Optimal Operating Point for MIMO Multiple Access Channel with Bursty Traffic

Size: px
Start display at page:

Download "Optimal Operating Point for MIMO Multiple Access Channel with Bursty Traffic"

Transcription

1 Optiml Operting Point for MIMO Multiple Access Chnnel with Bursty Trffic Somsk Kittipiykul nd Tr Jvidi Abstrct Multiple ntenns t the trnsmitters nd receivers in multiple ccess chnnel (MAC) cn provide simultneous diversity, sptil multiplexing, nd spce-division multiple ccess gins. The fundmentl trdeoff in the symptoticlly lrge SNR regime is shown by Tse et l (24). On the other hnd, MAC scheduling cn provide sttisticl-multiplexing gin to improve the dely performnce s shown by Bertsims et l (998) nd Stolyr et l (2). In this pper, we formulte nd nlyticlly derive bounds on the optiml operting point for MIMO-MAC chnnel for bursty sources with dely constrints. Our system model brings together the four types of gins: diversity, sptil multiplexing, spce-division multiple-ccess, nd sttisticl-multiplexing gins. Our objective is to minimize the end-to-end performnce s defined by the dely bound violtion probbility s well s the chnnel decoding error probbility. We find the optiml diversity gin nd rte region in which the system should operte. As n exmple, we illustrte our technique nd the optiml operting point for the cse of compound Poisson source. In ddition, we note n interesting interply between the intensity of the trffic nd resource pooling with regrd to both multiple-ccess nd sttisticl-multiplexing gins. Index Terms MIMO multiple ccess chnnel, diversitymultiplexing trdeoff, cross-lyer optimiztion, sttisticlmultiplexing. I. INTRODUCTION Multiple ntenns cn be used to enhnce the performnce of wireless systems. The multiple ntenns cn be used to simultneously boost the relibility (providing diversity gin) nd the dt rte (providing sptil multiplexing gin). In ddition, in multiple ccess scenrios where multiple users re trnsmitting to common receiver, multiple receive ntenns lso provide multiple-ccess gin by llowing for sptil seprtion of the signls of different users. Tse, Viswnth, nd Zheng [2] hve chrcterized the fundmentl trdeoff between these three types of gins t high SNR. Our gol in this pper is to nswer the question first posed by Hollidy nd Goldsmith in [5]: given the diversitymultiplexing region, where should one choose to operte?. Hollidy nd Goldsmith nswered this question in the context of the cross-lyer design of point-to-point system where source encoder is conctented with MIMO link. Their gol in this context ws to minimize the end-to-end distortion. In their lter ppers [6] nd [7], Hollidy et l generlized their formultion to include the distortion due to dely, where the dely is cused by rndom service time of the ARQ process. In this pper, we nswer the sme question in multi-user context. We consider cross-lyer queue-chnnel optimiztion problem for bursty nd dely-sensitive trffic sources. We consider system where ech user hs bursty source conctented with n infinite buffer nd MIMO multiple ccess chnnel (MIMO-MAC). The end-to-end performnce metric of interest is the totl bit loss probbility, where loss cn be due to either dely violtion or decoding errors in the MIMO-MAC chnnel. From user s perspective, we fce the following trdeoff: the higher the multiplexing gin the better the dely performnce, but the inevitble decrese in diversity results in n increse in MIMO chnnel errors. At the sme time, the sttisticl vrition in the trffic ptterns mong users provides us with flexibility in llocting the resources. The min contribution of this pper is the formultion of cross-lyer optiml operting point for MIMO-MAC chnnel with bursty sources nd dely constrints. In prticulr, we provide methodology for chrcterizing the optiml diversity gin nd rte region in which the system should operte in MIMO-MAC chnnel with given high SNR nd description of the bursty trffic sources. To chieve this, we ssume n optiml scheduler design which dynmiclly controls users trnsmission rtes (or equivlently, the multiplexing gins) s function of queue bcklogs. This dynmic dpttion of multiplexing gins ccounts for sttisticl-multiplexing while leverging the known trdeoff between diversity, sptil multiplexing, nd multiple-ccess gins given in [2]. From scheduling perspective, sttisticl-multiplexing is key mechnism by which the network resources re used to improve the dely performnce for bursty users. In prticulr, sttisticlmultiplexing cpitlizes on the fct tht peks in trffic of simultneously ongoing trffic strems rrely coincide. We believe tht our result cn be viewed s first step in integrting the known sptil diversity nd multiplexing nd multiple-ccess gins with tht of the sttisticl-multiplexing. In other words, for the first time, our model brings together the four types of gins offered t MIMO MAC. The reminder of the pper is orgnized s follows. In Section II, we provide detiled description of the system model s well s the problem formultion for generl number of users. In Section III, we provide the min nlyticl results nd bounds on the optiml chnnel diversity gin. We lso discuss the notion of sttisticl-multiplexing nd its benefits. In Section VI, we find the optiml operting diversity gin d or its bounds for prticulr clss of compound Poisson sources. Finlly, in Section V, we discuss the shortcomings of the present pper, possible extensions, nd future work. II. SYSTEM MODEL We consider the rchitecture shown in Figure. The system is time-slotted nd consists of three min components, ech shown with different number. The first component consists

2 2 Source Buffer r W t t Rte Scheduler Spce-Time Encoder rte r t 3 Tx MIMO-MAC (diversity d) Rx Joint-ML Decoder 2 Rx App. Bit loss (decoding error & dely violtion) Source K W K t r K t Spce-Time Encoder rte r K t Tx Rx App. K Fig.. System model nd the two cuses of bit loss: dely violtion nd chnnel decoding error. of K homogeneous users, ech of which hs n identicl but independent bursty source: ech source genertes informtion bits ccording to stochstic process. When pproprite, the bits re buffered prior to trnsmission over the chnnel. The second component of interest is MIMO multiple ccess chnnel without chnnel stte informtion (CSI) t the trnsmitters but with perfect CSI t the receiver. The receiver consists of joint mximum-likelihood decoder. In the bsence of CSI t the trnsmitters, we ssume tht the MIMO-MAC opertes t common diversity gin, which in turn specifies the corresponding cpcity region of the MIMO MAC chnnel s given in [2]. However, the individul rte of ech user is determined dynmiclly by the rte scheduler which is the third component in our system. This is centrlized rte scheduler tht dynmiclly determines the trnsmission rtes of the individul users given queue stte informtion (QSI) of ech user. In this pper we re interested in the following questions: given sttisticl description of the sources, lrge dely bound D, nd high SNR vlue of the chnnel, wht is the optiml design of the scheduler, nd wht is the optiml operting diversity gin of the MIMO MAC chnnel. The notion of optimlity needs to tke into ccount the chnnel decoding error s well s the dely bound violtion probbilities. We ssume no restrnsmission of the bits in error nd mp our objective to the sum of the probbility of dely violtion nd the probbility of chnnel error. In order to mthemticlly define this problem, we now model ech of the bove components precisely. A. Source Model We ssume tht the totl number of informtion bits generted by user i (i =,..., K) is given by sequence S i = St, i t =, 2,..., where St i is the totl number of bits of user i generted up to timeslot t nd S i. In ddition, we ssume tht the rrivl processes S i, i =,..., K, re identicl nd mutully independent. We lso ssume tht ech rrivl process S i hs sttionry increments nd stisfies Lrge Devitions Principle (LDP). In the ppendix, we discuss n dditionl smple pth LDP ssumption (Assumption B) on the rrivl processes. Here, to keep the flow of the pper, in this section we only provide the LDP ssumption nd the consequent chrcteriztion of the sources which is bsed on LDP. In generl, consider source process S generting sequence S t, t =, 2,... of rndom vribles, where S t is the totl number of bits generted up to timeslot t. Definition : A source S is sid to stisfy n LDP with decy function Λ : R [, ] if, for lrge enough t nd for smll ǫ >, Pr [ St t ] ( ǫ, + ǫ) e tλ () where Λ is lower semicontinuous function nd hs compct level sets 2 (see [3] nd [9] for more discussions in LDP). Fct : (Gärtner-Ellis theorem) Suppose source S stisfies the following: Assumption A: ) The limiting log-moment generting function Λ(θ) := lim t t log ] (2) E[eθSt exists for ll θ, where ± re llowed both s elements of the sequence nd s limit points. 2) The origin is in the interior of the domin D Λ := θ Λ(θ) < of Λ(θ). 3) Λ(θ) is differentible in the interior of D Λ nd the derivtive tends to infinity s θ pproches the boundry of D Λ. 4) Λ(θ) is lower semicontinuous, i.e. liminf θn θ Λ(θ n ) Λ(θ) for ll θ. Then the source S stisfies n LDP nd its decy function described by () is given s () Λ () = sup [θ Λ(θ)]. (3) θ Remrk : It cn be shown tht Λ is convex function tking vlues in [, ] such tht Λ (E[S ]) = where E[S ] is the verge rrivl rte of process S [9]. Remrk 2: Mny source models commonly used to model bursty trffic in communiction networks stisfy Assumptions A nd B. Such source models include renewl processes, Mrkov-modulted processes, nd more generlly sttionry processes with mild mixing conditions [3]. In lrge devitions literture, the Λ function is typiclly clled rte function. Here we use the nme decy function to void confusion with trnsmission rte. 2 The level set x : Λ (x) is compct for every rel

