ] is ( ) [ ] = In order for the FT to exist in the finite magnitude sense,
3. Z Trasform Referece: Etire Chapter 3 of text. Recall that the Fourier trasform (FT) of a DT sigal x [ ] is ω ( ) [ ] X e = j jω k = xe I order for the FT to exist i the fiite magitude sese, S = x [
More information