Fixed Point Analysis of Single Cell IEEE e WLANs: Uniqueness, Multistability and Throughput Differentiation

Size: px
Start display at page:

Download "Fixed Point Analysis of Single Cell IEEE e WLANs: Uniqueness, Multistability and Throughput Differentiation"

Transcription

1 Fxed Pont Analyss of Sngle Cell IEEE 82.e WLANs: Unqueness, Multstablty and Throughput Dfferentaton Venkatesh Ramayan ECE Department Indan Insttute of Scence Bangalore, Inda Anurag Kumar ECE Department Indan Insttute of Scence Bangalore, Inda Etan Altman INRIA Sopha-Antpols France ABSTRACT We consder the vector fxed pont equatons arsng out of the analyss of the saturaton throughput of a sngle cell IEEE 82.e wreless local area network wth nodes that have dfferent back-off parameters, ncludng dfferent Arbtraton InterFrame Space (AIFS) values. We consder balanced and unbalanced solutons of the fxed pont equatons arsng n homogeneous and nonhomogeneous networks. We are concerned, n partcular, wth () whether the fxed pont s balanced wthn a class, and () whether the fxed pont s unque. Our smulatons show that when multple unbalanced fxed ponts exst n a homogeneous system then the tme behavour of the system demonstrates severe short term unfarness (or multstablty). Implcatons for the use of the fxed pont formulaton for performance analyss are also dscussed. We provde a condton for the fxed pont soluton to be balanced wthn a class, and also a condton for unqueness. We then provde an extenson of our general fxed pont analyss to capture AIFS based dfferentaton; agan a condton for unqueness s establshed. An asymptotc analyss of the fxed pont s provded for the case n whch packets are never abandoned, and the number of nodes goes to. Fnally the fxed pont equatons are used to obtan nsghts nto the throughput dfferentaton provded by dfferent ntal back-offs, persstence factors, and AIFS, for fnte number of nodes, and for dfferentaton parameter values smlar to those n the standard. Categores and Subect Descrptors C.2.5 [Computer-Communcaton Networks]: Local and Wde-Area Networks Access schemes; I.6.4 [Smulaton and Modelng]: Model Valdaton and Analyss General Terms Performance Permsson to make dgtal or hard copes of all or part of ths work for personal or classroom use s granted wthout fee provded that copes are not made or dstrbuted for proft or commercal advantage and that copes bear ths notce and the full ctaton on the frst page. To copy otherwse, to republsh, to post on servers or to redstrbute to lsts, requres pror specfc permsson and/or a fee. SIGMETRICS 5, June 6, 25, Banff, Alberta, Canada. Copyrght 25 ACM /5/6...$5.. Keywords Performance of Wreless LANs, Short term Unfarness, QoS n Wreless LANs, EDCF Analyss. INTRODUCTION A new component of the IEEE 82.e medum access control (MAC) s an enhanced DCF (EDCF), whch provdes dfferentated channel access to packets by allowng dfferent backoff parameters (see [2]). Several traffc classes are supported, the classes beng dstngushed by dfferent back-off parameters. Thus, whereas n the legacy DCF all nodes have a sngle back-off state machne all wth the same back-off parameters (we say that the nodes are homogeneous), n EDCF the nodes can have multple back-off state machnes wth dfferent parameters, and hence are permtted to be nonhomogeneous. Ths paper s concerned wth the saturaton throughput analyss of sngle cell networks wth nonhomogeneous nodes. We lmt our study to the case n whch each node has one EDCF queue; the further generalsaton to multple traffc classes per node can also be done usng the same technques but s not reported here for lack of space. Thus n the nonhomogeneous case our analyss s applcable to a sngle cell ad hoc network of IEEE 82.e nodes (sngle cell meanng that all nodes are wthn control channel range of each other), wth each node offerng traffc of a sngle class. We consder an deal channel (wthout capture, fadng or frame error) and assume that packets are lost only due to collson of smultaneous transmssons. Much work has been reported on the performance evaluaton of EDCF to support dfferentated servce. Most of the analytcal work reported has been based on a decouplng approxmaton proposed ntally by Banch ([3]). Whle keepng the basc decouplng approxmaton, n [] Kumar et al. presented a sgnfcant smplfcaton and generalsaton of the analyss of the IEEE 82. back-off mechansm. Ths analyss led to a certan one dmensonal fxed pont equaton for the collson probablty experenced by the nodes n a homogeneous system (.e., one n whch all the nodes have the same back-off parameters). In ths paper we consder multdmensonal fxed pont equatons for a homogeneous system of nodes, and also for a nonhomogeneous system of nodes. The nonhomogenety could arse due to dfferent ntal back-offs, or dfferent back-off multplers, or dfferent amounts of tme that nodes wat after a transmsson before restartng ther back-off counters (.e., the AIFS (Arbtra-

2 ton InterFrame Space) mechansm of IEEE 82.e). We consder balanced solutons of the resultng multdmensonal fxed pont equatons (.e., solutons n whch all the coordnates are equal), and also unbalanced fxed ponts. The man contrbutons of ths paper are the followng:. We provde examples of homogeneous systems n whch, even though a unque balanced fxed pont exsts, there can be multple unbalanced fxed ponts, thus suggestng multstablty. We demonstrate by smulaton that, n such cases, sgnfcant short term unfarness can be observed and the unque balanced fxed pont fals to capture the system performance. 2. Next, n the case where the backoff ncreases multplcatvely (as n IEEE 82.), we establsh a smple suffcent condton for the unqueness of the soluton of the multdmensonal fxed pont equaton n the homogeneous and the nonhomogeneous cases. 3. We perform an analytcal study of the dfferentaton provded by each of the three mechansms that we model. We then provde an asymptotc analyss of the servce dfferentaton (as the number of nodes become large), and also some approxmate results for a fnte number of nodes. A survey of the lterature: There has been much research actvty on modelng the performance of IEEE 82. and IEEE 82.e medum access standards. The general approach has been to extend the decouplng approxmaton ntroduced by Banch ([3]). Wthout modelng the AIFS mechansm, the extenson s straghtforward. Only the ntal back-off, and the back-off multpler (persstence factor) are modeled. In [4], [5] and [6], such a scheme s studed by extendng Banch s Markov model per traffc category. In ths paper, n Secton 3, we wll provde a generalsaton and smplfcaton of ths approach. We wll provde examples where nonunque fxed ponts can exst, the consequences of such nonunqueness, and also condtons that guarantee unqueness. AIFS technque s a further enhancement n IEEE 82.e that provdes a sort of prorty to nodes that have smaller values of AIFS. After any successful transmsson, whereas hgh prorty nodes (wth AIFS = DIFS) wat only for DIFS (DCF Interframe Space) to resume countng down ther backoff counters, low prorty nodes (wth AIFS > DIFS) defer the ntaton of countdown for an addtonal AIFS DIFS slots. Thus a hgh prorty node decrements ts back-off counter faster than a low prorty node and also has fewer collsons. Among the approaches that have been proposed for modelng the AIFS mechansm (for example, [7], [8], [9], [], [] and [2]) the ones n [] and [2] come much closer to capturng the servce dfferentaton provded by the AIFS feature. In [] the authors propose a Markov model to capture both the dfferent back-off wndow expanson approach and AIFS. AIFS s modelled by expandng the state-space of the Markov chan to nclude the number of slots elapsed snce the prevous transmsson attempt on the channel. In [2] the authors observe that the system exsts n states n whch only nodes of certan traffc classes can attempt. The approach s to model the evoluton of these states as a Markov chan. The transton probabltes of ths Markov chan are obtaned from the assumed, decoupled attempt probabltes. Ths approach yelds a fxed pont formulaton. Ths s the approach we wll dscuss n Secton 6. Relaton of the exstng lterature to our work: We note that the analyses n [] and [2] are based on Banch s approach to modelng the resdual back-off by a Markov chan. In ths paper, we have extended the smplfcaton reported n [] (whch was for a homogeneous system of nodes) to nonhomogeneous nodes wth dfferent backoff parameters and also AIFS based prorty schemes. Thus, n our work, we have provded a smplfed and ntegrated model to capture all the essental servce dfferentaton mechansms of IEEE 82.e. In the prevous lterature, t s assumed that the collson rate experenced by a node of each traffc category s constant over tme. There appears to have been no attempt to study the phenomenon of short term unfarness n the fxed pont framework. Also, all the exstng work assumes that the collson probabltes of all the nodes of a gven traffc category are the same. Thus there appears to have been no earler work on studyng the possblty of unbalanced solutons of the fxed pont equatons. In addton, the possblty of nonunqueness of the soluton of the fxed pont equatons arsng n the analyses seems to have been mssed n the earler lterature. In our work, we study the fxed pont equatons for IEEE 82.e networks and take nto account all these possbltes. Outlne of the paper: In Secton 2 we revew the generalsed back-off model that was frst presented n []. In Secton 3 we develop the multdmensonal fxed pont equaton for the homogeneous and nonhomogeneous cases, and obtan the necessary and suffcent condtons satsfed by the solutons to the fxed pont equatons. We provde examples n Secton 4 to show that even n the homogeneous case there can exst multple unbalanced fxed ponts and show the consequence of ths. In Secton 5., we analyse the fxed pont equatons for a homogeneous system of nodes and obtan a condton for the exstence of only one fxed pont. In Sectons 5.2 and 6, we extend the analyss to nonhomogeneous system of nodes, wth dfferent back-off parameters. In Secton 7 we provde an analytcal study of the servce dfferentaton provded by the varous mechansms. Secton 8 concludes the paper and dscusses future work. The proofs of all lemmas and theorems, f not n the paper, are provded n [8]. 2. THE GENERALISED BACK-OFF MODEL There are n nodes, ndexed by, n, each wth one EDCF queue. We adopt the notaton n [] whose authors consder a generalsaton of the back-off behavour of the nodes, and defne the followng back-off parameters (for node ) K := At the (K + )th attempt ether the packet beng attempted by node succeeds or s dscarded b,k := The mean back-off (n slots) at the kth attempt for a packet beng attempted by node, k K Defnton 2.. A system of n nodes s sad to be homogeneous, f the back-off parameters of the nodes K, b,k, k K are the same for all, n. A system of nodes s called nonhomogeneous f the back-off parameters of the nodes are not dentcal.