3 3 Remrk 3: From LDP, source is fully chrcterized by either Λ( ) or Λ ( ). An lterntive to the LDP chrcteriztion of source is the well-known effective bndwidth (see [8]) which ws used in our previous study for point-to-point scenrio [4]. r Multiplexing gin region R d= d=.5 d= d=2 d=3.8 B. MIMO-MAC Chnnel Model nd PHY Model We use the sme symmetric MIMO multiple ccess chnnel model s described in [2] which ssumes symmetric trnsmitters seeing i.i.d. fding chnnels, perfect symbol synchroniztion nd perfect CSI t the receiver but no CSI t ny trnsmitters. Ech trnsmitter hs M trnsmit ntenns, while the receiver hs N receive ntenns. Spce-time coding hppens over chnnel coherence time which is ssumed to contin T symbols 3. Since the trnsmitters re ssumed to know only the chnnel sttistics, including the verge received SNR, they lwys trnsmit t the mximum powers which re equl for ll trnsmitters. The chnnel fding processes of the trnsmitters re ssumed to be sttionry over time, mutully independent, nd identicl. For ech trnsmitter, the chnnel fding for different ntenn pths re ssumed to be slow block-fding with i.i.d. Ryleigh fding. We denote by ρ the verge received signl-to-noise rtio (SNR) t ech receive ntenn. From system perspective, t ech SNR level ρ, the PHY lyer for the MIMO-MAC chnnel provides trdeoff between the relibility of the trnsmissions nd the trnsmission rtes. Equivlently, we cn sy tht the PHY lyer provides trdeoff between common diversity d nd the multiplexing gin region, denoted by R, where d nd R re s defined in [] nd [2]. We stte these definitions below. Definition 2: (Definition in []) A code scheme C(ρ), which is fmily of codes (coding over one single coherence block) with one codebook for ech SNR level ρ nd provides dt rte R(ρ) nd verge error probbility P e (ρ), chieves multiplexing gin r nd diversity gin d if R(ρ) lim ρ logρ = r nd lim log P e (ρ) = d. (4) ρ log ρ For nottionl simplicity we shorten (4) s R(ρ). = r log ρ nd P e (ρ). = ρ d. We lso use. nd. if nd hold in the limit. 4 Definition 3: (Theorem 2 in [2]) Let r i t be the multiplexing gin of user i, i =,...,K, t time t. Given common diversity requirement d for ll users, i.e., P i e. SNR d, i =,...,K, (5) where Pe i is the verge error probbility for user i. Then the sptil multiplexing gins (rt,...,rk t ) t ny timeslot t must 3 We ssume either sufficiently lrge symbol rte or sufficiently smll number of ntenns such tht T KM + N. 4 Note tht we use the nturl log insted of log 2 nd hence use nts insted of bits. This is for convenience of the presenttion r Fig. 2. Exmple of the multiplexing gin region R for M = N = 2 cse. be within the (time-independent) multiplexing gin region R = (r,..., r K ) : s S r s r S M,N, S,...,K where rm,n for ny integers m nd n is the lrgest multiplexing gin chieved for n m n point-to-point MIMO link for given diversity d nd is defined s piecewise liner function joining the points ((m k)(n k), k) for k =,...,min(m, n). In this pper, we consider system which lwys opertes t common diversity gin d t ny time t. This d directly determines the multiplexing region R nd its shpe. In prticulr, d determines the sum of ll the rtes t ll time, which is independent of time. However, the individul rte t time t, rt, i i =,...,K, is determined dynmiclly by the rte scheduler discussed lter. In Figure 2, we illustrte the dependence of the shpe of R nd d for simple cse of K = 2 users nd M = N = 2. As seen in this figure, there exists diversity gin d (in this exmple, d = 2) such tht, for lrge d (d > d ), the shpe of R follows rectngulr shpe (single-user performnce regime), while, for smll d (d < d ), R is polymtroid shpe (ntenn-pooling regime). Furthermore, [2] shows tht d is the unique solution to t, i.e. K i= ri t r KM,N. (6) r KM,N (d ) = Kr M,N (d ). (7) Lter we will see the impct of this chnge of shpe on the working of the scheduler block. C. Rte Scheduler Given tht ech user opertes t fixed nd common diversity gin d nd given n verge SNR of ρ in the MIMO- MAC subsystem, the function of scheduler g d : R K + R is to llocte, t the beginning of every timeslot t, the set of fesible multiplexing gins to the users. This is done, equivlently, by selecting vector of sptil multiplexing gins (rt,..., rt K ) from the multiplexing gin region R. The decision is bsed on the dely of the hed-of-the-line bit in queue i, denoted by Wt i, i =,...,K, t the beginning of

4 4 timeslot t. Specificlly, we ssume tht (r t,..., r K t ) = g d (W t,..., W K t ). Without loss of optimlity, one cn ssume tht the rte scheduler lwys ssigns the highest possible sum rte. At timeslot t, n mount of rt i T log ρ bits re tken out from hed-of-the-line of the buffer of user i. We ssume tht if ny buffers do not hve enough dt to trnsmit, the null dt is used to fulfill the rtes. Remrk 4: Recll tht the shpe of the multiplexing gin R depends on d (e.g. see Figure 2). As result, the choice of diversity gin d determines the clss of fesible dynmic schedulers. In the single-user performnce regime (d d ), the users re decoupled nd independent from one nother, hence, reducing the scheduler to sttic (nd decoupled) choice of multiplexing gin rt i = r M,N for ll i =,...,K nd ll time t Z. For the ntenn-pooling regime (d < d ), on the other hnd, R is polymtroid. In other words, in this regime, the multiplexing gins of the users re dependent on one nother nd must be jointly llocted. Remrk 5: The model in this pper ssumes tht there is no CSI vilble t the trnsmitters nd the centrl scheduler. However, the scheduler hs perfect knowledge of the queue stte informtion (QSI). This is not unrelistic given the fct tht it is less bndwidth consuming nd more ccurte to send the QSI of ech buffer (n observble sclr number) to the centrlized scheduler thn to estimte CSI for MIMO chnnels (K mtrices, ech of dimension M N) t the receiver nd feed bck these mtrices to the trnsmitters. Remrk 6: Due to lck of CSI, the role of the rte scheduler in this pper is not to minimize the chnnel error performnce; insted, the scheduler improves the dely violtion probbility by tking dvntge of the sttisticl-multiplexing gin provided by the multiple bursty sources shring the multiple ccess chnnel. D. Arrivl Rte Scling nd Stbility Condition Since the rtes of trnsmission in the MIMO-MAC chnnel re scled s log ρ, we scle the rrivl rtes with log ρ s well. In other words, we ssume tht the verge bit rrivl rte λ of ech user is λ = λt log ρ (8) bits per timeslot for given constnt positive λ. In ddition, to gurntee system stbility, we require tht the totl verge rrivl rte to be no greter thn the (sum) cpcity of the MIMO-MAC chnnel [4]. In prticulr, we ssume tht or equivlently K λ < min(km, N)T log ρ, λ < min(m, N/K). (9) Since the system is stble, it reches stedy stte. We let W i nd L i denote the stedy-stte dely nd queue length, respectively, for queue i, i =,...,K. For the rest of the pper, we denote r v s the verge multiplexing gin t the common diversity d, defined s rm,n r v := if d d, K r KM,N if d < d (). nd denote C v := r v T log ρ (bits per timeslot) () s the verge per-queue chnnel cpcity t diversity d. E. Objective Our system objective is to find the optiml operting chnnel diversity gin d nd the corresponding multiplexing gin region R(d ) in which the system should operte. This diversity d minimizes the end-to-end totl bit loss probbility cused by two phenomens: ) dely violtion of the dely bound D nd 2) chnnel decoding error. 5 In prticulr, we define the following probbilities: Pe i := Pr[decoding error for user i] P e := Pr[decoding error for ny user] (2) Pq i := Pr[dely violtion of user i] = Pr[W i > D] P q := Pr[dely violtion for ny user] = Pr[ mx i=,...,k W i > D]. (3) With the bove definitions, the totl loss probbility P tot is expressed s P tot := mx Pr[bit loss for user i] i=,...,k = mx P i e + ( Pe i )P q i i=,...,k where Pe i+( P e i)p q i is the totl bit loss probbility of user i due to chnnel nd dely violtion. We will lter show tht Pe i.. = P e = ρ d nd Pq i.. = P q = ρ g where g is functionl tking positive vlues. Hence, the symptotic lrge-snr expression of P tot is given s P tot. = P e + P q. = ρ d + ρ g. (4) We note tht both probbilities P e nd P q re functions of the diversity gin d s well s the verge SNR ρ. However, there is trdeoff between the two probbilities s function of d: Intuitively, for fixed ρ, we expect tht high diversity gin, which trnsltes into smller trnsmission rte region, results in fster queue build-up nd lrger delys. On the other hnd, this higher diversity gin yields better chnnel performnce. In the reminder of the pper, we will derive nlyticlly lrge SNR pproximtions for P q nd P e nd show tht given fixed nd high ρ, P q is incresing on d while P e is decresing on d (confirming the bove intuition). Furthermore, we find the best PHY lyer operting point, 5 We ssume no retrnsmission for the lost bits due to chnnel decoding errors or dely violtion. Furthermore, the source processes re not effected by the lost bits.

5 5 i.e. diversity gin d, so s to minimize the totl bit loss probbility P tot in the high SNR regime. In other words, we will find d tht blnces the exponents of the two probbilities. III. PROBLEM ANALYSIS In this section, we nlyticlly derive the two loss probbilities, P e nd P q, for symptoticlly lrge SNR. As we will see, the two probbilities decy exponentilly with SNR. For the chnnel, the definition of diversity gin [2] gives direct symptotic pproximtion of P e for lrge SNR. Obtining the symptotic P q, however, requires more work. Depending on the vlue of d, we either directly compute the symptotic P q or provide lower nd upper bounds of the symptotic P q. A. Asymptotic P e The symptotic expression of P e for lrge SNR comes directly from the definition of diversity gin given in Definitions 2 nd 3. By the union bound nd the symmetry mong users, we hve the following bounds: P e P e KP e where Pe is the probbility of decoding error for user. Using Pe =. ρ d in Definition 2 nd the fct tht K is constnt independent of ρ, we hve B. Asymptotic P q P e. = ρ d. (5) Similrly, by the union bound nd the symmetry mong users, we hve the following bounds: Pr[W > D] P q K Pr[W > D], (6) where Pr[W > D] implicitly depends on d. Now, let us first focus on Pr[W > D]. To get n nlyticl expression for the symptotic Pr[W > D], we consider two cses depending on the vlue of d. Cse : Single-user performnce regime (d d < MN) As discussed in Section II-C, the multiplexing gin region R in this regime is squre nd the scheduler ssign decoupled rtes to the queues. Hence, the optiml scheduler simply ssigns fixed trnsmission rte of C v given in () to ech user (i.e. we cll this the symmetric sttic scheduler). Therefore, the symptotic pproximtion (when D is sufficiently lrge) of the dely violtion probbility is given s follows. Lemm : For d d MN nd sufficiently lrge D, the symptotic lrge-snr pproximtion of Pr[W > D] is such tht log Pr [ W > D ] lim = σ s DTr v (7) ρ log ρ where σ s is defined such tht Λ(σ s ) = σ s C v. (8) Equivlently, we cn write (7) s Pr[W > D]. = ρ σsdtrv. (9) The proof of this Lemm is given in Appendix II. Cse 2: Antenn-pooling regime ( < d < d ) In this cse, the multiplexing gin region R is polymtroid nd the trnsmission rtes of the users must be jointly llocted by scheduler. Since the optiml policy (with respect to the dely violtion probbility objective) is unknown, we provide the following lower bound nd upper bound to Pr[W > D]. ) Upper Bound on Pr[W > D]: An upper bound on Pr[W > D] is esily found since ny fesible scheduling policy cn provide n upper bound. In prticulr, to rrive t the upper bound P u, we consider the sme symmetric sttic scheduler s described in Cse : the symmetric sttic scheduler lwys ssigns the symmetric rte of C v to ech user t ll time. By Lemm, the symptotic pproximtion of P u for lrge D is given s P u. = ρ σsdtrv. (2) We note tht this upper bound becomes tighter s d increses to d since R pproches K-dimensionl hypercube. 2) Lower Bound on Pr[W > D]: The lower bound P l on Pr[W > D] is obtined from Fct 2, the construction of fictitious system, nd Fct 3, s follow. Fct 2: Consider two systems whose multiplexing gin regions re given by R nd R 2, respectively, where R R 2. The dely violtion probbility ssocited with the second system is no greter thn tht of the first system. For given d, consider fictitious system whose multiplexing gin region is given by K R fic := (r,...,r K ) : r i rkm,n. (2) i= Since R R fic, Fct 2 sttes tht the dely violtion probbility for this system is lower bound for Pr[W > D]. Stolyr nd Rmnn [4] hve shown tht the lrgest-delyfirst (LDF) policy chieves the minimum symptotic dely violtion probbility for this fictitious system. Fct 3: (Theorem 2.2 in [4]) Consider single-server queuing model with K users (illustrted in Figure 4(iii) for K = 2). For the sources considered in this pper, the lrgest-delyfirst (LDF) policy chieves the minimum dely violtion probbility Pr[mx i=,...,k W i > D] when D is lrge. 6 Hence, we compute the minimum symptotic dely violtion probbility for this fictitious system to rrive t lower bound, P l, for Pr[W > D], s given in the following Lemm: Lemm 2: For < d < d nd sufficiently lrge D, the symptotic lrge-snr pproximtion of P l is given by P l. = ρ KσsDTrv. (22) 6 See more detils of this fct in Fct 4. The result in [4] is much more generl thn this. It works with ny weighted delys, i.e. Pr[mx i=,...,k W i /α i > D], where α i is the weight for user i.

