Online Learning Algorithms for Stochastic Water-Filling

Size: px
Start display at page:

Download "Online Learning Algorithms for Stochastic Water-Filling"

Transcription

1 Onlne Lernng Algorthms for Stochstc Wter-Fllng Y G nd Bhskr Krshnmchr Mng Hseh Deprtment of Electrcl Engneerng Unversty of Southern Clforn Los Angeles, CA 90089, USA Eml: {yg, bkrshn}@usc.edu Abstrct Wter-fllng s the term for the clssc soluton to the problem of lloctng constrned power to set of prllel chnnels to mxmze the totl dt-rte. It s used wdely n prctce, for exmple, for power llocton to sub-crrers n mult-user OFDM systems such s WMx. The clssc wterfllng lgorthm s determnstc nd requres perfect knowledge of the chnnel gn to nose rtos. In ths pper we consder how to do power llocton over stochstclly tme-vryng (..d.) chnnels wth unknown gn to nose rto dstrbutons. We dopt n onlne lernng frmework bsed on stochstc multrmed bndts. We consder two vrtons of the problem, one n whch the gol s to fnd power llocton to mxmze E[log(1 + SNR )], nd nother n whch the gol s to fnd power llocton to mxmze log(1+e[snr ]). For the frst problem, we propose cogntve wter-fllng lgorthm tht we cll CWF1. We show tht CWF1 obtns regret (defned s the cumultve gp over tme between the sum-rte obtned by dstrbuton-wre gene nd ths polcy) tht grows polynomlly n the number of chnnels nd logrthmclly n tme, mplyng tht t symptotclly cheves the optml tme-verged rte tht cn be obtned when the gn dstrbutons re known. For the second problem, we present n lgorthm clled CWF2, whch s, to our knowledge, the frst lgorthm n the lterture on stochstc mult-rmed bndts to explot non-lner dependences between the rms. We prove tht the number of tmes CWF2 pcks the ncorrect power llocton s bounded by functon tht s polynoml n the number of chnnels nd logrthmc n tme, mplyng tht ts frequency of ncorrect llocton tends to zero. I. INTRODUCTION A fundmentl resource llocton problem tht rses n mny settngs n communcton networks s to llocte constrned mount of power cross mny prllel chnnels n order to mxmze the sum-rte. Assumng tht the powerrte functon for ech chnnel s proportonl to log(1+snr) s per the Shnnon s cpcty theorem for AWGN chnnels, t s well known tht the optml power llocton cn be determned by wter-fllng strtegy [1]. The clssc wterfllng soluton s determnstc lgorthm, nd requres perfect knowledge of ll chnnel gn to nose rtos. In prctce, however, chnnel gn-to-nose rtos re stochstc qunttes. To hndle ths rndomness, we consder n lterntve pproch, bsed on onlne lernng, specfclly Ths reserch ws sponsored n prt by the U.S. Army Reserch Lbortory under the Network Scence Collbortve Technology Allnce, Agreement Number W911NF stochstc mult-rmed bndts. We formulte the problem of stochstc wter-fllng s follows: tme s dscretzed nto slots; ech chnnel s gn-to-nose rto s modeled s n..d. rndom vrble wth n unknown dstrbuton. In our generl formulton, the power-to-rte functon for ech chnnel s llowed to be ny sub-ddtve functon 1. We seek power llocton tht mxmzes the expected sum-rte (.e., n optmzton of the form E[ log(1 + SNR )]). Even f the chnnel gn-to-nose rtos re rndom vrbles wth known dstrbutons, ths turns out to be hrd combntorl stochstc optmzton problem. Our focus n ths pper s thus on more chllengng cse. In the clsscl mult-rmed bndt, there s plyer plyng K rms tht yeld stochstc rewrds wth unknown mens t ech tme n..d. fshon over tme. The plyer seeks polcy to mxmze ts totl expected rewrd over tme. The performnce metrc of nterest n such problems s regret, defned s the cumultve dfference n expected rewrd between modelwre gene nd tht obtned by the gven lernng polcy. And t s of nterest to show tht the regret grows sub-lnerly wth tme so tht the tme-verged regret symptotclly goes to zero, mplyng tht the tme-verged rewrd of the modelwre gene s obtned symptotclly by the lernng polcy. We show tht t s possble to mp the problem of stochstc wter-fllng to n MAB formulton by tretng ech possble power llocton s n rm (we consder dscrete power levels n ths pper; f there re P possble power levels for ech of N chnnels, there would be P N totl rms.) We present novel combntorl polcy for ths problem tht we cll CWF1, tht yelds regret growng polynomlly n N nd logrthmclly over tme. Despte the exponentl growng set of rms, the CWF1 observes nd mntns nformton for P N vrbles, one correspondng to ech power-level nd chnnel, nd explots lner dependences between the rms bsed on these vrbles. Typclly, the wy the rndomness n the chnnel gn to nose rtos s delt wth s tht the men chnnel gn to nose rtos re estmted frst bsed on vergng fnte set of trnng observtons nd then the estmted gns re used n determnstc wter-fllng procedure. Essentlly ths 1 A functon f s subddtve f f(x + y) f(x) + f(y); for ny concve functon g, f g(0) 0 (such s log(1 + x)), g s subddtve.