3 Remark: IEEE 82.e permts dfferent backoff parameters to dfferentate channel access obtaned by the nodes n an attempt to provde QoS. The above defntons capture the possblty of havng dfferent CW mn and CW max values, dfferent exponental back-off multpler values and even dfferent number of permtted attempts. For ease of dscusson and understandng, we wll postpone the topc of AIFS untl Secton 6. Hence n the dscussons up to Secton 5.2, all the nodes wat only for a DIFS after a busy channel. It has been shown n [] that under the decouplng assumpton, ntroduced by Banch n [3], the attempt probablty of node (condtoned on beng n back-off) for gven collson probablty γ s gven by, G (γ ) := + γ + + γ K b, + γ b, + + γ K b,k () Remark: When the system s homogeneous then we wll drop the subscrpt from G ( ), and wrte the functon smply as G( ). 3. THE FIXED POINT EQUATION It s mportant to note that n the present dscusson all rates are condtoned on beng n the back-off perods;.e., we have elmnated all duratons other than those n whch nodes are countng down ther back-off counters (see []). Ths suffces to obtan the collson probabltes. Now consder a nonhomogeneous system of n nodes. Let γ be the vector of collson probabltes of each node. Wth the slotted model for the back-off process and the decouplng assumpton, the natural mappng of the attempt probabltes of other nodes to the collson probablty of a node s gven by γ = Γ (β, β 2,..., β n) = n =, ( β ) where β = G (γ ). We can now expect that the equlbrum behavour of the system wll be charactersed by the solutons of the followng system of equatons. For n, γ = Γ (G (γ ),, G n(γ n)) We wrte these n equatons compactly n the form of the followng multdmensonal fxed pont equaton. γ = Γ(G(γ)) (2) Snce Γ(G(γ)) s a composton of contnuous functons t s contnuous. We thus have a contnuous mappng from [, ] n to [, ] n. Hence by Brouwer s fxed pont theorem there exsts a fxed pont n [, ] n for the equaton γ = Γ(G(γ)). Consder the th component of the fxed pont equaton,.e., γ = ( G (γ )) or equvalently, ( γ ) = n, n, ( G (γ )) Multplyng both sdes by ( G (γ )), we get, ( γ )( G (γ )) = ( G (γ )) n Thus a necessary and suffcent condton for a vector of collson probabltes γ = (γ,, γ n) to be a fxed pont soluton s that, for all n, n ( γ )( G (γ )) = ( G (γ )) (3) = where the rght-hand sde s seen to be ndependent of. Defne F (γ) := ( γ)( G (γ)). From Equaton 3 we see that f γ s a soluton of Equaton 2, then for all,,, n, F (γ ) = F (γ ) (4) Notce that ths s only a necessary condton. For example, n a homogeneous system of nodes, the vector γ such that γ = γ for all n, satsfes Equaton 4 for any γ, but not all such ponts are solutons of the fxed pont Equaton 2. Defnton 3.. We say that a fxed pont γ (.e., a soluton of γ = Γ(G(γ))) s a balanced fxed pont f γ = γ for all, n; otherwse, γ s sad to be an unbalanced fxed pont. Remarks 3... It s clear that f there exsts an unbalanced fxed pont for a homogeneous system, then every permutaton s also a fxed pont and hence, n such cases, we do not have a unque fxed pont. 2. In the homogeneous case, by symmetry, the average collson probablty at each node must be the same for every node. If the collson probabltes correspond to a fxed pont (see 3, next), then ths fxed pont wll be of the form (γ, γ,, γ) where γ solves γ = Γ(G(γ)) (snce Γ ( ) = Γ( ) and G ( ) = G( ) for all n). Such a fxed pont of γ = Γ(G(γ)) s guaranteed by Brouwer s Fxed Pont. The unqueness of such a balanced fxed pont was studed n []. We reproduce ths result n Theorem There s, however, the possblty that even n the homogeneous case, there s an unbalanced soluton of γ = Γ(G(γ)). By smulaton examples we observe n Secton 4 that when there exst unbalanced fxed ponts, the balanced fxed pont of the system does not characterse the average performance, even f there exsts only one balanced fxed pont. In Secton 5., we provde a condton for IEEE 82. type nodes wth geometrc backoff under whch there s a unque balanced fxed pont and no unbalanced fxed pont for a homogeneous system of nodes. In such cases, t s now well establshed, that the unque balanced fxed pont accurately predcts the saturaton throughput of the system. 4. For the homogeneous case the back-off process can be exactly modeled by a postve recurrent Markov chan (see []). Hence the attempt and collson processes wll be ergodc and, by symmetry, the nodes wll have equal attempt and collson probabltes. In such a stuaton the exstence of multple unbalanced fxed ponts wll suggest short term unfarness or short term multstablty. We wll observe ths phenomenon n Secton 4.

4 5. Consder a system of homogeneous nodes havng unbalanced solutons for the fxed pont equaton γ = Γ(G(γ)) (.e., there exsts, such that γ γ ), then from Equaton 4, we see that F (γ ) = F (γ ), or the functon F s many-to-one. Hence for a homogeneous system of nodes, f the functon F s one-to-one then there cannot exst unbalanced fxed ponts. In Secton 5.2 we use ths observaton to obtan a suffcent condton for the unqueness of the fxed pont n the nonhomogeneous case. 4. NONUNIQUE FIXED POINTS AND MUL- TISTABILIT: SIMULATION EXAMPLES 4. Example Consder a homogeneous system (let us call t System-I) wth n = nodes. The functon G( ) of the nodes s gven by, G(γ) = + γ + γ 2 + γ γ + γ 2 + γ (γ 4 + γ ) The system corresponds to the case where K =, b = b = b 2 = b 3 = and b 4 = b 5 = b 6 = = 64. From the form of functon G, we can see that a node whch s currently at backoff stage s more lkely to reman at that stage as t takes 4 successve collsons to make the attempt rate of the node <. Lkewse, a node that s n the larger back-off stages b 4 = b 5 = = 64, wll retry contnuously wth mean nter-attempt slots of 64 untl t succeeds. Observe that only one node can be at backoff stage at any tme. Ths leads to the apparent multstablty of the system. Fgure plots G(γ), the correspondng F (γ) = ( γ)( G(γ)) and shows the balanced fxed pont of the system for n = nodes. The balanced fxed pont of the system shown n the fgure s obtaned usng the fxed pont equaton γ = ( G(γ)) 9. Observe that the functon F ( ) s not one-to-one (the functon F ( ) not beng one-to-one does not necessarly mply that there exst multple fxed pont solutons; see Remarks 3., 5). Fgure 2 shows the exstence of unbalanced fxed ponts for System-I. These fxed ponts are obtaned as follows. Assume that we are nterested n fxed ponts such that γ γ 2 = = γ n. Gven γ 2 = = γ n, the attempt probablty of the nodes s gven by G(γ 2). Hence, the collson probablty of node s gven by γ = ( G(γ 2)) n. The attempt probablty of node would then be G(γ ). Usng the decouplng assumpton, the collson probablty of any of the n nodes would then be, ( G(γ 2)) n 2 ( G(γ )) = γ 2. Thus we obtan a fxed pont equaton for γ 2 (and hence for all the other γ, 3 n). In Fgure 2 we plot ( G(γ)) 8 ( G( ( G(γ)) 9 )) (plotted as lne marked wth dots), the ntersecton of whch wth the y=x lne shows the solutons for γ 2 = = γ n. In the same way, by elmnatng γ 2 from the multdmensonal system of equatons, we can obtan a fxed pont equaton for γ. Ths functon s also plotted n Fgure 2 (usng pluses and lnes) and the nterseton of ths curve wth the y=x lne shows the solutons for γ. We see that there are three solutons n each case. The smallest values of γ (approx..4) pars up wth the largest value of γ 2 = = γ n (approx..97). Notce that the balanced fxed pont of the system s also a fxed pont n the plot (compare wth Fgure ). Then there G(γ) ( γ)( G(γ)) ( G(γ)) 9 Balanced Fxed Pont collson probablty (γ) Fgure : Example System-I: The balanced fxed pont. Plots of G(γ), F (γ) = ( γ)( G(γ)) and ( G(γ)) 9 vs. the collson probablty γ; we also show the y=x lne. s one remanng unbalanced fxed pont whose values can be read off the plot. We note that there could exst many other unbalanced fxed ponts for ths system of equatons, as we have consdered only a partcular varety of fxed ponts that have the property that γ γ 2 = = γ n. In order to examne the consequences of multple unbalanced fxed ponts we smulated the back-off process wth the back-off parameters of System-I. The followng remarks summarse our smulaton approach n ths paper. Remarks 4. (On the Smulaton Approach used).. All the smulaton results reported n ths paper are based on smulatons of the coupled multdmensonal back-off processes of the varous nodes. We are not smulatng the actual Wreless LAN system (as s done n an ns2 smulaton). The man am of the smulatons s to understand the backoff behavour of the nodes wth respect to the dfferent backoff parameters. From the pont of vew of performance analyss, t may also be noted that once the back-off behavour s correctly modelled the channel actvty can easly be added analytcally, and thus throughput results can be obtaned (see [3] and []). Note that a good match between analyss that uses a decoupled Markov model of the back-off process and ns2 smulatons has already been reported n earler work (see the lterature survey n Secton ). 2. Thus our smulaton s programmed as follows. The system evolves over back-off slots. All the nodes are assumed to be n perfect slot synchronsaton. The actual coupled evoluton of the backoff process s modeled. The backoff dstrbuton s unform and the resdual backoff tme s the state for each node. At every slot, dependng on the state of the back-off process, there are three possbltes: the slot s dle, there s a successful transmsson, or there s a collson. Ths causes further evoluton of the back-off process.