6 6 R r 3 r* 3M,N nd, in prticulr, when d d < MN, P q. = ρ σsdtrv (24) R fic nd, when < d d, ρ KσsDTrv. P q. ρ σsdtrv. (25) r* 3M,N r < d < d 2 r* 3M,N Fig. 3. The MIMO-MAC multiplexing gin region R nd the multiplexing gin region of the fictitious system R fic for K = 3 users. r 2 In summry, so fr, we hve seen tht P q nd P e exponentilly decy with ρ. The rte of decy of P e is known. When d > d, the rte of decy of P q is known vi (24). The rte of decy of P q when d < d is, however, unknown but is bounded s in (25). Next, with P q nd P e t hnd, we proceed with the minimiztion of the totl bit loss probbility. + b/2 + b/2 b 2+b (i) (ii) (iii) Fig. 4. Queuing models of the upper nd lower bounds of P[W > D] for the cse of K = 2 users nd the ntenn-pooling regime (r2m,n 2rM,N ). nd b re defined such tht + b = r M,N nd 2 + b = r2m,n. The proof is given in Appendix II. We lso note tht P l becomes tighter bound s d. Remrk 7: Compring the exponents in (2) nd (22), we see tht the LDF scheduler improves the exponent of the dely violtion probbility by K times of tht of the symmetric sttic scheduler. Tlking in the lnguge of chnnel diversity, the LDF scheduler improves the diversity gin by K folds by tking dvntge of sttisticl-multiplexing of the sources. However, we wnt to emphsize tht the lower bound in P l derived from the fictitious system with the multiplexing gin region R fic becomes more loose s the number of users K grows. This is expected becuse the ctul multiplexing gin region R in (6) is polymtroid (see n exmple of K = 3 users in Figure 3) while tht of the fictitious system is just the K-dimensionl simplex given by the constrint K i= r i rkm,n. Thus, the lower-bound becomes more optimistic s the number of users increses. Remrk 8: For n exmple of K = 2 users, Figure 4 summrizes the two bounds with the queuing models in mind. The upper bound P l is the til probbility of system (i) which lwys serves ech queue with multiplexing gin r2m,n /2. The lower bound P l is the til probbility of system (iii) which ssigns the single server of multiplexing gin r2m,n bsed on LDF scheduling. System (ii) is the queuing model given by the multiplexing gin region R. Now, using the bove two cses nd the bounds in (6), we rrive t n symptotic chrcteriztion of P q s follows P q. = Pr[W > D], (23) C. Minimizing Asymptotic Totl Loss Probbility From the symptotic expressions of P e given in (5) nd P q in (24) nd (25), the symptotic chrcteriztion of the totl loss probbility P tot. = P q + P e is immedite: For d d MN, For < d < d, P tot. = ρ σsdtrv + ρ d. (26) ρ KσsDTrv + ρ d. P tot. ρ σsdtrv + ρ d. (27) Since the term σ s DTr v is decresing in d while the term d is incresing on d, the minimum of P tot in (26) or its bounds in (27) hppen when the vlue of d mkes the exponents of the two terms re within o() of ech other (note tht if the exponents were not in the sme order, one term would dominte in the sum s ρ ). We now introduce n lgorithm which gurntees such choices of d: Algorithm : ) Solve for d which is solution of σ s DTr v = d. (28) If d d, then d = d u = d l = d nd stop. Otherwise, set d l = d. Go to Step 2. 2) Solve for d which is solution of nd set d u = min(d, d ). Kσ s DTr v = d (29) Theorem : Algorithm results in closed intervl [d l, d u ] in which the optiml common diversity gin d lies. The proof of this theorem is given in Appendix II. Notice tht the optiml diversity d nd its bounds depend on the sttisticl chrcteristics of the symmetric sources (Λ, µ, λ), the prmeters of the MIMO-MAC chnnel (e.g. T, M, N), nd the dely bound D.

7 7 D. Sttisticl-Multiplexing nd Optiml Diversity Gin From the bove nlysis, we obtin the following criticl observtion. Given dely constrint, the sttisticl property of the source hs significnt impct on the level of diversity well-designed system cn enjoy. In other words, the optiml scheduler which sttisticlly multiplexes the MIMO resources llows the combined bursty sources to perceive s smller ggregte trffic nd hence higher degree of diversity. Rigorously, this performnce improvement cn be ttributed to sttisticl-multiplexing gin s follows: Definition 4: An optiml dynmic scheduler with the totl bit loss probbility Ptot provides sttisticl-multiplexing gin of s over the sttic rte scheduler with Ptot, f where s := lim ρ log Ptot log P f tot log ρ From this definition nd the fct tht P f tot. (3). = ρ d l, the following proposition is immedite. Lemm 3: Consider the system model in Section II. The optiml sttisticl-multiplexing gin s is given by s = d d l. (3) Furthermore, it is bounded bove by d u d l. Remrk 9: The two concepts of sttisticl-multiplexing gin nd multi-user diversity gin re relted conceptully. The former tkes dvntge of the troughs (due to burstiness) of the trffic of different users while the ltter tkes dvntge of the peks (due to fdings) of the chnnels of different users. But their impcts on the design re sufficiently different, s multi-user diversity gin requires chnnel CSI t the trnsmitters while sttisticl-multiplexing requires QSI. E. Resource Pooling nd Sttisticl-Multiplexing Here we discuss the effect of the rrivl rte λ nd the verge dely bound D to the performnce region of the MIMO-MAC. The reltionship between (λ, D) nd the system performnce is summrized in Figure 5. The system performnce is divided into three min regions: the singleuser performnce region, the ntenn-pooling with significnt sttisticl-multiplexing region, nd the ntenn-pooling with insignificnt sttisticl-multiplexing region. In the single-user performnce region, the chieved optiml diversity gin d is equivlent to the cse when only one user is in the system, i.e. the cse d d. Specificlly, this cse hppens when λ is sufficiently smll nd D is sufficiently lrge. We denote this region s A. d A := (λ, D) : λ r, D σ s (d )Tr where r = r M,N (d ). 7 Since in this region the trnsmission rte of ech user is independent, there is no resource shring nd hence no sttisticl-multiplexing gin. On the other hnd, the significnce of sttisticlmultiplexing gin outside A is impcted by the rte of rrivls λ s well s the verge dely bound D. In prticulr, for 7 d Note tht is n incresing function on λ since σ σ s(d )Tr s(d ) which is the dely violtion exponent is itself decresing on the rrivl rte λ. Fig. 5. D Single User Performnce A Antenn Pooling, significnt stt. mux. Antenn Pooling, insignificnt stt. mux. min(m,n/k) λ The reltion between (λ, D) to the system performnce. (λ, D) in the neighborhood of A, the sttisticl-multiplexing gin is not significnt s the queues still behve in roughly independent mnner. Similrly, s λ increses to n overlod sitution or the dely bound D becomes very tight, the benefits of juggling resources diminishes. In contrst, for medium vlues of λ nd D nd under the optiml dynmic scheduler, ech queue perceives the whole (pooled) resource to itself, compred to /K of the resource s in cse of symmetric sttic scheduler. To illustrte the pproch shown in this pper nd the corresponding clcultion, we look t simple exmple of compound Poisson source with K = 2 users in the next section. IV. EXAMPLE: COMPOUND POISSON SOURCES AND K = 2 In this section, we illustrte the proposed pproch vi n exmple. We consider two independent but identicl source processes. For ech source i, rrivls re independent cross timeslots. The number of bits tht rrive in timeslot t, A i t, is n ggregtion of rndom number of pckets whose sizes re lso rndom, i.e. A i t = N n= Y n. Furthermore, we ssume tht the number of pckets t ech slot, N, is n independent Poisson rndom vrible with rte ν pckets per timeslot, while the length of the pckets, Y i, i =, 2,..., re i.i.d. rndom vribles with exponentil distribution of men /µ. The verge bit rrivl rte λ for ech source is equl to ν/µ nd scles with log ρ s in (8), i.e. λ = ν/µ = λt log ρ. (32) Proposition : For the compound Poisson source with exponentil pcket length, the σ s defined in Lemms is given s σ s = µ( λ ). (33) r v This proposition is proved in Appendix II. Note tht the rtio of the per-queue verge bit rrivl rte λ over the verge service rte, r v, cn be clled the trffic lod per queue. It is importnt to note tht the dely violtion exponent in (33) is decresing function of the verge pcket size /µ, for fixed pcket rrivl rte ν. A lrger pcket size in effectively cretes more burstiness in the rrivls, hence higher dely violtion probbility.