2 2 pproch tres to dentfy the power llocton tht mxmzes pseudo-sum-rte, whch s determned bsed on the powerrte equton ppled to the men chnnel gn-to-nose rtos (.e., n optmzton of the form log(1 + E[SNR ]). We lso present dfferent stochstc wter-fllng lgorthm tht we cll CWF2, whch lerns to do ths n n onlne fshon. Ths lgorthm observes nd mntns nformton for N vrbles, one correspondng to ech chnnel, nd explots non-lner dependences between the rms bsed on these vrbles. To our knowledge, CWF2 s the frst MAB lgorthm to explot non-lner dependences between the rms. We show tht the number of tmes CWF2 plys non-optml combnton of powers s unformly bounded by functon tht s logrthmc n tme. Under some restrctve condtons, CWF2 my lso solve the frst problem more effcently. II. RELATED WORK The clssc wter-fllng strtegy s descrbed n [1]. There re few other stochstc vrtons of wter-fllng tht hve been covered n the lterture tht re dfferent n sprt from our formulton. When fdng dstrbuton over the gns s known pror, the power constrnt s expressed over tme, nd the nstntneous gns re lso known, then determnstc jont frequency-tme wter-fllng strtegy cn be used [2], [3]. In [4], stochstc grdent pproch bsed on Lgrnge dulty s proposed to solve ths problem when the fdng dstrbuton s unknown but stll nstntneous gns re vlble. By contrst, n our work we do not ssume tht the nstntneous gns re known, nd focus on keepng the sme power constrnt t ech tme whle consderng unknown gn dstrbutons. Another work [5] consders wter-fllng over stochstc non-sttonry fdng chnnels, nd proposes n dptve lernng lgorthm tht trcks the tme-vryng optml power llocton by ncorportng forgettng fctor. However, the focus of ther lgorthm s on mnmzng the mxmum men squred error ssumng mperfect chnnel estmtes, nd they prove only tht ther lgorthm would converge n sttonry settng. Although ther lgorthm cn be vewed s lernng mechnsm, they do not tret stochstc wterfllng from the perspectve of mult-rmed bndts, whch s novel contrbuton of our work. In our work, we focus on sttonry settng wth perfect chnnel estmtes, but prove stronger results, showng tht our lernng lgorthm not only converges to the optml llocton, t does so wth sub-lner regret. There hs been long lne of work on stochstc multrmed bndts nvolvng plyng rms yeldng stochstclly tme vryng rewrds wth unknown dstrbutons. Severl uthors [6] [9] present lernng polces tht yeld regret growng logrthmclly over tme (symptotclly, n the cse of [6] [8] nd unformly over tme n the cse of [9]). Our lgorthms buld on the UCB1 lgorthm proposed n [9] but mke sgnfcnt modfctons to hndle the combntorl nture of the rms n ths problem. CWF1 hs some commonltes wth the LLR lgorthm we recently developed for completely dfferent problem, tht of stochstc combntorl bprtte mtchng for chnnel llocton [10], but s modfed to ccount for the non-lner power-rte functon n ths pper. Other recent work on stochstc MAB hs consdered decentrlzed settngs [11] [14], nd non-..d. rewrd processes [15] [19]. Wth respect to ths lterture, the problem settng for stochstc wter-fllng s novel n tht t nvolves non-lner functon of the cton nd unknown vrbles. In prtculr, s fr s we re wre, our CWF2 polcy s the frst to explot the non-lner dependences between rms to provbly mprove the regret performnce. III. PROBLEM FORMULATION We defne the stochstc verson of the clssc communcton theory problem of power llocton for mxmzng rte over prllel chnnels (wter-fllng) s follows. We consder system wth N chnnels, where the chnnel gn-to-nose rtos re unknown rndom processes X (n), 1 N. Tme s slotted nd ndexed by n. We ssume tht X (n) evolves s n..d. rndom process over tme (.e., we consder block fdng), wth the only restrcton tht ts dstrbuton hs fnte support. Wthout loss of generlty, we normlze X (n) [0, 1]. We do not requre tht X (n) be ndependent cross. Ths rndom process s ssumed to hve men θ = E[X ] tht s unknown to the users. We denote the set of ll these mens by Θ = {θ }. At ech decson perod n (lso referred to nterchngebly s tme slot), n N-dmensonl cton vector (n), representng power llocton on these N chnnels, s selected under polcy π(n). We ssume tht the power levels re dscrete, nd we cn put ny constrnt on the selectons of power lloctons such tht they re from fnte set F (.e., the mxmum totl power constrnt, or n upper bound on the mxmum llowed power per subcrrer). We ssume (n) 0 for ll 1 N. When prtculr power llocton (n) s selected, the chnnel gn-to-nose rtos correspondng to nonzero components of (n) re reveled,.e., the vlue of X (n) s observed for ll such tht (n) 0. We denote by A (n) = { : (n) 0, 1 N} the ndex set of ll (n) 0 for n llocton. We dopt generl formulton for wter-fllng, where the sum rte 2 obtned t tme n by lloctng set of powers (n) s defned s: R (n) (n) = (n) f ( (n), X (n)). (1) where for ll, f ( (n), X (n)) s nonlner contnuous ncresng sub-ddtve functon n X (n), nd f ( (n), 0) = 0 for ny (n). We ssume f s defned on R + R +. Our formulton s generl enough to nclude s specl cse of the rte functon obtned from Shnnon s cpcty theorem for AWGN, whch s wdely used n communcton 2 We refer to rte nd rewrd nterchngebly n ths pper.