5 collson probablty (γ ) γ 2... γ γ Balanced Fxed Pont short term average collson probablty Node Node 2 Avg collson prob collson probablty (γ ) Fgure 2: Example System-I: Demonstraton of unbalanced fxed ponts. Plots of ( G(γ)) 8 ( G( ( G(γ)) 9 )) (the curve drawn wth dots and lnes) and the functon for the unbalanced fxed pont equaton for γ (see text). P k= C () 3. In Fgures 3 and 5, for the purpose of reportng the short term unfarness results, the entre duraton of smulaton s dvded nto k frames, where the sze of each frame s, slots. The short-term average of the collson probablty of each node, n, s calculated as C () A where C () () and A () correspond to the number of collsons and attempts n frame, k, for node. The long-term aver- P age s smlarly calculated as n P n = where k= A () n s the number of nodes. Notce that the long-term average collson rate s a batch based average of the short-term collson rates. Hence, when lookng at the graphs, t wll be ncorrect to vsually average the short-term collson rate plots n an attempt to obtan the long-term average collson rate. Ths s because when a node s shown to have a low collson probablty, t s the one that s attemptng every slot (whle the other nodes attempt wth a mean gap of 64 slots), and hence t sees a low probablty of collson. In ths case A ( ) s large and C ( ) A ( ). On the other hand, when a node s shown to have a hgh collson probablty t s attemptng at an average rate of 64 and almost all ts attempts collde wth the node that s then attemptng n every slot. In ths case A ( ) s small and C ( ) A ( ). Thus, n obtanng the lnear average, t s essental to account for the large varaton n A ( ) between the two cases. In Fgure 3 we plot a (smulaton) snap shot of the short term average collson probablty of 2 of the nodes of System-I and the average collson probablty of the nodes (The average s calculated over all frames and all nodes. Snce the nodes are dentcal, the average collson probablty s the same for all the nodes). Observe that the short term average has a huge varance around the long term average. It s evdent that over s of slots one node or the other monopolses the channel (and the remanng nodes see a collson probablty of durng those slots). Ths could be tme n slots Fgure 3: Example System-I: Snap-shot of short term average collson probablty of 2 of the nodes. Also plotted s the average collson probablty of the nodes (averaged over all slots and nodes). The 95% confdence nterval for the average collson probablty les wthn.7% of the mean value. descrbed as short term multstablty. A look nto the farness ndex (see Fgure 6) plotted as a functon of the frame sze used to calculate throughput suggests that System-I exhbts sgnfcant unfarness n servce even over reasonably large tme ntervals. Implcaton for the use of the balanced fxed pont: Notce also that the average collson rate shown n Fgure 3 s about.25, whereas the balanced fxed pont shown n Fgure shows a collson probablty of about.62. Hence we see that n ths case, where there are multple fxed ponts, the balanced fxed pont does not capture the actual system performance. 4.2 Example 2 Let us now consder yet another homogeneous example (let us call t System-II) wth n = 2 nodes. The functon G( ) of the nodes s gven by, G(γ) = + γ + γ γ 7 + 3γ + 9γ γ γ 7 The system corresponds to the case where K = 7, b =, p = 3 and b k = p k b for all k K. We notce that n ths example the way the back-off expands s smlar to the way t expands n the IEEE 82. standard, except that the ntal back-off s very small ( slot), and the multpler s 3, rather than 2. We observe that, smlar to Example System-I, ths system also has multple (unbalanced) fxed ponts and exhbts short-term unfarness n servce (A detaled comment on System-II s provded n [8]). Dscusson of Examples and 2: From the smulaton examples, we can make the followng nferences. When there are multple unbalanced fxed ponts n a homogeneous system then the system can dsplay short term multstablty, whch manfests tself as sgnfcant short term unfarness n channel access. 2. When there are multple unbalanced fxed ponts n a homogeneous system then the collson probablty

6 Balanced Fxed Pont G(γ) ( γ)( G(γ)) ( G(γ)) 9 short term average collson probablty Node Node 2 Avg collson prob collson probablty (γ) Fgure 4: Example System-III: Plots of G(γ), F (γ) = ( γ)( G(γ)) and ( G(γ)) 9 vs. the collson probablty γ; the lne y=x s also shown. obtaned from the balanced fxed pont may be a poor approxmaton to the long term average collson probablty. 3. Smlar conclusons can be drawn for nonhomogeneous systems when the system of fxed pont equatons have multple solutons. It appears that the exstence of multple-fxed ponts s a consequence of the form of the G( ) functon n the above examples, where G( ) s smlar to a swtchng curve; see, for example, Fgure where there s a very hgh attempt probablty at low collson probabltes and a very low attempt probablty at hgh collson probabltes. 4.3 Example 3 Consder a homogeneous system n whch backoff ncreases multplcatvely as n IEEE 82. DCF (let us call t System- III), wth n = nodes. The functon G( ) s gven by, G(γ) = + γ + γ γ γ + 64γ γ 7 The system corresponds to the case where K = 7, p = 2 and b = 6 and b k = p k b for all k K. Fgure 4 plots G( ), the correspondng F (γ) = ( γ)( G(γ)) and the unque balanced fxed pont of the system (Notce that F s one-to-one and unqueness of the fxed pont wll be proved n Secton 5.) The balanced fxed pont of the system s obtaned usng the fxed pont equaton γ = ( G(γ)) 9. The balanced fxed pont yelds a collson probablty of approxmately.29. Fgure 5 plots a snap shot of the short term average collson probablty (from smulaton) of 2 of the nodes and the average collson probablty of the nodes of the Example System-III. Notce that the short term average collson rate s close to the average collson rate (the vertcal scale n ths fgure s much fner than n the correspondng fgures for System-I and System-II, see [8]). Also, the average collson rate matches well wth the balanced fxed pont soluton obtaned n Fgure 4. Remark: Thus we see that n a stuaton n whch there s a unque fxed pont not only s there lack of short term tme n slots Fgure 5: Example System-III: Snap-shot of short term average collson probablty of 2 of the nodes. Also plotted s the average collson probablty obtaned by the nodes. The 95% confdence nterval of the average collson rate les wthn.2% of the mean value. multstablty, but also the fxed pont soluton yelds a good approxmaton to the long run average behavour. 4.4 Short Term Farness n Examples, 2, 3 ( P n = τ ) 2 P n= τ 2 Fgure 6 plots the throughput farness ndex n (where τ s the average throughput of node over the measurement frame, see [5]) aganst the frame sze used to measure throughput. The farness ndex s obtaned for each frame and s averaged over the duraton of the smulaton. Also plotted n the fgure s the 95% confdence nterval. We note that values of ths ndex wll le n the nterval [, ], and smaller values of the ndex correspond to greater unfarness between the nodes. The performance of all the three example systems are compared. Notce that Example System-III (smlar to IEEE 82. DCF) has the best farness propertes. The system acheves farness of.9 over s of slots (or packets). However, for Example System-I and II, smlar performance s acheved only over, and,, slots (or packets). The unfarness of Example Systems-I and II can be attrbuted to ther apparent mult-stablty. In the subsequent sectons we establsh condtons for the unqueness of the solutons to the multdmensonal fxed pont equaton. 5. ANALSIS OF THE FIXED POINT 5. The Homogeneous Case The followng two results are adopted from []. Lemma 5.. G(γ) s nonncreasng n γ f b k, k, s a nondecreasng sequence. In that case, unless b k = b for all k, G(γ) s strctly decreasng n γ. Theorem 5.. For a homogeneous system of nodes, Γ(G(γ)) : [, ] [, ], has a unque fxed pont f b k, k, s a nondecreasng sequence. Remark: The fxed pont (γ, γ,, γ) s the unque balanced fxed pont for γ = Γ(G(γ)). From Equaton 4, we

7 Jans farness measure Example System I Example System II Example System III averagng nterval (n slots) Fgure 6: Jan s throughput farness ndex s plotted aganst the number of slots used to measure throughput. The dotted lnes mark the 95% confdence nterval. see that a necessary condton for the exstence of unbalanced fxed ponts n a homogeneous system of nodes s that the functon F (γ) = ( γ)( G(γ)) needs to be manyto-one. In other words, f the functon ( γ)( G(γ)) s one-to-one and f γ = (γ, γ 2,..., γ n) s a soluton of the system γ = Γ(G(γ)), then γ = γ for all,. Consder the multplcatvely ncreasng back-off model for whch G( ) s gven by, G(γ) = + γ + γ γ K b ( + pγ + p 2 γ p K γ K ) Clearly, G(γ) s a contnuously dfferentable functon and so s F (γ) = ( γ)( G(γ)). The followng smple lemma s a consequence of the mean value theorem. Lemma 5.2. F (γ) s one-to-one f F (γ) for all γ. Remarks 5.. When F ( ) s one-to-one, the followng hold () F (γ) = ff γ =, () F () >, snce G(γ) for all γ, and () F (γ) s a decreasng functon of γ. Now the dervatve of F s F (γ) = + G(γ) G (γ)( γ) Lemma 5.3. If K, p 2 and G( ) s as n Equaton 5, then G (γ) 2p b for all γ. Clearly, G(γ) b and G (γ) and ( γ) for all γ. Substtutng nto the expresson for F (γ), we get, F (γ) + + 2p b The followng result s then mmedate. Theorem 5.2. For a functon G( ) defned as n Equaton 5 f K, p 2 and b > 2p +, then the system γ = Γ(G(γ)) has a unque fxed pont whch s balanced. (5) Remark: It can be shown that f Lemma 5.3 holds for G( ) as n Equaton 5 t also holds for any case n whch b k = p k b for k m K and b k = p m b for m < k K. The latter s the stuaton n the IEEE 82. standard (wth b = 6, p = 2, K = 7, m = 5). Hence a homogeneous IEEE 82. WLAN has a unque fxed pont whch s also balanced. In general, f the functon G( ) s arbtrary (as n Equaton ) but monotone decreasng, there exsts a unque balanced fxed pont for the system as long as the functon ( γ)( G(γ)) s one-to-one. 5.2 The Nonhomogeneous Case In ths secton, we wll extend our results to systems wth nonhomogeneous nodes. AIFS wll be ntroduced n Secton 6. Nonhomogenety s ntroduced by varyng b, p and K of the nodes. Consder a nonhomogeneous system of n nodes, wth G ( ) a monotoncally decreasng functon and the functon ( γ)( G (γ)) beng one-to-one for all. Let there be two fxed pont solutons γ = (γ,, γ,2,..., γ,n) and γ 2 = (γ 2,, γ 2,2,..., γ 2,n) for the above system. From the necessary condton (Equaton 4) we requre that, for all, and for some J > and J 2 >, ( γ,)( G (γ,)) = J ( γ 2,)( G (γ 2,)) = J 2 Snce ( γ)( G (γ)) s one-to-one, we requre J J 2. Wthout loss of generalty, assume J < J 2. Hence, γ, > γ 2, for all. Usng Equaton 3 we have, γ 2, = ( G (γ 2,)) ( G (γ,)) = γ, a contradcton. Hence, we requre J = J 2 or there exsts a unque fxed pont. Notce that the arguments above mmedately mply the followng result. Theorem 5.3. If G (γ) s a decreasng functon of γ for all and ( γ)( G (γ)) s a monotone functon n [, ], then the system of equatons β = G (γ ) and γ = Γ (β,..., β,..., β n) has a unque fxed pont. Where nodes use exponentally ncreasng back-off, the next result then follows. Theorem 5.4. For a system of nodes wth G ( ) as n Equaton 5 that satsfy K, p 2 and b > 2p +, there a exsts a unque fxed pont for the system of equatons γ = Q ( G(γ)) for n. Remark: The above result has relevance n the context of the IEEE 82.e standard where the proposal s to use dfferences n back-off parameters to dfferentate the throughputs obtaned by the varous nodes. Whle Theorem 5.4 only states a suffcent condton, t does pont to a cauton n choosng the back-off parameters of the nodes. 6. ANALSIS OF THE AIFS MECHANISM Our approach for obtanng the fxed pont equatons s the same as the one developed n [2]. However, we develop