8 8 Exponents of P q nd P e d * l Upperbound on exponent of P q Diversity Gin d d * u Liner estimte of exponent of P q Lowerbound on exponent of P q Exponents of P q nd P e d * l d * u Diversity Gin d d * d* l.5.5 Exponents of P q nd P e () λ =.5, D = Diversity Gin d d * l = d * u = d* (c) λ =.5, D = 5 Exponents of P q nd P e (b) λ =, D = Diversity Gin d λ =, D = 5 Fig. 6. Plots of the exponents of P q nd P e for two different verge rrivl rtes (λ =.5 nd ) nd for two different dely bounds D = 2 nd 5. For d < d = 7.3, the upper nd lower bounds of the exponent of P q re shown. In ddition, we drw simple liner estimte (dotted line) of the exponent of P q between the two bounds. With Proposition in hnd, we re now redy to use Algorithm to obtin d (or its bounds). Figure 6 shows the optiml d nd its bounds when M = N = 4, the verge pcket size /µ is nts, nd the symbol rte such tht there re T = 2M + N = symbols per timeslot. In these figures, we plot the exponents of P q or its bounds, i.e. σ s DTr v nd 2σ s DTr v, nd the exponent of P e, i.e. d. To better illustrte the procedure followed by Algorithm, we plot the exponents of P q nd P e seprtely. Note tht when d d (d = 7.3 in this exmple), we only hve lower nd upper bounds for the exponents of P q. In this cse, in ddition to the bounds, we plot liner pproximtion (dotted line) to emphsize the tightness of the lower bound round d nd the upper bound round. The optiml diversity gin d or its bounds (d l nd d u) re shown in ech plot s the crossing of the exponents. As we discussed in Section III-E, the sttisticl-multiplexing gin s defined in (3) depends on the optiml choice of d which itself is function of the rrivl rte λ nd the verge dely bound D. Depending on λ nd D, we my or my not hve sttisticl-multiplexing gin. For exmple, Figure 6(c) shows tht, in the cse of sufficiently low rrivl rtes nd lrge dely bound, the optiml Ptot hppens when the users operte in the single-user performnce region. Hence, in this cse, there is no sttisticl-multiplexing gin to be chieved by dynmic scheduler. Here, the dominnt form of loss occurs on the chnnels. On the other hnd, Figure 6(b) corresponds to the cse of lrge rrivl rte nd smll dely bound, where the loss probbility due to dely violtion domintes tht of the chnnel. In this cse, the optiml diversity d necessittes resource shring in form of ntenn pooling. As result, the impct of n optiml dynmic scheduler, in form of sttisticlmultiplexing gin, becomes more significnt. Figure 7 illustrtes the performnce region discussed in Section III-E. In prticulr, the figure gives chrcteriztion d * l d * u 4 D D () d d l.5 λ vs. (λ, D).5.5 λ (b) Contour Plot d d l Fig. 7. 3D nd contour plots chrcterizing the sttisticl-multiplexing gin, pproximted by d d l, v.s. dely bound D nd rrivl rte λ. of the region shown in Figure 5 for the compound Poisson cse. We pproximte the sttisticl-multiplexing gin s = d d l chieved with n optiml dynmic scheduler with tht of simple liner pproximtion d d l, where d is the optiml diversity gin derived from the dotted line in Figure 6..5 V. SUMMARY AND FUTURE WORK In this pper, we considered system of bursty nd delysensitive symmetric sources conctented with symmetric MIMO-MAC chnnel. We ssumed no CSI informtion vilble to the trnsmitters nd block fding model with block coding whose block lengths re mtched to the coherence time of the chnnel. Furthermore, we ssumed fixed nd equl high trnsmission power t ech trnsmitter, i.e. high SNR regime. We ddressed the optiml choice of the sptil diversity gin d such tht it minimizes n endto-end loss performnce where loss cn occur due to dely violtion s well s chnnel decoding error. We showed how n optiml choice of diversity gin d depends on queuebsed scheduler module whose job is to sttisticlly multiplex the resources of the MIMO-MAC. In doing so, we integrted the notion of sttisticl-multiplexing gin with those of sptil diversity, multiplexing, nd multi-ccess gins provided by the MIMO-MAC. To strengthen the result presented in this pper, we will need to nlyze the performnce of the optiml dynmic scheduler, rther thn working with bounds. As emphsized in the introduction to this pper, we view the result of

9 9 this pper s first step to fully integrte the notions of scheduling nd sttisticl-multiplexing with other spects of MIMO technology. As such, strengthening the current result is not our most criticl concern. Some future works tht cn extend the utiliztion of the system resources in other spects re s follows: Time diversity: In the present work, we ssume coding whose block length is mtched to the coherence time of the chnnel. This is rther limiting ssumption, given the dely bound in the order of multiple coherence times we considered in this pper. We hope to extend our optimiztion of sptil diversity to the time-diversity, in form of coding over multiple coherence times (see [5]) or hybrid ARQ (see [3], [6], nd [7]). Controlling SNR: In the current formultion, the verge SNR is fixed cross ll users nd time. In the bsence of CSI t the trnsmitters nd of dely constrints (D = ), this fixed SNR ssumption incurs no loss in optimlity. But for smll nd medium D, it is nturl to expect tht n improvement in performnce is possible when the trnsmit powers for users re functions of the queue sttes. Dynmic control of PHY lyer operting point: In this pper, we ssumed tht the sole responsibility of the dynmic scheduler is to control the trnsmission rtes or multiplexing gins while the diversity gin nd the sum of the rtes of ll users re kept constnt. It is cler tht llowing for dynmic control of the diversity gin, s proposed by [7], will only improve the sttisticlmultiplexing gin nd the performnce of the system. This cn nturlly be extended to time-diversity, nd is roughly relted to dynmic control of the trdeoff mong vrious forms of diversity gin. In this context, our work cn be viewed s providing lower bound on the performnce of n optimlly designed system. ACKNOWLEDGMENT This reserch is supported in prt by the Center for Wireless Communictions, UCSD nd UC Discovery Grnt No. Com4-76, ARO-MURI grnt No. W9NF , NSF CAREER Awrd No. CNS-53335, nd AFOSR Grnt No. FA REFERENCES [] L. Zheng nd D. Tse, Diversity-multiplexing: fundmentl trdeoff in multiple-ntenn chnnels, IEEE Trns. Info. Theory, v. 49, no. 5, My 23. [2] D. Tse, P. Viswnth, nd L. Zheng, Diversity-multiplexing trdeoff in multiple-ccess chnnels, IEEE Trns. Info. Theory, v. 5, no. 9, Sept 24. [3] D. Bertsims, I. C. Pschlidis, nd J. N. Tsitsiklis, Asymptotic buffer overflow probbilities in multiclss multiplexers: n optiml control pproch, IEEE Trns. Automtic Control, v. 43, no. 3, Mrch 998. [4] A. L. Stolyr nd K. Rmnn, Lrgest weighted dely first scheduling: Lrge devitions nd optimlity. Ann. Applied Probbility, v., pp. -48, 2. [5] T. Hollidy nd A. Goldsmith, Joint source nd chnnel coding for MIMO systems, Proc. Allerton Conf., 24. [6] T. Hollidy nd A. Goldsmith, Optimizing End-to-End Distortion in MIMO Systems, IEEE ISIT 5, 25. [7] T. Hollidy, A. Goldsmith, nd H. V. Poor, The Impct of Dely on the Diversity Multiplexing ARQ Trdeoff, ICC, 26. [8] F. Kelly, Notes on effective bndwidths, Stochstic Networks: Theory nd Applictions, vol. 4, 996. [9] A. Dembo nd O. Zeitouni, Lrge Devition Techniques nd Applictions, Boston, MA: Jones nd Brtlett, 992. [] A. Dembo nd T. Zjic, Lrge devitions: from empiricl men nd mesure to prtil sum process, Stochstic Processes nd their Apps., v. 57, 995. [] A. Gnesh, N. O Connel, nd D. Wischik, Big Queues, Springer-Verlg, Berlin, 24. [2] J. Wlrnd nd P. Vriy, High-Performnce Communiction Networks, 2nd ed., Morgn Kufmnn Publishers, 2. [3] H. El Gml, G. Cire, nd M. O. Dmen, The MIMO ARQ chnnel: diversity-multiplexing-dely trdeoff, IEEE Trns. Info. Theory, v. 52, Aug 26. [4] S. Kittipiykul nd T. Jvidi, Optiml operting point in MIMO chnnel for dely-sensitive nd bursty trffic, IEEE ISIT 6, July 26. [5] P. Eli, S. Kittipiykul, nd T. Jvidi, On the Responsiveness-Diversity- Multiplexing trdeoff, WiOpt, 27. APPENDIX I ADDITIONAL ASSUMPTION ON SOURCE MODEL Here we give the dditionl ssumption B on the source we consider in this pper. Assumption B is required in the proof of Lemm 2. Assumption B: Smple pth LDP (see [3], [9] nd []) For n rrivl sequence S, S 2,..., for ll m N, for every ǫ, ǫ 2 >, nd for every sclr,..., m, there exists M > such tht for ll n M nd ll k,..., k m with = k k k m = n, m exp nǫ 2 (k i+ k i )Λ ( i ) i= Pr [ Ski+ S ki (k i+ k i ) i nǫ, i =,..., m ] m exp nǫ 2 (k i+ k i )Λ ( i ) i= APPENDIX II PROOFS OF LEMMAS AND PROPOSITION (34) Proof of Lemm : Proof: Since the symmetric sttic scheduler lwys ssigns the service rte C v to ech queue, we hve tht Pr[W > D] = Pr[L > DC v ]. (35) To be more specific, the sttement holds becuse ny bits delyed more thn D timeslots see t lest DC v bits before them, nd ny bits delyed less thn D timeslots must see less thn DC v bits before them. This is vlid becuse of the first-come-first-serve discipline ssumption. Hence, the two events W > D nd L > DC v re equivlent nd hve the sme probbility. Now, since Pr[L > DC v ] is equl to the til probbility for buffer which is served t fixed cpcity of C v nd whose rrivl process is described by Λ( ) nd stisfying LDP, one cn clculte the til probbility for single queue system with fixed service rte c s (see [8], [] nd [2]) lim B B log Pr[L > B] = θ

10 where θ is the lrgest positive root of eqution Λ(θ) θ = c. By replcing B with DC v nd c with C v nd using (), we hve log Pr[L > DC v ] lim = σ s DTr v ρ log ρ where σ s is given s the solution to Λ(σ s ) = σ s C v. Proof of Lemm 2: Before showing the proof of Lemm 2, we recll the following result on the symptotic til probbility of the mximl weighted dely under the longest-weighted-delyfirst (LWDF) scheduling discipline from [4] nd simplify the result to our specific ssumptions of symmetric users with LDF scheduling discipline. Fct 4: (Theorem 2.2 in [4]) Consider single server of fixed service rte nd K mutully independent source processes with sttionry increments. The totl number of informtion bits generted by source i (i =,...,K) is given by sequence Ŝi t, t =, 2,... where Ŝi t is the cumultive totl number of work rrived until time t from source i. We ssume Ŝi t, t =, 2,... stisfies LDP nd smple pth LDP (Assumptions A nd B) with the convex decy function ˆΛ i nd the convex log moment generting function ˆΛ i. Assume KE[Ŝ ] < for stbility. Let α i be the weight for user i (ssuming < α α 2 α K ). Consider the longestweighted-dely-first (LWDF) scheduling discipline, which lwys ssign the server to the longest witing (i.e. hed-ofthe-line)) customer of the source i which hs the mximl weighted dely. Then, the LWDF scheduling discipline mximizes the exponentil decy rte of the sttionry distribution of the mximl dely, mong ll cusl nd work-conserving scheduling disciplines. Furthermore, the probbility is given s lim sup n log Pr n [ n mx i,...,k ŵi > ] J (36) where nd ŵ i is the sttionry dely for user i, i =...,K, nd J is given s: subject to nd J = min j;x,...,x j γ j ( α i γ)ˆλ i (x i) (37) i= j,...,k,x i >, j x i > (38) i= j i= < γ = x i α j j+ i= α (39) ix i α j with α K+. Since in this pper we consider symmetric users nd LDF scheduling which is the LWDF discipline with equl weights, the following corollry gives specific expression of J which will be used to show Lemm 2. Corollry : Under the ssumptions of symmetric users with equl weights, i.e. ˆΛ i = ˆΛ, ˆΛ i = ˆΛ, nd α i = for ll i =,...,K, J in Fct 4 is reduced to J = sup θ : ˆΛ(θ) θ/k. (4) Proof: Under the ssumption of equl weights (i.e. α i = for ll i =,...,K), there re fesible vlues of γ in (39) only when j = K. Hence, the minimiztion in (37) is reduced to subject to nd J = γ min x,...,x K γ K ˆΛ (x i ) (4) i= K x i >, x i >, i =,...,K (42) i= K i= = < γ = x i α K K+ i= x. (43) i However, we notice tht condition (43) is stisfied with ny choices of x i stisfying condition (42). Hence, plugging the expression of γ into (4), we get J = min x,...,x K K i= x i K ˆΛ (x i ) (44) subject to (42). We cn simplify J further by using the convexity property of ˆΛ, i.e. K K i= i= ˆΛ (x i ) ˆΛ ( K i= x i K ) =: ˆΛ ( K + K ), where we let be such tht K = K i= x i > by the condition in (42). The equlity holds when x i = + /K for ll i =,..., K. Hence, we cn rewrite (44) nd its conditions concisely s ) J = min > ˆΛ ( + K. (45) To finish the proof, we expnd ˆΛ using its definition, s follows: J = min > ˆΛ ( + /K) = min > sup θ( + /K) ˆΛ(θ) θ R = sup min θ R > = sup θ R = sup, θ + θ/k ˆΛ(θ) if θ/k < ˆΛ(θ), θ, if θ/k ˆΛ(θ) θ : ˆΛ(θ) θ/k, where the third equlity holds becuse the function θ + is convex on nd concve on θ (since ˆΛ is convex). θ/k ˆΛ(θ) Proof: (Lemm 2) Consider scled version of the system in Fct 4 where the service rte is scled to C (which

11 is equl to KC v ) nd the rrivls re lso scled up by C. We cn think of this scling s chnge of mesurement units. We denote Ws i s the sttionry dely of rrivls of user i, i =,...,K, for this scled single-server system with LDF scheduling. Since scling of the service nd rrivls do not chnge the distribution of the delys, we hve from Fct 4 tht lim sup n [ ] n log P n mx W s i > J. (46) i,...,k Noticing tht the log moment generting function Λ of the scled system is given s Λ(θ) = lim t t log E[eθCŜ t ] = ˆΛ(θC), (47) we hve, by using Corollry, = sup θ : ˆΛ(θ) θ/k J = sup θ : Λ(θ/C) θ/k = C sup θ : Λ( θ) θc/k = θcv = Cσ s (48) where σ s > is defined s the unique solution to Λ(σ s ) = σ s C v. The second equlity in (48) follows by using (47); the third equility follows by letting θ = θ/c; nd the lst equlity by using tht fct tht Λ is strictly convex nd Λ () = λt log ρ < C v (the stbility condition in (9) nd the fct tht ˆΛ () is the verge rrivl rte per source [8]) nd hence the supremum is ttined with θ = σ s. Replcing n with D nd J = Cσ s = KC v σ s = Kσ s Tr v log ρ in (46), we hve [ ] P mx W s i.= > D ρ Kσ sdtr v, i,...,k for lrge vlue of D. From symmetry, on the other hnd, we hve [ ] P[Ws > D] P mx W s i > D KP[Ws > D]. i,...,k This provides the ssertion of the lemm: P l := P[W s > D]. = ρ KσsDTrv. Proof: The limiting log moment generting function Λ( ) for compound Poisson source with exponentil pcket length is derived in [4], which is Λ(θ) = lim n n log E[ exp(θst )] [ ] n = lim n n log E exp(θ A t) t= ] = log E [ exp θa = νθ µ θ if θ < µ, if θ µ. (49) From Lemm, σ s is the solution to Λ(σ s ) = σ s C v. From (49), this reduces to finding σ s such tht νσ s µ σ s = σ s C v µλt log ρ µ σ s = r v T log ρ where we hve replced ν nd C v from (32) nd (). Hence, σ s = µ( λ/r v ). Proof of Theorem : Proof: We first show the existence of solution d in (28) of Algorithm. The LHS term is decresing on d nd equl to for d d, for some d such tht the rrivl rte λt log ρ is equl to the service rte C v ( d) (in which cse, σ s ( d) = ). On nother hnd, the RHS term is incresing on d nd is equl to when d =. Hence, (28) must hold for some d (, d). Next, if d solving (28) is less thn d, this d is the lower bound d l (i.e. symptoticlly mximizing the RHS term in (27)) nd the upper bound d u is obtined from mximizing the LHS in (27). The existence of d solving (29) cn be shown similrly. Proof of Proposition I:

The Regulated and Riemann Integrals

The Regulated and Riemann Integrals Chpter 1 The Regulted nd Riemnn Integrls 1.1 Introduction We will consider severl different pproches to defining the definite integrl f(x) dx of function f(x). These definitions will ll ssign the sme vlue

More information

Lecture 1. Functional series. Pointwise and uniform convergence.

Lecture 1. Functional series. Pointwise and uniform convergence. 1 Introduction. Lecture 1. Functionl series. Pointwise nd uniform convergence. In this course we study mongst other things Fourier series. The Fourier series for periodic function f(x) with period 2π is

More information

Tests for the Ratio of Two Poisson Rates

Tests for the Ratio of Two Poisson Rates Chpter 437 Tests for the Rtio of Two Poisson Rtes Introduction The Poisson probbility lw gives the probbility distribution of the number of events occurring in specified intervl of time or spce. The Poisson

More information

Online Supplements to Performance-Based Contracts for Outpatient Medical Services

Online Supplements to Performance-Based Contracts for Outpatient Medical Services Jing, Png nd Svin: Performnce-bsed Contrcts Article submitted to Mnufcturing & Service Opertions Mngement; mnuscript no. MSOM-11-270.R2 1 Online Supplements to Performnce-Bsed Contrcts for Outptient Medicl

More information

Continuous Random Variables

Continuous Random Variables STAT/MATH 395 A - PROBABILITY II UW Winter Qurter 217 Néhémy Lim Continuous Rndom Vribles Nottion. The indictor function of set S is rel-vlued function defined by : { 1 if x S 1 S (x) if x S Suppose tht

More information

Advanced Calculus: MATH 410 Notes on Integrals and Integrability Professor David Levermore 17 October 2004

Advanced Calculus: MATH 410 Notes on Integrals and Integrability Professor David Levermore 17 October 2004 Advnced Clculus: MATH 410 Notes on Integrls nd Integrbility Professor Dvid Levermore 17 October 2004 1. Definite Integrls In this section we revisit the definite integrl tht you were introduced to when

More information

THE EXISTENCE-UNIQUENESS THEOREM FOR FIRST-ORDER DIFFERENTIAL EQUATIONS.

THE EXISTENCE-UNIQUENESS THEOREM FOR FIRST-ORDER DIFFERENTIAL EQUATIONS. THE EXISTENCE-UNIQUENESS THEOREM FOR FIRST-ORDER DIFFERENTIAL EQUATIONS RADON ROSBOROUGH https://intuitiveexplntionscom/picrd-lindelof-theorem/ This document is proof of the existence-uniqueness theorem

More information

Review of basic calculus

Review of basic calculus Review of bsic clculus This brief review reclls some of the most importnt concepts, definitions, nd theorems from bsic clculus. It is not intended to tech bsic clculus from scrtch. If ny of the items below

More information

Recitation 3: More Applications of the Derivative

Recitation 3: More Applications of the Derivative Mth 1c TA: Pdric Brtlett Recittion 3: More Applictions of the Derivtive Week 3 Cltech 2012 1 Rndom Question Question 1 A grph consists of the following: A set V of vertices. A set E of edges where ech

More information

Review of Calculus, cont d

Review of Calculus, cont d Jim Lmbers MAT 460 Fll Semester 2009-10 Lecture 3 Notes These notes correspond to Section 1.1 in the text. Review of Clculus, cont d Riemnn Sums nd the Definite Integrl There re mny cses in which some

More information

Chapter 5 : Continuous Random Variables

Chapter 5 : Continuous Random Variables STAT/MATH 395 A - PROBABILITY II UW Winter Qurter 216 Néhémy Lim Chpter 5 : Continuous Rndom Vribles Nottions. N {, 1, 2,...}, set of nturl numbers (i.e. ll nonnegtive integers); N {1, 2,...}, set of ll

More information

Properties of Integrals, Indefinite Integrals. Goals: Definition of the Definite Integral Integral Calculations using Antiderivatives

Properties of Integrals, Indefinite Integrals. Goals: Definition of the Definite Integral Integral Calculations using Antiderivatives Block #6: Properties of Integrls, Indefinite Integrls Gols: Definition of the Definite Integrl Integrl Clcultions using Antiderivtives Properties of Integrls The Indefinite Integrl 1 Riemnn Sums - 1 Riemnn

More information

W. We shall do so one by one, starting with I 1, and we shall do it greedily, trying

W. We shall do so one by one, starting with I 1, and we shall do it greedily, trying Vitli covers 1 Definition. A Vitli cover of set E R is set V of closed intervls with positive length so tht, for every δ > 0 nd every x E, there is some I V with λ(i ) < δ nd x I. 2 Lemm (Vitli covering)

More information

The First Fundamental Theorem of Calculus. If f(x) is continuous on [a, b] and F (x) is any antiderivative. f(x) dx = F (b) F (a).