8 the analyss n the more general framework ntroduced n []. We show that under the condton that F ( ) s one-toone there exsts a unque fxed pont for ths problem as well. The analyss s presented here for the two prorty class case, but can be extended to any number of classes. Let us begn by recallng the basc dea of AIFS based servce dfferentaton (see [3]). In legacy DCF, a node decrements ts back-off counter and attempts to transmt only after t senses an dle medum for more than a DCF nterframe space (DIFS). However, n EDCF (Enhanced Dstrbuted Coordnaton Functon) based on the access category of a node (and ts AIFS value), a node attempts to transmt only after t senses the medum dle for more than ts AIFS. Hgher prorty nodes have smaller values of AIFS (though not less than DIFS), and hence obtan a lower average collson probablty, snce these nodes can decrement ther back-off counters, and even transmt, n slots n whch lower prorty nodes (watng to complete ther AIFSs) cannot. Thus, nodes of hgher prorty (lower AIFS) not only tend to transmt more often but also have fewer collsons compared to nodes of lower prorty (larger AIFS). 6. The Fxed Pont Equatons Let us consder two classes of nodes of two dfferent prortes. The prorty for a class s supported by usng AIFS as well as b, p and K. All the nodes of a partcular prorty have the same values for all these parameters. There are n () nodes of Class and n () nodes of Class. Class corresponds to a hgher prorty of servce. The AIFS for Class s DIFS, and for Class the AIFS s DIFS+l slots. Thus, after every transmsson actvty n the channel, only Class nodes attempt to transmt n the frst l slots followng an dle DIFS, whle Class nodes wat to complete ther AIFS. Also, f there s any transmsson actvty (by Class nodes) durng those l slots, then agan the Class nodes wat for another l slots followng an dle DIFS, and so on. As n [3] and [], we need to model only the evoluton of the back-off process of a node (.e., the back-off slots after removng any channel actvty such as transmssons or collsons) to obtan the collson probabltes. For convenence, let us call the slots n whch only Class nodes can attempt as Excess AIFS slots, whch wll correspond to the subscrpt EA n the notaton. In the remanng slots (correspondng to the subscrpt R n the notaton) nodes of ether class can attempt. Let us vew such groups of slots, where dfferent sets of nodes contend for the channel, as dfferent contenton perods. Let us defne β () := the attempt probablty of a Class node for all, n (), n the slots n whch a Class node can attempt (.e., all the slots) β () := the attempt probablty of a Class node for all, n (), n the contenton perods durng whch Class nodes can attempt (.e., slots that are not Excess AIFS slots) Note that n makng these defntons we are modelng the attempt probabltes for Class as beng constant over all slots,.e., the Excess AIFS slots and the remanng slots. Ths smplfcaton s ust an extenson of the basc decouplng approxmaton, and has been shown to yeld results that match well wth smulatons (see [2]). q EA q EA q EA q EA l q q EA EA q R Fgure 7: AIFS dfferentaton mechansm: Markov model for remanng number of AIFS slots. Now the collson probabltes experenced by nodes wll depend on the contenton perod that the system s n. The approach s to model the evoluton over contenton perods as a Markov Chan over the states (,, 2,, l), where the state s, s (l ), denotes that an amount of tme equal to DIFS +s slots has elapsed snce the end of the prevous channel actvty. These states correspond to the Excess AIFS perod n whch only Class nodes can attempt. In the remanng slots, where the state s s = l, all nodes can attempt. In order to obtan the transton probabltes for ths Markov chan we need the probablty that a slot s dle. Usng the decouplng assumpton, the dle probablty n any slot durng the Excess AIFS perod s obtaned as, q EA = n () = ( β () ) (6) Smlarly, the dle probablty n any of the remanng slots s obtaned as, q R = n () = n () ( β () ) ( β () ) (7) = The transton structure of the Markov chan s shown n Fgure 7. As compared to [2], we have used a smplfcaton that the maxmum contenton wndow s much larger than l. If ths were not the case then some nodes would certanly attempt before reachng l. In practce, l s small (e.g., slot or 5 slots; see [2]) compared to the maxmum contenton wndow. Let π(ea) be the statonary probablty of the system beng n the Excess AIFS perod;.e., ths s the probablty that the above Markov chan s n states, or, or, or (l ). In addton, let π(r) be the steady state probablty of the system beng n the any of the remanng slots,.e., state l of the Markov chan. Solvng the balance equatons for the steady state probabltes, we obtan, π(ea) = π(r) = + q EA + q 2 EA + + q l EA + q EA + q 2 EA + + ql EA + ql EA q R q l EA q R (8) + q EA + qea ql EA + ql EA q R Average collson probablty of a node s then obtaned by averagng the collson probablty experenced by a node over the dfferent contenton perods. Average collson prob- q R

9 ablty for Class nodes s gven by, for all, n (), γ () = π(ea)( + π(r)( ( n () =, n () =, ( β () )) ( β () n () ) = ( β () ))) (9) Smlarly, the average collson probablty of a Class node s gven by, for all, n (), n () = ( ( β () ) γ () = n () =, ( β () )) () Our analyss n the remanng secton now generalses the analyss of [2] and also establshes unqueness of the fxed pont and the property that the fxed pont s balanced over nodes n the same class. Defne G () ( ) and G () ( ) as n Equaton (except that the superscrpts here denote the class dependent back-off parameters, wth nodes wthn a class havng the same parameters). Then the average collson probablty obtaned from the prevous equatons can be used to obtan the attempt rates by usng the relatons β () = G () (γ () ), and β () = G () (γ () ) () for all n (), n (). We obtan fxed pont equatons for the collson probabltes by substtutng the attempt probabltes from Equaton nto Equatons 9 and (and also nto Equatons 6 and 7). We have a contnuous mappng from [, ] n() +n () to [, ] n() +n (). It follows from Brouwer s fxed pont theorem that there exsts a fxed pont. 6.2 Unqueness of the Fxed Pont Lemma 6.. If F ( ) s one-to-one, then collson probabltes of all the nodes of the same class are dentcal;.e., the fxed ponts are balanced wthn each class. Theorem 6.. The set of Equatons 9, and (together wth 8, 6 and 7), representng the fxed pont for the AIFS model, has a unque soluton f the correspondng functons F () ( ) and F () ( ) are one-to-one. Remark: It follows from the earler results n ths paper (see, for example, Theorem 5.2) that f G () ( ) and G () ( ) are of the form n Equaton 5, and f K (), p () 2, and b () 2p () +, for =,, then the fxed pont wll be unque. 7. THROUGHPUT DIFFERENTIATION: AN ANALTICAL STUD It should be noted that all the results n ths secton are for the fxed pont soluton. Hence, when we use the term collson probablty and attempt rate t s only n so far as a good match between the fxed pont analyss and smulaton has already been reported n earler lterature (see Secton ). We wll consder two alternatves for K, the maxmum retransmsson attempts allowed for a packet, namely K = and K fnte. In ths secton, for the fnte K case, the form of the functon G(γ), for all γ, γ s, G(γ) = + γ + γ γ K b ( + pγ + p 2 γ p K γ K ) (2) It s clear that for fnte K the attempt rate of a node s lower bounded, and hence as the number of nodes ncreases to nfnty the collson probablty of any node goes to. Hence, for ths case, we wll obtan nsghts regardng performance dfferentaton only for a fntely large number of nodes. For the nfnte K case, however, we wll study (as n []) the asymptotcs of performance dfferentaton as the number of nodes tends to. In the K = case, the functon G(γ) smplfes to, G (γ) = ( ( γp) b ( γ) γ < p γ p (3) In the nonhomogeneous case we wll wrte G () (γ) and G () (γ). For the homogeneous case wth K =, the (balanced fxed pont) asymptotc analyss as n was performed n []. Consder a set of nodes, dvded nto two classes, Class and Class, wth Class correspondng to a hgher prorty of servce. For smplcty, we assume that n () and n (), the number of nodes of Class and Class respectvely, are related as, n () = αn, n () = ( α)n for some n and α, < α <. Let γ () (K, n) and β () (K, n) be the fxed pont solutons for the collson probablty and attempt rate of a Class node for a gven K and total number of nodes n. Smlarly, let γ () (K, n) and β () (K, n) be the correspondng values for a Class node. We wll study three cases: Case : b () < b (), p() = p () = p, AIF S () = AIF S () = DIF S Case 2: b () = b () = b, p () < p (), AIF S () = AIF S () = DIF S Case 3: b () = b () = b, p () = p () = p, AIF S () < AIF S () Note that n the analyss n earler sectons, we used the Bnomal model for the number of attempts n a slot. Wth n, n ths secton, we wll use the Posson batch model for the number of attempts n a slot (as n []). 7. Case : Dfferentaton by b 7.. K =, Asymptotc Analyss as n Wth the random number of attempts of each class n a back-off slot beng modeled as Posson dstrbuted, the collson probabltes γ ( ) (, n) and the attempt rates β ( ) (, n) are related by γ () (, n) = e ((n() )β () (,n)+n () β () (,n)) γ () (, n) = e (n() β () (,n)+(n () )β () (,n)) (4) Substtutng β ( ) (, n) = G ( ) (γ ( ) (, n)) n the above equatons gves the desred fxed pont equatons governng the system. Trvally, we see that, ( γ () (, n))e β() (,n) = ( γ () (, n))e β() (,n) (5)