The First Fundamental Theorem of Calculus. If f(x) is continuous on [a, b] and F (x) is any antiderivative. f(x) dx = F (b) F (a). The Fundmentl Theorems of Clculus Mth 4, Section 0, Spring 009 We now know enough bout definite integrls to give precise formultions of the Fundmentl Theorems of Clculus. We will lso look t some bsic emples

More information

1.9 C 2 inner variations

1.9 C 2 inner variations 46 CHAPTER 1. INDIRECT METHODS 1.9 C 2 inner vritions So fr, we hve restricted ttention to liner vritions. These re vritions of the form vx; ǫ = ux + ǫφx where φ is in some liner perturbtion clss P, for

More information

A SHORT NOTE ON THE MONOTONICITY OF THE ERLANG C FORMULA IN THE HALFIN-WHITT REGIME. Bernardo D Auria 1

A SHORT NOTE ON THE MONOTONICITY OF THE ERLANG C FORMULA IN THE HALFIN-WHITT REGIME. Bernardo D Auria 1 Working Pper 11-42 (31) Sttistics nd Econometrics Series December, 2011 Deprtmento de Estdístic Universidd Crlos III de Mdrid Clle Mdrid, 126 28903 Getfe (Spin) Fx (34) 91 624-98-49 A SHORT NOTE ON THE

More information

Unit #9 : Definite Integral Properties; Fundamental Theorem of Calculus

Unit #9 : Definite Integral Properties; Fundamental Theorem of Calculus Unit #9 : Definite Integrl Properties; Fundmentl Theorem of Clculus Gols: Identify properties of definite integrls Define odd nd even functions, nd reltionship to integrl vlues Introduce the Fundmentl

More information

8 Laplace s Method and Local Limit Theorems

8 Laplace s Method and Local Limit Theorems 8 Lplce s Method nd Locl Limit Theorems 8. Fourier Anlysis in Higher DImensions Most of the theorems of Fourier nlysis tht we hve proved hve nturl generliztions to higher dimensions, nd these cn be proved

More information

State space systems analysis (continued) Stability. A. Definitions A system is said to be Asymptotically Stable (AS) when it satisfies

State space systems analysis (continued) Stability. A. Definitions A system is said to be Asymptotically Stable (AS) when it satisfies Stte spce systems nlysis (continued) Stbility A. Definitions A system is sid to be Asymptoticlly Stble (AS) when it stisfies ut () = 0, t > 0 lim xt () 0. t A system is AS if nd only if the impulse response

More information

Math 1B, lecture 4: Error bounds for numerical methods

Math 1B, lecture 4: Error bounds for numerical methods Mth B, lecture 4: Error bounds for numericl methods Nthn Pflueger 4 September 0 Introduction The five numericl methods descried in the previous lecture ll operte by the sme principle: they pproximte the

More information

Non-Linear & Logistic Regression

Non-Linear & Logistic Regression Non-Liner & Logistic Regression If the sttistics re boring, then you've got the wrong numbers. Edwrd R. Tufte (Sttistics Professor, Yle University) Regression Anlyses When do we use these? PART 1: find

More information

S. S. Dragomir. 2, we have the inequality. b a

S. S. Dragomir. 2, we have the inequality. b a Bull Koren Mth Soc 005 No pp 3 30 SOME COMPANIONS OF OSTROWSKI S INEQUALITY FOR ABSOLUTELY CONTINUOUS FUNCTIONS AND APPLICATIONS S S Drgomir Abstrct Compnions of Ostrowski s integrl ineulity for bsolutely

More information

20 MATHEMATICS POLYNOMIALS

20 MATHEMATICS POLYNOMIALS 0 MATHEMATICS POLYNOMIALS.1 Introduction In Clss IX, you hve studied polynomils in one vrible nd their degrees. Recll tht if p(x) is polynomil in x, the highest power of x in p(x) is clled the degree of

More information

1 Online Learning and Regret Minimization

1 Online Learning and Regret Minimization 2.997 Decision-Mking in Lrge-Scle Systems My 10 MIT, Spring 2004 Hndout #29 Lecture Note 24 1 Online Lerning nd Regret Minimiztion In this lecture, we consider the problem of sequentil decision mking in

More information

Duality # Second iteration for HW problem. Recall our LP example problem we have been working on, in equality form, is given below.

Duality # Second iteration for HW problem. Recall our LP example problem we have been working on, in equality form, is given below. Dulity #. Second itertion for HW problem Recll our LP emple problem we hve been working on, in equlity form, is given below.,,,, 8 m F which, when written in slightly different form, is 8 F Recll tht we

More information

ECO 317 Economics of Uncertainty Fall Term 2007 Notes for lectures 4. Stochastic Dominance

ECO 317 Economics of Uncertainty Fall Term 2007 Notes for lectures 4. Stochastic Dominance Generl structure ECO 37 Economics of Uncertinty Fll Term 007 Notes for lectures 4. Stochstic Dominnce Here we suppose tht the consequences re welth mounts denoted by W, which cn tke on ny vlue between

More information

APPROXIMATE INTEGRATION

APPROXIMATE INTEGRATION APPROXIMATE INTEGRATION. Introduction We hve seen tht there re functions whose nti-derivtives cnnot be expressed in closed form. For these resons ny definite integrl involving these integrnds cnnot be

More information

Heat flux and total heat

Heat flux and total heat Het flux nd totl het John McCun Mrch 14, 2017 1 Introduction Yesterdy (if I remember correctly) Ms. Prsd sked me question bout the condition of insulted boundry for the 1D het eqution, nd (bsed on glnce

More information

p-adic Egyptian Fractions

p-adic Egyptian Fractions p-adic Egyptin Frctions Contents 1 Introduction 1 2 Trditionl Egyptin Frctions nd Greedy Algorithm 2 3 Set-up 3 4 p-greedy Algorithm 5 5 p-egyptin Trditionl 10 6 Conclusion 1 Introduction An Egyptin frction

More information

P 3 (x) = f(0) + f (0)x + f (0) 2. x 2 + f (0) . In the problem set, you are asked to show, in general, the n th order term is a n = f (n) (0)

P 3 (x) = f(0) + f (0)x + f (0) 2. x 2 + f (0) . In the problem set, you are asked to show, in general, the n th order term is a n = f (n) (0) 1 Tylor polynomils In Section 3.5, we discussed how to pproximte function f(x) round point in terms of its first derivtive f (x) evluted t, tht is using the liner pproximtion f() + f ()(x ). We clled this

More information

New Expansion and Infinite Series

New Expansion and Infinite Series Interntionl Mthemticl Forum, Vol. 9, 204, no. 22, 06-073 HIKARI Ltd, www.m-hikri.com http://dx.doi.org/0.2988/imf.204.4502 New Expnsion nd Infinite Series Diyun Zhng College of Computer Nnjing University

More information

Acceptance Sampling by Attributes

Acceptance Sampling by Attributes Introduction Acceptnce Smpling by Attributes Acceptnce smpling is concerned with inspection nd decision mking regrding products. Three spects of smpling re importnt: o Involves rndom smpling of n entire

More information

Riemann is the Mann! (But Lebesgue may besgue to differ.)

Riemann is the Mann! (But Lebesgue may besgue to differ.) Riemnn is the Mnn! (But Lebesgue my besgue to differ.) Leo Livshits My 2, 2008 1 For finite intervls in R We hve seen in clss tht every continuous function f : [, b] R hs the property tht for every ɛ >

More information

Numerical integration

Numerical integration 2 Numericl integrtion This is pge i Printer: Opque this 2. Introduction Numericl integrtion is problem tht is prt of mny problems in the economics nd econometrics literture. The orgniztion of this chpter

More information

Entropy and Ergodic Theory Notes 10: Large Deviations I

Entropy and Ergodic Theory Notes 10: Large Deviations I Entropy nd Ergodic Theory Notes 10: Lrge Devitions I 1 A chnge of convention This is our first lecture on pplictions of entropy in probbility theory. In probbility theory, the convention is tht ll logrithms

More information

Stuff You Need to Know From Calculus

Stuff You Need to Know From Calculus Stuff You Need to Know From Clculus For the first time in the semester, the stuff we re doing is finlly going to look like clculus (with vector slnt, of course). This mens tht in order to succeed, you

More information

Goals: Determine how to calculate the area described by a function. Define the definite integral. Explore the relationship between the definite

Goals: Determine how to calculate the area described by a function. Define the definite integral. Explore the relationship between the definite Unit #8 : The Integrl Gols: Determine how to clculte the re described by function. Define the definite integrl. Eplore the reltionship between the definite integrl nd re. Eplore wys to estimte the definite

More information

Numerical Analysis: Trapezoidal and Simpson s Rule

Numerical Analysis: Trapezoidal and Simpson s Rule nd Simpson s Mthemticl question we re interested in numericlly nswering How to we evlute I = f (x) dx? Clculus tells us tht if F(x) is the ntiderivtive of function f (x) on the intervl [, b], then I =

More information

Conservation Law. Chapter Goal. 5.2 Theory

Conservation Law. Chapter Goal. 5.2 Theory Chpter 5 Conservtion Lw 5.1 Gol Our long term gol is to understnd how mny mthemticl models re derived. We study how certin quntity chnges with time in given region (sptil domin). We first derive the very

More information

MATH34032: Green s Functions, Integral Equations and the Calculus of Variations 1

MATH34032: Green s Functions, Integral Equations and the Calculus of Variations 1 MATH34032: Green s Functions, Integrl Equtions nd the Clculus of Vritions 1 Section 1 Function spces nd opertors Here we gives some brief detils nd definitions, prticulrly relting to opertors. For further

More information

Lecture 14: Quadrature

Lecture 14: Quadrature Lecture 14: Qudrture This lecture is concerned with the evlution of integrls fx)dx 1) over finite intervl [, b] The integrnd fx) is ssumed to be rel-vlues nd smooth The pproximtion of n integrl by numericl