10 Lemma 7.. For {, }, F () (γ) := ( γ)e G() (γ) s one-to-one for all γ, γ f b 2p +. Theorem 7.. In Case, wth K =, when F () one-to-one for {, },. γ () (, n) < γ () (, n) for all n 2. lm n γ () (, n) p, lmn γ() (, n) p 3. lm n (n () β () (, n)+n () β () (, n)) ln( p p ) Theorem 7.2. In Case, wth K =, the rato of the throughputs of Class and Class 2 converges to b() n. s p as b () p Remark: Thus, for example, f b () = 6, b () = 32, and p = 2 then the rato of the Class to Class node throughput wll be approxmately 3/4 for large n Fnte K, Approxmate Analyss for Large n Wth fnte K, as the number of nodes ncreases, the collson probablty of ether class ncreases to (snce the attempt rate s lower bounded) and G ( ) s small (snce t decreases lke, see Equaton 2). Then the dfference b p K+ between the collson probabltes (we drop the arguments K and n n the followng) γ () γ () = (G () (γ () ) G () (γ () )) ( G () (γ () )) (n() ) ( G () (γ () )) (n() ) also becomes nsgnfcant. Hence, we can assume that γ () γ (). For equal packet length transmsson, the rato of the throughputs of a Class node to a Class node corresponds to the rato of ther success probabltes, hence the throughput rato s gven by, G () (γ () )( G () (γ () )) n() ( G () (γ () )) n() G () (γ () )( G () (γ () )) n() ( G () (γ () )) n() = G () (γ () ) ( G () (γ () )) G () (γ () ) ( G () (γ () )) (6) Usng γ () γ (), wrtng ths as γ, and usng the fact that G ( ) (γ) for large n, we have, G () (γ) ( G () (γ)) G () (γ) ( G () (γ)) G() (γ) G () (γ) = b() b () It follows that when servce dfferentaton s provded by the back-off wndow, for a large number of nodes, the throughput rato roughly corresponds to b() b (), whch, for large values of b () and b () s almost that same as that obtaned for the asymptotc analyss wth K = n Theorem 7.2 Remark: For fnte K case, ths observaton (throughput rato s approxmately equal to b() b () ) s well known. Ths result has been shown analytcally (usng smlar approxmatons) and also has been observed n smulatons (see [6], [] and [4]). It has been observed n [] that for a gven number of nodes, n, there wll exst a K(n) such that the system performance wll not vary much for all K > K(n). Hence, an asymptotc analyss would suffce for such cases. Moreover, we have obtaned ths result n a much more general settng, usng the functon G( ). 7.2 Case 2: Dfferentaton by p It may be noted that n the current verson of IEEE 82.e standard ths mechansm no longer exsts [2] K =, Asymptotc Analyss as n The fxed pont equaton governng the collson probablty and the attempt rate s the same as Equaton 4. The followng theorem summarzes the man results for Case 2. Theorem 7.3. In Case 2, wth K =, when F () one-to-one for {, }, the followng hold:. γ () (, n) < γ () (, n) for all n 2. lm n γ () (, n) p (), lm n γ () (, n) p () 3. lm n n () β () (, n) ln( p() p () ) 4. lm n n () β () (, n) = Remark: Thus we see that, wth K = and a large number of nodes, unlke ntal back-off based dfferentaton, the persstence factor based dfferentaton completely suppresses the class wth the larger value of p Fnte K, Approxmate Analyss for Large n For fnte K, wth the approxmaton γ () γ () and the fact that G ( ) (γ ( ) ), the throughput rato approxmates to (+p() γ+p () 2 γ 2 + +p ()K γ K ) (see Equaton 6). Hence, as (+p () γ+p () 2 γ 2 + +p ()K γ K ) the collson probablty of the system ncreases wth load, the rato of the throughputs of Class to Class also ncreases (dependng on p (), p () and the value of K). We note that as n, the throughput rato for the fnte K case s fnte, unlke the asymptotc case (K = ). However, the rato tends to nfnty when we consder K. 7.3 Case 3: Dfferentaton by AIFS 7.3. K =, Asymptotc Analyss for n In ths case servce dfferentaton s provded only by AIFS and we let G () = G () = G (.e., the back-off parameters b and p are the same). Wth the assumpton that the number of attempts n each slot s Posson dstrbuted, the fxed pont equatons for the AIFS model are (see Equatons 9 and ) γ () (, n) = π(ea)( e (n() )β () (,n) ) + π(r)( e (n() )β () (,n) n () β () (,n) ) γ () (, n) = ( e n() β () (,n) (n () )β () (,n) ) Theorem 7.4. In Case 3, wth K =, when F () one-to-one for {, },. γ () (, n) < γ () (, n) for all n 2. lm n γ () (, n) p, lmn γ() (, n) p s s

11 3. lm n n () β () (, n) ln( p p ) 4. lm n n () β () (, n) = Remark: Agan we see that usng AIFS for dfferentaton, when K = and large n, completely suppresses the class wth the larger value of AIFS. Observe that Parts 3 and 4 of Theorem 7.4 mply that the ndvdual node attempt rato β () (,n) goes to as n. Some nsght nto ths result β () (,n) wll be obtaned from the analyss n the followng secton Fnte K, Approxmate Analyss rato of throughput AIFS = /, b = 6/6, n () = n () AIFS = /, b = 6/32, n () = n () AIFS = /, b =6/6, n () = 5 AIFS = /, b = 6/32, n () = n () Lemma 7.2. In Case 3 for fnte K, wth l =, f the fxed pont collson probabltes are γ () and γ (), then the rato of the throughputs of Class to Class s gven by G () (γ () ) ( G () (γ () )) G () (γ () ) ( G () (γ () )) Usng ths result and approxmatng ( G () (γ () )) as before, the rato of throughput equals G () (γ () ) ( G () (γ () )) G () (γ () ) ( G () (γ () )) q R G() (γ () ) (7) q R G () (γ () ) q R For general l, we can expect a factor lke n the prevous q R l expresson. For low loads, when q R s not close to, the domnatng term n the prevous expresson s G() (γ () ). At G () (γ () ) hgh loads, both the terms contrbute to throughput dfferentaton dependng on the values of n () and n (). 7.4 Numercal Study and Dscusson In Fgure 8 we plot throughput ratos obtaned from a smulaton of the coupled back-off processes of two classes of nodes (the smulaton approach s explaned n Remarks 4.). We note that ths s the throughput rato f the packet szes of the two classes are equal. If the packet szes are unequal then we only need to multply the throughput rato plotted here by the rato of the packet lengths of the two classes. The followng remarks help n nterpretng the results n Fgure 8. Remarks 7... For fnte K the attempt rates are bounded below, and the term G() (γ () ) s bounded, but as G () (γ () ) (n() + n () ) the dle probablty q R ensurng (see Equaton 7) that the ndvdual node throughput rato goes to for fnte K as well (smlar to the asymptotc results n Theorem 7.4). In addton, when n () ncreases, π(ea) ncreases to. Hence, the lower prorty nodes (wth larger AIFS) rarely get a chance to attempt and the throughput rato goes to nfnty; ths s demonstrated by the smulaton results n Fgure 8, plots wth + and. When n () s kept constant and n () s ncreased (whch s more typcal), the collson probablty of Class nodes ncreases to and ther success probablty tends to. However, the collson probablty of Class nodes remans much less than n () + n () Fgure 8: Rato of the throughput of a Class (hgher prorty) node to the throughput of a Class node (lower prorty). Analyss results (sold lnes) and smulaton results (symbols). Four cases are consdered: +: dfferentaton only by AIFS wth equal number of nodes, n () = n () ; : dfferentaton by AIFS and by b wth equal number of nodes, n () = n () ; : dfferentaton only by b wth equal number of nodes, n () = n () ; : dfferentaton only by AIFS wth, 5 = n () n (). In all cases p = 2 and K = 7 for ether class. For the smulaton results, the 95% confdence nterval les wthn % of the average value. dependng on the value of n () and hence agan the throughput rato tends to (see Fgure 8, plots wth ). Fgure 8 also shows the throughput rato when only b s used for dfferentaton (plots wth ); notce that, as shown earler, the throughput rato s ust the recprocal of the ratos of the ntal back-off duratons, and does not change wth n. 2. For Case 3, n general, γ () and γ () are dfferent, unlke n Cases and 2. Ths s captured by the frst term n the expresson G() (γ () ) G () (γ () ) q R. 3. Notce that the above results for AIFS hold even when the functons G () and G () are not dentcal (see Fgure 8, plot wth ). A comparson between the plots wth + and n Fgure 8 shows the effect of usng both b and AIFS for throughput dfferentaton. The b based dfferentaton causes the entre curve to shft up (n favour of the hgher prorty class), and AIFS stll causes the rato to ncrease wth ncreasng n. 8. SUMMAR In ths paper we have studed a multdmensonal fxed pont equaton arsng from a model of the back-off process of the EDCF access mechansm n IEEE 82. and e Wreless LANs. Our frst concern was the consequences of the nonunqueness of the fxed pont soluton and condtons for unqueness. We demonstrated va examples of homogeneous systems that even when the balanced fxed pont s unque, the exstence of unbalanced fxed ponts coexsts wth the observaton of severe short term unfarness n smulatons.

More metrics on cartesian products

More metrics on cartesian products More metrcs on cartesan products If (X, d ) are metrc spaces for 1 n, then n Secton II4 of the lecture notes we defned three metrcs on X whose underlyng topologes are the product topology The purpose of

More information

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur Module 3 LOSSY IMAGE COMPRESSION SYSTEMS Verson ECE IIT, Kharagpur Lesson 6 Theory of Quantzaton Verson ECE IIT, Kharagpur Instructonal Objectves At the end of ths lesson, the students should be able to:

More information

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X Statstcs 1: Probablty Theory II 37 3 EPECTATION OF SEVERAL RANDOM VARIABLES As n Probablty Theory I, the nterest n most stuatons les not on the actual dstrbuton of a random vector, but rather on a number

More information

Chapter 13: Multiple Regression

Chapter 13: Multiple Regression Chapter 13: Multple Regresson 13.1 Developng the multple-regresson Model The general model can be descrbed as: It smplfes for two ndependent varables: The sample ft parameter b 0, b 1, and b are used to

More information

Analysis of Discrete Time Queues (Section 4.6)

Analysis of Discrete Time Queues (Section 4.6) Analyss of Dscrete Tme Queues (Secton 4.6) Copyrght 2002, Sanjay K. Bose Tme axs dvded nto slots slot slot boundares Arrvals can only occur at slot boundares Servce to a job can only start at a slot boundary

More information

NUMERICAL DIFFERENTIATION

NUMERICAL DIFFERENTIATION NUMERICAL DIFFERENTIATION 1 Introducton Dfferentaton s a method to compute the rate at whch a dependent output y changes wth respect to the change n the ndependent nput x. Ths rate of change s called the

More information

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE Analytcal soluton s usually not possble when exctaton vares arbtrarly wth tme or f the system s nonlnear. Such problems can be solved by numercal tmesteppng

More information

Module 9. Lecture 6. Duality in Assignment Problems

Module 9. Lecture 6. Duality in Assignment Problems Module 9 1 Lecture 6 Dualty n Assgnment Problems In ths lecture we attempt to answer few other mportant questons posed n earler lecture for (AP) and see how some of them can be explaned through the concept

More information

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems Numercal Analyss by Dr. Anta Pal Assstant Professor Department of Mathematcs Natonal Insttute of Technology Durgapur Durgapur-713209 emal: anta.bue@gmal.com 1 . Chapter 5 Soluton of System of Lnear Equatons

More information

Difference Equations

Difference Equations Dfference Equatons c Jan Vrbk 1 Bascs Suppose a sequence of numbers, say a 0,a 1,a,a 3,... s defned by a certan general relatonshp between, say, three consecutve values of the sequence, e.g. a + +3a +1

More information

Temperature. Chapter Heat Engine

Temperature. Chapter Heat Engine Chapter 3 Temperature In prevous chapters of these notes we ntroduced the Prncple of Maxmum ntropy as a technque for estmatng probablty dstrbutons consstent wth constrants. In Chapter 9 we dscussed the

More information

Structure and Drive Paul A. Jensen Copyright July 20, 2003

Structure and Drive Paul A. Jensen Copyright July 20, 2003 Structure and Drve Paul A. Jensen Copyrght July 20, 2003 A system s made up of several operatons wth flow passng between them. The structure of the system descrbes the flow paths from nputs to outputs.

More information

Lecture 4: November 17, Part 1 Single Buffer Management

Lecture 4: November 17, Part 1 Single Buffer Management Lecturer: Ad Rosén Algorthms for the anagement of Networs Fall 2003-2004 Lecture 4: November 7, 2003 Scrbe: Guy Grebla Part Sngle Buffer anagement In the prevous lecture we taled about the Combned Input

More information

Lecture 12: Discrete Laplacian

Lecture 12: Discrete Laplacian Lecture 12: Dscrete Laplacan Scrbe: Tanye Lu Our goal s to come up wth a dscrete verson of Laplacan operator for trangulated surfaces, so that we can use t n practce to solve related problems We are mostly

More information

Numerical Heat and Mass Transfer

Numerical Heat and Mass Transfer Master degree n Mechancal Engneerng Numercal Heat and Mass Transfer 06-Fnte-Dfference Method (One-dmensonal, steady state heat conducton) Fausto Arpno f.arpno@uncas.t Introducton Why we use models and

More information

The optimal delay of the second test is therefore approximately 210 hours earlier than =2.

The optimal delay of the second test is therefore approximately 210 hours earlier than =2. THE IEC 61508 FORMULAS 223 The optmal delay of the second test s therefore approxmately 210 hours earler than =2. 8.4 The IEC 61508 Formulas IEC 61508-6 provdes approxmaton formulas for the PF for smple

More information

Linear Regression Analysis: Terminology and Notation

Linear Regression Analysis: Terminology and Notation ECON 35* -- Secton : Basc Concepts of Regresson Analyss (Page ) Lnear Regresson Analyss: Termnology and Notaton Consder the generc verson of the smple (two-varable) lnear regresson model. It s represented

More information

APPENDIX A Some Linear Algebra

APPENDIX A Some Linear Algebra APPENDIX A Some Lnear Algebra The collecton of m, n matrces A.1 Matrces a 1,1,..., a 1,n A = a m,1,..., a m,n wth real elements a,j s denoted by R m,n. If n = 1 then A s called a column vector. Smlarly,

More information

Convergence of random processes

Convergence of random processes DS-GA 12 Lecture notes 6 Fall 216 Convergence of random processes 1 Introducton In these notes we study convergence of dscrete random processes. Ths allows to characterze phenomena such as the law of large

More information

2E Pattern Recognition Solutions to Introduction to Pattern Recognition, Chapter 2: Bayesian pattern classification

2E Pattern Recognition Solutions to Introduction to Pattern Recognition, Chapter 2: Bayesian pattern classification E395 - Pattern Recognton Solutons to Introducton to Pattern Recognton, Chapter : Bayesan pattern classfcaton Preface Ths document s a soluton manual for selected exercses from Introducton to Pattern Recognton

More information

Psychology 282 Lecture #24 Outline Regression Diagnostics: Outliers

Psychology 282 Lecture #24 Outline Regression Diagnostics: Outliers Psychology 282 Lecture #24 Outlne Regresson Dagnostcs: Outlers In an earler lecture we studed the statstcal assumptons underlyng the regresson model, ncludng the followng ponts: Formal statement of assumptons.

More information

Linear Approximation with Regularization and Moving Least Squares

Linear Approximation with Regularization and Moving Least Squares Lnear Approxmaton wth Regularzaton and Movng Least Squares Igor Grešovn May 007 Revson 4.6 (Revson : March 004). 5 4 3 0.5 3 3.5 4 Contents: Lnear Fttng...4. Weghted Least Squares n Functon Approxmaton...

More information

Econ107 Applied Econometrics Topic 3: Classical Model (Studenmund, Chapter 4)

Econ107 Applied Econometrics Topic 3: Classical Model (Studenmund, Chapter 4) I. Classcal Assumptons Econ7 Appled Econometrcs Topc 3: Classcal Model (Studenmund, Chapter 4) We have defned OLS and studed some algebrac propertes of OLS. In ths topc we wll study statstcal propertes

More information

COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS

COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS Avalable onlne at http://sck.org J. Math. Comput. Sc. 3 (3), No., 6-3 ISSN: 97-537 COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS

More information

Exercise Solutions to Real Analysis

Exercise Solutions to Real Analysis xercse Solutons to Real Analyss Note: References refer to H. L. Royden, Real Analyss xersze 1. Gven any set A any ɛ > 0, there s an open set O such that A O m O m A + ɛ. Soluton 1. If m A =, then there

More information

Notes on Frequency Estimation in Data Streams

Notes on Frequency Estimation in Data Streams Notes on Frequency Estmaton n Data Streams In (one of) the data streamng model(s), the data s a sequence of arrvals a 1, a 2,..., a m of the form a j = (, v) where s the dentty of the tem and belongs to

More information

/ n ) are compared. The logic is: if the two

/ n ) are compared. The logic is: if the two STAT C141, Sprng 2005 Lecture 13 Two sample tests One sample tests: examples of goodness of ft tests, where we are testng whether our data supports predctons. Two sample tests: called as tests of ndependence

More information

TCOM 501: Networking Theory & Fundamentals. Lecture 7 February 25, 2003 Prof. Yannis A. Korilis

TCOM 501: Networking Theory & Fundamentals. Lecture 7 February 25, 2003 Prof. Yannis A. Korilis TCOM 501: Networkng Theory & Fundamentals Lecture 7 February 25, 2003 Prof. Yanns A. Korls 1 7-2 Topcs Open Jackson Networks Network Flows State-Dependent Servce Rates Networks of Transmsson Lnes Klenrock

More information

The Second Anti-Mathima on Game Theory

The Second Anti-Mathima on Game Theory The Second Ant-Mathma on Game Theory Ath. Kehagas December 1 2006 1 Introducton In ths note we wll examne the noton of game equlbrum for three types of games 1. 2-player 2-acton zero-sum games 2. 2-player

More information

The Feynman path integral

The Feynman path integral The Feynman path ntegral Aprl 3, 205 Hesenberg and Schrödnger pctures The Schrödnger wave functon places the tme dependence of a physcal system n the state, ψ, t, where the state s a vector n Hlbert space

More information

Lecture Note 3. Eshelby s Inclusion II

Lecture Note 3. Eshelby s Inclusion II ME340B Elastcty of Mcroscopc Structures Stanford Unversty Wnter 004 Lecture Note 3. Eshelby s Incluson II Chrs Wenberger and We Ca c All rghts reserved January 6, 004 Contents 1 Incluson energy n an nfnte

More information

The Geometry of Logit and Probit

The Geometry of Logit and Probit The Geometry of Logt and Probt Ths short note s meant as a supplement to Chapters and 3 of Spatal Models of Parlamentary Votng and the notaton and reference to fgures n the text below s to those two chapters.

More information

n α j x j = 0 j=1 has a nontrivial solution. Here A is the n k matrix whose jth column is the vector for all t j=0

n α j x j = 0 j=1 has a nontrivial solution. Here A is the n k matrix whose jth column is the vector for all t j=0 MODULE 2 Topcs: Lnear ndependence, bass and dmenson We have seen that f n a set of vectors one vector s a lnear combnaton of the remanng vectors n the set then the span of the set s unchanged f that vector

More information

Graph Reconstruction by Permutations

Graph Reconstruction by Permutations Graph Reconstructon by Permutatons Perre Ille and Wllam Kocay* Insttut de Mathémathques de Lumny CNRS UMR 6206 163 avenue de Lumny, Case 907 13288 Marselle Cedex 9, France e-mal: lle@ml.unv-mrs.fr Computer

More information

College of Computer & Information Science Fall 2009 Northeastern University 20 October 2009

College of Computer & Information Science Fall 2009 Northeastern University 20 October 2009 College of Computer & Informaton Scence Fall 2009 Northeastern Unversty 20 October 2009 CS7880: Algorthmc Power Tools Scrbe: Jan Wen and Laura Poplawsk Lecture Outlne: Prmal-dual schema Network Desgn:

More information

ANSWERS. Problem 1. and the moment generating function (mgf) by. defined for any real t. Use this to show that E( U) var( U)

ANSWERS. Problem 1. and the moment generating function (mgf) by. defined for any real t. Use this to show that E( U) var( U) Econ 413 Exam 13 H ANSWERS Settet er nndelt 9 deloppgaver, A,B,C, som alle anbefales å telle lkt for å gøre det ltt lettere å stå. Svar er gtt . Unfortunately, there s a prntng error n the hnt of

More information

Distributed Optimal TXOP Control for Throughput Requirements in IEEE e Wireless LAN

Distributed Optimal TXOP Control for Throughput Requirements in IEEE e Wireless LAN Dstrbuted Optmal TXOP Control for Throughput Requrements n IEEE 80.e Wreless LAN Ju Yong Lee, Ho Young Hwang, Jtae Shn, and Shahrokh Valaee KAIST Insttute for Informaton Technology Convergence, KAIST,

More information

One-sided finite-difference approximations suitable for use with Richardson extrapolation

One-sided finite-difference approximations suitable for use with Richardson extrapolation Journal of Computatonal Physcs 219 (2006) 13 20 Short note One-sded fnte-dfference approxmatons sutable for use wth Rchardson extrapolaton Kumar Rahul, S.N. Bhattacharyya * Department of Mechancal Engneerng,

More information

Lectures - Week 4 Matrix norms, Conditioning, Vector Spaces, Linear Independence, Spanning sets and Basis, Null space and Range of a Matrix

Lectures - Week 4 Matrix norms, Conditioning, Vector Spaces, Linear Independence, Spanning sets and Basis, Null space and Range of a Matrix Lectures - Week 4 Matrx norms, Condtonng, Vector Spaces, Lnear Independence, Spannng sets and Bass, Null space and Range of a Matrx Matrx Norms Now we turn to assocatng a number to each matrx. We could

More information

Resource Allocation with a Budget Constraint for Computing Independent Tasks in the Cloud

Resource Allocation with a Budget Constraint for Computing Independent Tasks in the Cloud Resource Allocaton wth a Budget Constrant for Computng Independent Tasks n the Cloud Wemng Sh and Bo Hong School of Electrcal and Computer Engneerng Georga Insttute of Technology, USA 2nd IEEE Internatonal

More information

ECE559VV Project Report

ECE559VV Project Report ECE559VV Project Report (Supplementary Notes Loc Xuan Bu I. MAX SUM-RATE SCHEDULING: THE UPLINK CASE We have seen (n the presentaton that, for downlnk (broadcast channels, the strategy maxmzng the sum-rate

More information

The Multiple Classical Linear Regression Model (CLRM): Specification and Assumptions. 1. Introduction

The Multiple Classical Linear Regression Model (CLRM): Specification and Assumptions. 1. Introduction ECONOMICS 5* -- NOTE (Summary) ECON 5* -- NOTE The Multple Classcal Lnear Regresson Model (CLRM): Specfcaton and Assumptons. Introducton CLRM stands for the Classcal Lnear Regresson Model. The CLRM s also

More information

Week3, Chapter 4. Position and Displacement. Motion in Two Dimensions. Instantaneous Velocity. Average Velocity

Week3, Chapter 4. Position and Displacement. Motion in Two Dimensions. Instantaneous Velocity. Average Velocity Week3, Chapter 4 Moton n Two Dmensons Lecture Quz A partcle confned to moton along the x axs moves wth constant acceleraton from x =.0 m to x = 8.0 m durng a 1-s tme nterval. The velocty of the partcle

More information

AN EXTENDED CLASS OF TIME-CONTINUOUS BRANCHING PROCESSES. Rong-Rong Chen. ( University of Illinois at Urbana-Champaign)

AN EXTENDED CLASS OF TIME-CONTINUOUS BRANCHING PROCESSES. Rong-Rong Chen. ( University of Illinois at Urbana-Champaign) AN EXTENDED CLASS OF TIME-CONTINUOUS BRANCHING PROCESSES Rong-Rong Chen ( Unversty of Illnos at Urbana-Champagn Abstract. Ths paper s devoted to studyng an extended class of tme-contnuous branchng processes,

More information

Equilibrium Analysis of the M/G/1 Queue

Equilibrium Analysis of the M/G/1 Queue Eulbrum nalyss of the M/G/ Queue Copyrght, Sanay K. ose. Mean nalyss usng Resdual Lfe rguments Secton 3.. nalyss usng an Imbedded Marov Chan pproach Secton 3. 3. Method of Supplementary Varables done later!