More information

Calculus of Variations

Calculus of Variations Clculus of Vritions Com S 477/577 Notes) Yn-Bin Ji Dec 4, 2017 1 Introduction A functionl ssigns rel number to ech function or curve) in some clss. One might sy tht functionl is function of nother function

More information

MATH 144: Business Calculus Final Review

MATH 144: Business Calculus Final Review MATH 144: Business Clculus Finl Review 1 Skills 1. Clculte severl limits. 2. Find verticl nd horizontl symptotes for given rtionl function. 3. Clculte derivtive by definition. 4. Clculte severl derivtives

More information

SUMMER KNOWHOW STUDY AND LEARNING CENTRE

SUMMER KNOWHOW STUDY AND LEARNING CENTRE SUMMER KNOWHOW STUDY AND LEARNING CENTRE Indices & Logrithms 2 Contents Indices.2 Frctionl Indices.4 Logrithms 6 Exponentil equtions. Simplifying Surds 13 Opertions on Surds..16 Scientific Nottion..18

More information

Physics 116C Solution of inhomogeneous ordinary differential equations using Green s functions

Physics 116C Solution of inhomogeneous ordinary differential equations using Green s functions Physics 6C Solution of inhomogeneous ordinry differentil equtions using Green s functions Peter Young November 5, 29 Homogeneous Equtions We hve studied, especilly in long HW problem, second order liner

More information

f(x)dx . Show that there 1, 0 < x 1 does not exist a differentiable function g : [ 1, 1] R such that g (x) = f(x) for all

f(x)dx . Show that there 1, 0 < x 1 does not exist a differentiable function g : [ 1, 1] R such that g (x) = f(x) for all 3 Definite Integrl 3.1 Introduction In school one comes cross the definition of the integrl of rel vlued function defined on closed nd bounded intervl [, b] between the limits nd b, i.e., f(x)dx s the

More information

The steps of the hypothesis test

The steps of the hypothesis test ttisticl Methods I (EXT 7005) Pge 78 Mosquito species Time of dy A B C Mid morning 0.0088 5.4900 5.5000 Mid Afternoon.3400 0.0300 0.8700 Dusk 0.600 5.400 3.000 The Chi squre test sttistic is the sum of

More information

Analysis of PDCCH performance for M2M traffic in LTE

Analysis of PDCCH performance for M2M traffic in LTE 1 Anlysis of PDCCH performnce for M2M trffic in LTE Prjwl Osti, Psi Lssil, Smuli Alto, Ann Lrmo, Tuoms Tirronen Abstrct As LTE is strting to get widely deployed, the volume of M2M trffic is incresing very

More information

5.3 Nonlinear stability of Rayleigh-Bénard convection

5.3 Nonlinear stability of Rayleigh-Bénard convection 118 5.3 Nonliner stbility of Ryleigh-Bénrd convection In Chpter 1, we sw tht liner stbility only tells us whether system is stble or unstble to infinitesimlly-smll perturbtions, nd tht there re cses in

More information

Recitation 3: Applications of the Derivative. 1 Higher-Order Derivatives and their Applications

Recitation 3: Applications of the Derivative. 1 Higher-Order Derivatives and their Applications Mth 1c TA: Pdric Brtlett Recittion 3: Applictions of the Derivtive Week 3 Cltech 013 1 Higher-Order Derivtives nd their Applictions Another thing we could wnt to do with the derivtive, motivted by wht

More information

Theoretical foundations of Gaussian quadrature

Theoretical foundations of Gaussian quadrature Theoreticl foundtions of Gussin qudrture 1 Inner product vector spce Definition 1. A vector spce (or liner spce) is set V = {u, v, w,...} in which the following two opertions re defined: (A) Addition of

More information

Week 10: Line Integrals

Week 10: Line Integrals Week 10: Line Integrls Introduction In this finl week we return to prmetrised curves nd consider integrtion long such curves. We lredy sw this in Week 2 when we integrted long curve to find its length.

More information

Chapter 4 Contravariance, Covariance, and Spacetime Diagrams

Chapter 4 Contravariance, Covariance, and Spacetime Diagrams Chpter 4 Contrvrince, Covrince, nd Spcetime Digrms 4. The Components of Vector in Skewed Coordintes We hve seen in Chpter 3; figure 3.9, tht in order to show inertil motion tht is consistent with the Lorentz

More information

Chapter 14. Matrix Representations of Linear Transformations

Chapter 14. Matrix Representations of Linear Transformations Chpter 4 Mtrix Representtions of Liner Trnsformtions When considering the Het Stte Evolution, we found tht we could describe this process using multipliction by mtrix. This ws nice becuse computers cn

More information

Lecture 3 ( ) (translated and slightly adapted from lecture notes by Martin Klazar)

Lecture 3 ( ) (translated and slightly adapted from lecture notes by Martin Klazar) Lecture 3 (5.3.2018) (trnslted nd slightly dpted from lecture notes by Mrtin Klzr) Riemnn integrl Now we define precisely the concept of the re, in prticulr, the re of figure U(, b, f) under the grph of

More information

Monte Carlo method in solving numerical integration and differential equation

Monte Carlo method in solving numerical integration and differential equation Monte Crlo method in solving numericl integrtion nd differentil eqution Ye Jin Chemistry Deprtment Duke University yj66@duke.edu Abstrct: Monte Crlo method is commonly used in rel physics problem. The

More information

Integral points on the rational curve

Integral points on the rational curve Integrl points on the rtionl curve y x bx c x ;, b, c integers. Konstntine Zeltor Mthemtics University of Wisconsin - Mrinette 750 W. Byshore Street Mrinette, WI 5443-453 Also: Konstntine Zeltor P.O. Box

More information

Fig. 1. Open-Loop and Closed-Loop Systems with Plant Variations

Fig. 1. Open-Loop and Closed-Loop Systems with Plant Variations ME 3600 Control ystems Chrcteristics of Open-Loop nd Closed-Loop ystems Importnt Control ystem Chrcteristics o ensitivity of system response to prmetric vritions cn be reduced o rnsient nd stedy-stte responses

More information

Quadratic Forms. Quadratic Forms

Quadratic Forms. Quadratic Forms Qudrtic Forms Recll the Simon & Blume excerpt from n erlier lecture which sid tht the min tsk of clculus is to pproximte nonliner functions with liner functions. It s ctully more ccurte to sy tht we pproximte

More information

5.7 Improper Integrals

5.7 Improper Integrals 458 pplictions of definite integrls 5.7 Improper Integrls In Section 5.4, we computed the work required to lift pylod of mss m from the surfce of moon of mss nd rdius R to height H bove the surfce of the

More information

The Wave Equation I. MA 436 Kurt Bryan

The Wave Equation I. MA 436 Kurt Bryan 1 Introduction The Wve Eqution I MA 436 Kurt Bryn Consider string stretching long the x xis, of indeterminte (or even infinite!) length. We wnt to derive n eqution which models the motion of the string

More information

X Z Y Table 1: Possibles values for Y = XZ. 1, p

X Z Y Table 1: Possibles values for Y = XZ. 1, p ECE 534: Elements of Informtion Theory, Fll 00 Homework 7 Solutions ll by Kenneth Plcio Bus October 4, 00. Problem 7.3. Binry multiplier chnnel () Consider the chnnel Y = XZ, where X nd Z re independent

More information

Reinforcement Learning

Reinforcement Learning Reinforcement Lerning Tom Mitchell, Mchine Lerning, chpter 13 Outline Introduction Comprison with inductive lerning Mrkov Decision Processes: the model Optiml policy: The tsk Q Lerning: Q function Algorithm

More information

Discrete Mathematics and Probability Theory Spring 2013 Anant Sahai Lecture 17

Discrete Mathematics and Probability Theory Spring 2013 Anant Sahai Lecture 17 EECS 70 Discrete Mthemtics nd Proility Theory Spring 2013 Annt Shi Lecture 17 I.I.D. Rndom Vriles Estimting the is of coin Question: We wnt to estimte the proportion p of Democrts in the US popultion,

More information

( dg. ) 2 dt. + dt. dt j + dh. + dt. r(t) dt. Comparing this equation with the one listed above for the length of see that

( dg. ) 2 dt. + dt. dt j + dh. + dt. r(t) dt. Comparing this equation with the one listed above for the length of see that Arc Length of Curves in Three Dimensionl Spce If the vector function r(t) f(t) i + g(t) j + h(t) k trces out the curve C s t vries, we cn mesure distnces long C using formul nerly identicl to one tht we

More information

Partial Derivatives. Limits. For a single variable function f (x), the limit lim

Partial Derivatives. Limits. For a single variable function f (x), the limit lim Limits Prtil Derivtives For single vrible function f (x), the limit lim x f (x) exists only if the right-hnd side limit equls to the left-hnd side limit, i.e., lim f (x) = lim f (x). x x + For two vribles

More information

Credibility Hypothesis Testing of Fuzzy Triangular Distributions

Credibility Hypothesis Testing of Fuzzy Triangular Distributions 666663 Journl of Uncertin Systems Vol.9, No., pp.6-74, 5 Online t: www.jus.org.uk Credibility Hypothesis Testing of Fuzzy Tringulr Distributions S. Smpth, B. Rmy Received April 3; Revised 4 April 4 Abstrct

More information

Student Activity 3: Single Factor ANOVA

Student Activity 3: Single Factor ANOVA MATH 40 Student Activity 3: Single Fctor ANOVA Some Bsic Concepts In designed experiment, two or more tretments, or combintions of tretments, is pplied to experimentl units The number of tretments, whether

More information

ARITHMETIC OPERATIONS. The real numbers have the following properties: a b c ab ac

ARITHMETIC OPERATIONS. The real numbers have the following properties: a b c ab ac REVIEW OF ALGEBRA Here we review the bsic rules nd procedures of lgebr tht you need to know in order to be successful in clculus. ARITHMETIC OPERATIONS The rel numbers hve the following properties: b b

More information

4 The dynamical FRW universe

4 The dynamical FRW universe 4 The dynmicl FRW universe 4.1 The Einstein equtions Einstein s equtions G µν = T µν (7) relte the expnsion rte (t) to energy distribution in the universe. On the left hnd side is the Einstein tensor which

More information

Bases for Vector Spaces

Bases for Vector Spaces Bses for Vector Spces 2-26-25 A set is independent if, roughly speking, there is no redundncy in the set: You cn t uild ny vector in the set s liner comintion of the others A set spns if you cn uild everything

More information

Lecture 19: Continuous Least Squares Approximation

Lecture 19: Continuous Least Squares Approximation Lecture 19: Continuous Lest Squres Approximtion 33 Continuous lest squres pproximtion We begn 31 with the problem of pproximting some f C[, b] with polynomil p P n t the discrete points x, x 1,, x m for

More information

NUMERICAL INTEGRATION. The inverse process to differentiation in calculus is integration. Mathematically, integration is represented by.