More information

The Order Relation and Trace Inequalities for. Hermitian Operators

The Order Relation and Trace Inequalities for. Hermitian Operators Internatonal Mathematcal Forum, Vol 3, 08, no, 507-57 HIKARI Ltd, wwwm-hkarcom https://doorg/0988/mf088055 The Order Relaton and Trace Inequaltes for Hermtan Operators Y Huang School of Informaton Scence

More information

Games of Threats. Elon Kohlberg Abraham Neyman. Working Paper

Games of Threats. Elon Kohlberg Abraham Neyman. Working Paper Games of Threats Elon Kohlberg Abraham Neyman Workng Paper 18-023 Games of Threats Elon Kohlberg Harvard Busness School Abraham Neyman The Hebrew Unversty of Jerusalem Workng Paper 18-023 Copyrght 2017

More information

Lecture 21: Numerical methods for pricing American type derivatives

Lecture 21: Numerical methods for pricing American type derivatives Lecture 21: Numercal methods for prcng Amercan type dervatves Xaoguang Wang STAT 598W Aprl 10th, 2014 (STAT 598W) Lecture 21 1 / 26 Outlne 1 Fnte Dfference Method Explct Method Penalty Method (STAT 598W)

More information

Module 2. Random Processes. Version 2 ECE IIT, Kharagpur

Module 2. Random Processes. Version 2 ECE IIT, Kharagpur Module Random Processes Lesson 6 Functons of Random Varables After readng ths lesson, ou wll learn about cdf of functon of a random varable. Formula for determnng the pdf of a random varable. Let, X be

More information

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity LINEAR REGRESSION ANALYSIS MODULE IX Lecture - 30 Multcollnearty Dr. Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur 2 Remedes for multcollnearty Varous technques have

More information

CSci 6974 and ECSE 6966 Math. Tech. for Vision, Graphics and Robotics Lecture 21, April 17, 2006 Estimating A Plane Homography

CSci 6974 and ECSE 6966 Math. Tech. for Vision, Graphics and Robotics Lecture 21, April 17, 2006 Estimating A Plane Homography CSc 6974 and ECSE 6966 Math. Tech. for Vson, Graphcs and Robotcs Lecture 21, Aprl 17, 2006 Estmatng A Plane Homography Overvew We contnue wth a dscusson of the major ssues, usng estmaton of plane projectve

More information

Section 8.3 Polar Form of Complex Numbers

Section 8.3 Polar Form of Complex Numbers 80 Chapter 8 Secton 8 Polar Form of Complex Numbers From prevous classes, you may have encountered magnary numbers the square roots of negatve numbers and, more generally, complex numbers whch are the

More information

Supplementary Notes for Chapter 9 Mixture Thermodynamics

Supplementary Notes for Chapter 9 Mixture Thermodynamics Supplementary Notes for Chapter 9 Mxture Thermodynamcs Key ponts Nne major topcs of Chapter 9 are revewed below: 1. Notaton and operatonal equatons for mxtures 2. PVTN EOSs for mxtures 3. General effects

More information

Dirichlet s Theorem In Arithmetic Progressions

Dirichlet s Theorem In Arithmetic Progressions Drchlet s Theorem In Arthmetc Progressons Parsa Kavkan Hang Wang The Unversty of Adelade February 26, 205 Abstract The am of ths paper s to ntroduce and prove Drchlet s theorem n arthmetc progressons,

More information

9 Derivation of Rate Equations from Single-Cell Conductance (Hodgkin-Huxley-like) Equations

9 Derivation of Rate Equations from Single-Cell Conductance (Hodgkin-Huxley-like) Equations Physcs 171/271 - Chapter 9R -Davd Klenfeld - Fall 2005 9 Dervaton of Rate Equatons from Sngle-Cell Conductance (Hodgkn-Huxley-lke) Equatons We consder a network of many neurons, each of whch obeys a set

More information

Problem Set 9 Solutions

Problem Set 9 Solutions Desgn and Analyss of Algorthms May 4, 2015 Massachusetts Insttute of Technology 6.046J/18.410J Profs. Erk Demane, Srn Devadas, and Nancy Lynch Problem Set 9 Solutons Problem Set 9 Solutons Ths problem

More information

Markov chains. Definition of a CTMC: [2, page 381] is a continuous time, discrete value random process such that for an infinitesimal

Markov chains. Definition of a CTMC: [2, page 381] is a continuous time, discrete value random process such that for an infinitesimal Markov chans M. Veeraraghavan; March 17, 2004 [Tp: Study the MC, QT, and Lttle s law lectures together: CTMC (MC lecture), M/M/1 queue (QT lecture), Lttle s law lecture (when dervng the mean response tme

More information

8.592J: Solutions for Assignment 7 Spring 2005

8.592J: Solutions for Assignment 7 Spring 2005 8.59J: Solutons for Assgnment 7 Sprng 5 Problem 1 (a) A flament of length l can be created by addton of a monomer to one of length l 1 (at rate a) or removal of a monomer from a flament of length l + 1

More information

THE CHINESE REMAINDER THEOREM. We should thank the Chinese for their wonderful remainder theorem. Glenn Stevens

THE CHINESE REMAINDER THEOREM. We should thank the Chinese for their wonderful remainder theorem. Glenn Stevens THE CHINESE REMAINDER THEOREM KEITH CONRAD We should thank the Chnese for ther wonderful remander theorem. Glenn Stevens 1. Introducton The Chnese remander theorem says we can unquely solve any par of

More information

Economics 101. Lecture 4 - Equilibrium and Efficiency

Economics 101. Lecture 4 - Equilibrium and Efficiency Economcs 0 Lecture 4 - Equlbrum and Effcency Intro As dscussed n the prevous lecture, we wll now move from an envronment where we looed at consumers mang decsons n solaton to analyzng economes full of

More information

Global Sensitivity. Tuesday 20 th February, 2018

Global Sensitivity. Tuesday 20 th February, 2018 Global Senstvty Tuesday 2 th February, 28 ) Local Senstvty Most senstvty analyses [] are based on local estmates of senstvty, typcally by expandng the response n a Taylor seres about some specfc values

More information

Comparison of Regression Lines

Comparison of Regression Lines STATGRAPHICS Rev. 9/13/2013 Comparson of Regresson Lnes Summary... 1 Data Input... 3 Analyss Summary... 4 Plot of Ftted Model... 6 Condtonal Sums of Squares... 6 Analyss Optons... 7 Forecasts... 8 Confdence

More information

Assortment Optimization under MNL

Assortment Optimization under MNL Assortment Optmzaton under MNL Haotan Song Aprl 30, 2017 1 Introducton The assortment optmzaton problem ams to fnd the revenue-maxmzng assortment of products to offer when the prces of products are fxed.

More information

Negative Binomial Regression

Negative Binomial Regression STATGRAPHICS Rev. 9/16/2013 Negatve Bnomal Regresson Summary... 1 Data Input... 3 Statstcal Model... 3 Analyss Summary... 4 Analyss Optons... 7 Plot of Ftted Model... 8 Observed Versus Predcted... 10 Predctons...

More information

Transfer Functions. Convenient representation of a linear, dynamic model. A transfer function (TF) relates one input and one output: ( ) system

Transfer Functions. Convenient representation of a linear, dynamic model. A transfer function (TF) relates one input and one output: ( ) system Transfer Functons Convenent representaton of a lnear, dynamc model. A transfer functon (TF) relates one nput and one output: x t X s y t system Y s The followng termnology s used: x y nput output forcng

More information

COS 521: Advanced Algorithms Game Theory and Linear Programming

COS 521: Advanced Algorithms Game Theory and Linear Programming COS 521: Advanced Algorthms Game Theory and Lnear Programmng Moses Charkar February 27, 2013 In these notes, we ntroduce some basc concepts n game theory and lnear programmng (LP). We show a connecton

More information

Lecture 4 Hypothesis Testing

Lecture 4 Hypothesis Testing Lecture 4 Hypothess Testng We may wsh to test pror hypotheses about the coeffcents we estmate. We can use the estmates to test whether the data rejects our hypothess. An example mght be that we wsh to

More information

x = , so that calculated

x = , so that calculated Stat 4, secton Sngle Factor ANOVA notes by Tm Plachowsk n chapter 8 we conducted hypothess tests n whch we compared a sngle sample s mean or proporton to some hypotheszed value Chapter 9 expanded ths to

More information

Inductance Calculation for Conductors of Arbitrary Shape

Inductance Calculation for Conductors of Arbitrary Shape CRYO/02/028 Aprl 5, 2002 Inductance Calculaton for Conductors of Arbtrary Shape L. Bottura Dstrbuton: Internal Summary In ths note we descrbe a method for the numercal calculaton of nductances among conductors

More information

ELASTIC WAVE PROPAGATION IN A CONTINUOUS MEDIUM

ELASTIC WAVE PROPAGATION IN A CONTINUOUS MEDIUM ELASTIC WAVE PROPAGATION IN A CONTINUOUS MEDIUM An elastc wave s a deformaton of the body that travels throughout the body n all drectons. We can examne the deformaton over a perod of tme by fxng our look

More information

Applied Stochastic Processes

Applied Stochastic Processes STAT455/855 Fall 23 Appled Stochastc Processes Fnal Exam, Bref Solutons 1. (15 marks) (a) (7 marks) The dstrbuton of Y s gven by ( ) ( ) y 2 1 5 P (Y y) for y 2, 3,... The above follows because each of

More information

U.C. Berkeley CS294: Spectral Methods and Expanders Handout 8 Luca Trevisan February 17, 2016

U.C. Berkeley CS294: Spectral Methods and Expanders Handout 8 Luca Trevisan February 17, 2016 U.C. Berkeley CS94: Spectral Methods and Expanders Handout 8 Luca Trevsan February 7, 06 Lecture 8: Spectral Algorthms Wrap-up In whch we talk about even more generalzatons of Cheeger s nequaltes, and

More information

Introductory Cardinality Theory Alan Kaylor Cline

Introductory Cardinality Theory Alan Kaylor Cline Introductory Cardnalty Theory lan Kaylor Clne lthough by name the theory of set cardnalty may seem to be an offshoot of combnatorcs, the central nterest s actually nfnte sets. Combnatorcs deals wth fnte

More information

Foundations of Arithmetic

Foundations of Arithmetic Foundatons of Arthmetc Notaton We shall denote the sum and product of numbers n the usual notaton as a 2 + a 2 + a 3 + + a = a, a 1 a 2 a 3 a = a The notaton a b means a dvdes b,.e. ac = b where c s an

More information

Inner Product. Euclidean Space. Orthonormal Basis. Orthogonal

Inner Product. Euclidean Space. Orthonormal Basis. Orthogonal Inner Product Defnton 1 () A Eucldean space s a fnte-dmensonal vector space over the reals R, wth an nner product,. Defnton 2 (Inner Product) An nner product, on a real vector space X s a symmetrc, blnear,

More information

Kernel Methods and SVMs Extension

Kernel Methods and SVMs Extension Kernel Methods and SVMs Extenson The purpose of ths document s to revew materal covered n Machne Learnng 1 Supervsed Learnng regardng support vector machnes (SVMs). Ths document also provdes a general

More information

HMMT February 2016 February 20, 2016

HMMT February 2016 February 20, 2016 HMMT February 016 February 0, 016 Combnatorcs 1. For postve ntegers n, let S n be the set of ntegers x such that n dstnct lnes, no three concurrent, can dvde a plane nto x regons (for example, S = {3,

More information

Introduction to Vapor/Liquid Equilibrium, part 2. Raoult s Law:

Introduction to Vapor/Liquid Equilibrium, part 2. Raoult s Law: CE304, Sprng 2004 Lecture 4 Introducton to Vapor/Lqud Equlbrum, part 2 Raoult s Law: The smplest model that allows us do VLE calculatons s obtaned when we assume that the vapor phase s an deal gas, and

More information

Winter 2008 CS567 Stochastic Linear/Integer Programming Guest Lecturer: Xu, Huan

Winter 2008 CS567 Stochastic Linear/Integer Programming Guest Lecturer: Xu, Huan Wnter 2008 CS567 Stochastc Lnear/Integer Programmng Guest Lecturer: Xu, Huan Class 2: More Modelng Examples 1 Capacty Expanson Capacty expanson models optmal choces of the tmng and levels of nvestments

More information

See Book Chapter 11 2 nd Edition (Chapter 10 1 st Edition)

See Book Chapter 11 2 nd Edition (Chapter 10 1 st Edition) Count Data Models See Book Chapter 11 2 nd Edton (Chapter 10 1 st Edton) Count data consst of non-negatve nteger values Examples: number of drver route changes per week, the number of trp departure changes

More information

Markov Chain Monte Carlo (MCMC), Gibbs Sampling, Metropolis Algorithms, and Simulated Annealing Bioinformatics Course Supplement

Markov Chain Monte Carlo (MCMC), Gibbs Sampling, Metropolis Algorithms, and Simulated Annealing Bioinformatics Course Supplement Markov Chan Monte Carlo MCMC, Gbbs Samplng, Metropols Algorthms, and Smulated Annealng 2001 Bonformatcs Course Supplement SNU Bontellgence Lab http://bsnuackr/ Outlne! Markov Chan Monte Carlo MCMC! Metropols-Hastngs

More information

Affine transformations and convexity

Affine transformations and convexity Affne transformatons and convexty The purpose of ths document s to prove some basc propertes of affne transformatons nvolvng convex sets. Here are a few onlne references for background nformaton: http://math.ucr.edu/

More information

MMA and GCMMA two methods for nonlinear optimization

MMA and GCMMA two methods for nonlinear optimization MMA and GCMMA two methods for nonlnear optmzaton Krster Svanberg Optmzaton and Systems Theory, KTH, Stockholm, Sweden. krlle@math.kth.se Ths note descrbes the algorthms used n the author s 2007 mplementatons

More information

Queueing Networks II Network Performance

Queueing Networks II Network Performance Queueng Networks II Network Performance Davd Tpper Assocate Professor Graduate Telecommuncatons and Networkng Program Unversty of Pttsburgh Sldes 6 Networks of Queues Many communcaton systems must be modeled

More information

Simulated Power of the Discrete Cramér-von Mises Goodness-of-Fit Tests

Simulated Power of the Discrete Cramér-von Mises Goodness-of-Fit Tests Smulated of the Cramér-von Mses Goodness-of-Ft Tests Steele, M., Chaselng, J. and 3 Hurst, C. School of Mathematcal and Physcal Scences, James Cook Unversty, Australan School of Envronmental Studes, Grffth

More information

Supplement: Proofs and Technical Details for The Solution Path of the Generalized Lasso

Supplement: Proofs and Technical Details for The Solution Path of the Generalized Lasso Supplement: Proofs and Techncal Detals for The Soluton Path of the Generalzed Lasso Ryan J. Tbshran Jonathan Taylor In ths document we gve supplementary detals to the paper The Soluton Path of the Generalzed

More information

1. Inference on Regression Parameters a. Finding Mean, s.d and covariance amongst estimates. 2. Confidence Intervals and Working Hotelling Bands

1. Inference on Regression Parameters a. Finding Mean, s.d and covariance amongst estimates. 2. Confidence Intervals and Working Hotelling Bands Content. Inference on Regresson Parameters a. Fndng Mean, s.d and covarance amongst estmates.. Confdence Intervals and Workng Hotellng Bands 3. Cochran s Theorem 4. General Lnear Testng 5. Measures of

More information

Volume 18 Figure 1. Notation 1. Notation 2. Observation 1. Remark 1. Remark 2. Remark 3. Remark 4. Remark 5. Remark 6. Theorem A [2]. Theorem B [2].

Volume 18 Figure 1. Notation 1. Notation 2. Observation 1. Remark 1. Remark 2. Remark 3. Remark 4. Remark 5. Remark 6. Theorem A [2]. Theorem B [2]. Bulletn of Mathematcal Scences and Applcatons Submtted: 016-04-07 ISSN: 78-9634, Vol. 18, pp 1-10 Revsed: 016-09-08 do:10.1805/www.scpress.com/bmsa.18.1 Accepted: 016-10-13 017 ScPress Ltd., Swtzerland

More information

6. Stochastic processes (2)

6. Stochastic processes (2) Contents Markov processes Brth-death processes Lect6.ppt S-38.45 - Introducton to Teletraffc Theory Sprng 5 Markov process Consder a contnuous-tme and dscrete-state stochastc process X(t) wth state space

More information

1 Derivation of Rate Equations from Single-Cell Conductance (Hodgkin-Huxley-like) Equations

1 Derivation of Rate Equations from Single-Cell Conductance (Hodgkin-Huxley-like) Equations Physcs 171/271 -Davd Klenfeld - Fall 2005 (revsed Wnter 2011) 1 Dervaton of Rate Equatons from Sngle-Cell Conductance (Hodgkn-Huxley-lke) Equatons We consder a network of many neurons, each of whch obeys

More information

6. Stochastic processes (2)

6. Stochastic processes (2) 6. Stochastc processes () Lect6.ppt S-38.45 - Introducton to Teletraffc Theory Sprng 5 6. Stochastc processes () Contents Markov processes Brth-death processes 6. Stochastc processes () Markov process

More information

Randomness and Computation

Randomness and Computation Randomness and Computaton or, Randomzed Algorthms Mary Cryan School of Informatcs Unversty of Ednburgh RC 208/9) Lecture 0 slde Balls n Bns m balls, n bns, and balls thrown unformly at random nto bns usually

More information

Canonical transformations

Canonical transformations Canoncal transformatons November 23, 2014 Recall that we have defned a symplectc transformaton to be any lnear transformaton M A B leavng the symplectc form nvarant, Ω AB M A CM B DΩ CD Coordnate transformatons,

More information

Random Walks on Digraphs

Random Walks on Digraphs Random Walks on Dgraphs J. J. P. Veerman October 23, 27 Introducton Let V = {, n} be a vertex set and S a non-negatve row-stochastc matrx (.e. rows sum to ). V and S defne a dgraph G = G(V, S) and a drected

More information

PHYS 705: Classical Mechanics. Calculus of Variations II

PHYS 705: Classical Mechanics. Calculus of Variations II 1 PHYS 705: Classcal Mechancs Calculus of Varatons II 2 Calculus of Varatons: Generalzaton (no constrant yet) Suppose now that F depends on several dependent varables : We need to fnd such that has a statonary

More information

BOUNDEDNESS OF THE RIESZ TRANSFORM WITH MATRIX A 2 WEIGHTS

BOUNDEDNESS OF THE RIESZ TRANSFORM WITH MATRIX A 2 WEIGHTS BOUNDEDNESS OF THE IESZ TANSFOM WITH MATIX A WEIGHTS Introducton Let L = L ( n, be the functon space wth norm (ˆ f L = f(x C dx d < For a d d matrx valued functon W : wth W (x postve sem-defnte for all

More information

CS 798: Homework Assignment 2 (Probability)

CS 798: Homework Assignment 2 (Probability) 0 Sample space Assgned: September 30, 2009 In the IEEE 802 protocol, the congeston wndow (CW) parameter s used as follows: ntally, a termnal wats for a random tme perod (called backoff) chosen n the range

More information

CONJUGACY IN THOMPSON S GROUP F. 1. Introduction

CONJUGACY IN THOMPSON S GROUP F. 1. Introduction CONJUGACY IN THOMPSON S GROUP F NICK GILL AND IAN SHORT Abstract. We complete the program begun by Brn and Squer of charactersng conjugacy n Thompson s group F usng the standard acton of F as a group of

More information

Remarks on the Properties of a Quasi-Fibonacci-like Polynomial Sequence

Remarks on the Properties of a Quasi-Fibonacci-like Polynomial Sequence Remarks on the Propertes of a Quas-Fbonacc-lke Polynomal Sequence Brce Merwne LIU Brooklyn Ilan Wenschelbaum Wesleyan Unversty Abstract Consder the Quas-Fbonacc-lke Polynomal Sequence gven by F 0 = 1,

More information

Copyright 2017 by Taylor Enterprises, Inc., All Rights Reserved. Adjusted Control Limits for P Charts. Dr. Wayne A. Taylor

Copyright 2017 by Taylor Enterprises, Inc., All Rights Reserved. Adjusted Control Limits for P Charts. Dr. Wayne A. Taylor Taylor Enterprses, Inc. Control Lmts for P Charts Copyrght 2017 by Taylor Enterprses, Inc., All Rghts Reserved. Control Lmts for P Charts Dr. Wayne A. Taylor Abstract: P charts are used for count data

More information