NUMERICAL INTEGRATION. The inverse process to differentiation in calculus is integration. Mathematically, integration is represented by. NUMERICAL INTEGRATION 1 Introduction The inverse process to differentition in clculus is integrtion. Mthemticlly, integrtion is represented by f(x) dx which stnds for the integrl of the function f(x) with

More information

A REVIEW OF CALCULUS CONCEPTS FOR JDEP 384H. Thomas Shores Department of Mathematics University of Nebraska Spring 2007

A REVIEW OF CALCULUS CONCEPTS FOR JDEP 384H. Thomas Shores Department of Mathematics University of Nebraska Spring 2007 A REVIEW OF CALCULUS CONCEPTS FOR JDEP 384H Thoms Shores Deprtment of Mthemtics University of Nebrsk Spring 2007 Contents Rtes of Chnge nd Derivtives 1 Dierentils 4 Are nd Integrls 5 Multivrite Clculus

More information

Lecture 21: Order statistics

Lecture 21: Order statistics Lecture : Order sttistics Suppose we hve N mesurements of sclr, x i =, N Tke ll mesurements nd sort them into scending order x x x 3 x N Define the mesured running integrl S N (x) = 0 for x < x = i/n for

More information

CS 188 Introduction to Artificial Intelligence Fall 2018 Note 7

CS 188 Introduction to Artificial Intelligence Fall 2018 Note 7 CS 188 Introduction to Artificil Intelligence Fll 2018 Note 7 These lecture notes re hevily bsed on notes originlly written by Nikhil Shrm. Decision Networks In the third note, we lerned bout gme trees

More information

Math 426: Probability Final Exam Practice

Math 426: Probability Final Exam Practice Mth 46: Probbility Finl Exm Prctice. Computtionl problems 4. Let T k (n) denote the number of prtitions of the set {,..., n} into k nonempty subsets, where k n. Argue tht T k (n) kt k (n ) + T k (n ) by

More information

MAA 4212 Improper Integrals

MAA 4212 Improper Integrals Notes by Dvid Groisser, Copyright c 1995; revised 2002, 2009, 2014 MAA 4212 Improper Integrls The Riemnn integrl, while perfectly well-defined, is too restrictive for mny purposes; there re functions which

More information

Consequently, the temperature must be the same at each point in the cross section at x. Let:

Consequently, the temperature must be the same at each point in the cross section at x. Let: HW 2 Comments: L1-3. Derive the het eqution for n inhomogeneous rod where the therml coefficients used in the derivtion of the het eqution for homogeneous rod now become functions of position x in the

More information

Math 61CM - Solutions to homework 9

Math 61CM - Solutions to homework 9 Mth 61CM - Solutions to homework 9 Cédric De Groote November 30 th, 2018 Problem 1: Recll tht the left limit of function f t point c is defined s follows: lim f(x) = l x c if for ny > 0 there exists δ

More information

Reversals of Signal-Posterior Monotonicity for Any Bounded Prior

Reversals of Signal-Posterior Monotonicity for Any Bounded Prior Reversls of Signl-Posterior Monotonicity for Any Bounded Prior Christopher P. Chmbers Pul J. Hely Abstrct Pul Milgrom (The Bell Journl of Economics, 12(2): 380 391) showed tht if the strict monotone likelihood

More information

UNIT 1 FUNCTIONS AND THEIR INVERSES Lesson 1.4: Logarithmic Functions as Inverses Instruction

UNIT 1 FUNCTIONS AND THEIR INVERSES Lesson 1.4: Logarithmic Functions as Inverses Instruction Lesson : Logrithmic Functions s Inverses Prerequisite Skills This lesson requires the use of the following skills: determining the dependent nd independent vribles in n exponentil function bsed on dt from

More information

Infinite Geometric Series

Infinite Geometric Series Infinite Geometric Series Finite Geometric Series ( finite SUM) Let 0 < r < 1, nd let n be positive integer. Consider the finite sum It turns out there is simple lgebric expression tht is equivlent to

More information

We partition C into n small arcs by forming a partition of [a, b] by picking s i as follows: a = s 0 < s 1 < < s n = b.

We partition C into n small arcs by forming a partition of [a, b] by picking s i as follows: a = s 0 < s 1 < < s n = b. Mth 255 - Vector lculus II Notes 4.2 Pth nd Line Integrls We begin with discussion of pth integrls (the book clls them sclr line integrls). We will do this for function of two vribles, but these ides cn

More information

1 Probability Density Functions

1 Probability Density Functions Lis Yn CS 9 Continuous Distributions Lecture Notes #9 July 6, 28 Bsed on chpter by Chris Piech So fr, ll rndom vribles we hve seen hve been discrete. In ll the cses we hve seen in CS 9, this ment tht our

More information

Continuous Random Variables Class 5, Jeremy Orloff and Jonathan Bloom

Continuous Random Variables Class 5, Jeremy Orloff and Jonathan Bloom Lerning Gols Continuous Rndom Vriles Clss 5, 8.05 Jeremy Orloff nd Jonthn Bloom. Know the definition of continuous rndom vrile. 2. Know the definition of the proility density function (pdf) nd cumultive

More information

Chapter 0. What is the Lebesgue integral about?

Chapter 0. What is the Lebesgue integral about? Chpter 0. Wht is the Lebesgue integrl bout? The pln is to hve tutoril sheet ech week, most often on Fridy, (to be done during the clss) where you will try to get used to the ides introduced in the previous

More information

MATH SS124 Sec 39 Concepts summary with examples

MATH SS124 Sec 39 Concepts summary with examples This note is mde for students in MTH124 Section 39 to review most(not ll) topics I think we covered in this semester, nd there s exmples fter these concepts, go over this note nd try to solve those exmples

More information

Review of Gaussian Quadrature method

Review of Gaussian Quadrature method Review of Gussin Qudrture method Nsser M. Asi Spring 006 compiled on Sundy Decemer 1, 017 t 09:1 PM 1 The prolem To find numericl vlue for the integrl of rel vlued function of rel vrile over specific rnge

More information

Math 270A: Numerical Linear Algebra

Math 270A: Numerical Linear Algebra Mth 70A: Numericl Liner Algebr Instructor: Michel Holst Fll Qurter 014 Homework Assignment #3 Due Give to TA t lest few dys before finl if you wnt feedbck. Exercise 3.1. (The Bsic Liner Method for Liner

More information

Main topics for the First Midterm

Main topics for the First Midterm Min topics for the First Midterm The Midterm will cover Section 1.8, Chpters 2-3, Sections 4.1-4.8, nd Sections 5.1-5.3 (essentilly ll of the mteril covered in clss). Be sure to know the results of the

More information

f(x) dx, If one of these two conditions is not met, we call the integral improper. Our usual definition for the value for the definite integral

f(x) dx, If one of these two conditions is not met, we call the integral improper. Our usual definition for the value for the definite integral Improper Integrls Every time tht we hve evluted definite integrl such s f(x) dx, we hve mde two implicit ssumptions bout the integrl:. The intervl [, b] is finite, nd. f(x) is continuous on [, b]. If one

More information

Math 360: A primitive integral and elementary functions

Math 360: A primitive integral and elementary functions Mth 360: A primitive integrl nd elementry functions D. DeTurck University of Pennsylvni October 16, 2017 D. DeTurck Mth 360 001 2017C: Integrl/functions 1 / 32 Setup for the integrl prtitions Definition:

More information

Best Approximation. Chapter The General Case

Best Approximation. Chapter The General Case Chpter 4 Best Approximtion 4.1 The Generl Cse In the previous chpter, we hve seen how n interpolting polynomil cn be used s n pproximtion to given function. We now wnt to find the best pproximtion to given

More information

S. S. Dragomir. 1. Introduction. In [1], Guessab and Schmeisser have proved among others, the following companion of Ostrowski s inequality:

S. S. Dragomir. 1. Introduction. In [1], Guessab and Schmeisser have proved among others, the following companion of Ostrowski s inequality: FACTA UNIVERSITATIS NIŠ) Ser Mth Inform 9 00) 6 SOME COMPANIONS OF OSTROWSKI S INEQUALITY FOR ABSOLUTELY CONTINUOUS FUNCTIONS AND APPLICATIONS S S Drgomir Dedicted to Prof G Mstroinni for his 65th birthdy

More information

MAT 168: Calculus II with Analytic Geometry. James V. Lambers

MAT 168: Calculus II with Analytic Geometry. James V. Lambers MAT 68: Clculus II with Anlytic Geometry Jmes V. Lmbers Februry 7, Contents Integrls 5. Introduction............................ 5.. Differentil Clculus nd Quotient Formuls...... 5.. Integrl Clculus nd

More information

Overview of Calculus I

Overview of Calculus I Overview of Clculus I Prof. Jim Swift Northern Arizon University There re three key concepts in clculus: The limit, the derivtive, nd the integrl. You need to understnd the definitions of these three things,

More information

Exam 2, Mathematics 4701, Section ETY6 6:05 pm 7:40 pm, March 31, 2016, IH-1105 Instructor: Attila Máté 1

Exam 2, Mathematics 4701, Section ETY6 6:05 pm 7:40 pm, March 31, 2016, IH-1105 Instructor: Attila Máté 1 Exm, Mthemtics 471, Section ETY6 6:5 pm 7:4 pm, Mrch 1, 16, IH-115 Instructor: Attil Máté 1 17 copies 1. ) Stte the usul sufficient condition for the fixed-point itertion to converge when solving the eqution

More information

Improper Integrals, and Differential Equations

Improper Integrals, and Differential Equations Improper Integrls, nd Differentil Equtions October 22, 204 5.3 Improper Integrls Previously, we discussed how integrls correspond to res. More specificlly, we sid tht for function f(x), the region creted

More information

63. Representation of functions as power series Consider a power series. ( 1) n x 2n for all 1 < x < 1

63. Representation of functions as power series Consider a power series. ( 1) n x 2n for all 1 < x < 1 3 9. SEQUENCES AND SERIES 63. Representtion of functions s power series Consider power series x 2 + x 4 x 6 + x 8 + = ( ) n x 2n It is geometric series with q = x 2 nd therefore it converges for ll q =

More information