3 3 networks: R (n) (n) = N log(1 + (n)x (n)) In the typcl formulton there s totl power constrnt nd ndvdul power constrnts, the correspondng constrnt s F = { : N P totl 0 P, }. where P totl s the totl power constrnt nd P s the mxmum llowed power per chnnel. Our gol s to mxmze the expected sum-rte when the dstrbutons of ll X re unknown, s shown n (2). We refer to ths objectve s O 1. mx E[ f (, X ))] (2) Note tht even when X hve known dstrbutons, ths s hrd combntorl non-lner stochstc optmzton problem. In our settng, wth unknown dstrbutons, we cn formulte ths s mult-rmed bndt problem, where ech power llocton (n) F s n rm nd the rewrd functon s n combntorl non-lner form. The optml rms re the ones wth the lrgest expected rewrd, denoted s O = { }. For the rest of the pper, we use s the ndex ndctng tht prmeter s for n optml rm. If more thn one optml rm exsts, refers to ny one of them. We note tht for the combntorl mult-rmed bndt problem wth lner rewrds where the rewrd functon s defned by R (n) (n) = (n)x (n), s soluton to determnstc optmzton problem becuse mx E[ (n) X ] = mx E[X ]. Dfferent from the combntorl multrmed bndt problem wth lner rewrds, here s soluton to stochstc optmzton problem,.e., O = {ã : ã = rg mx E[ f (, X ))]}. (3) We evlute polces for O 1 wth respect to regret, whch s defned s the dfference between the expected rewrd tht could be obtned by gene tht cn pck n optml rm t ech tme, nd tht obtned by the gven polcy. Note tht mnmzng the regret s equvlent to mxmzng the expected rewrds. Regret cn be expressed s: n R π (n) = nr E[ R π(t) (t)], (4) t=1 where R = mx E[ f (, X ))], the expected rewrd of n optml rm. Intutvely, we would lke the regret R π (n) to be s smll s possble. If t s sub-lner wth respect to tme n, the tmeverged regret wll tend to zero nd the mxmum possble tme-verged rewrd cn be cheved. Note tht the number of rms F cn be exponentl n the number of unknown rndom vrbles N. We lso note tht for the stochstc verson of the wterfllng problems, typcl wy n prctce to del wth the unknown rndomness s to estmte the men chnnel gn to nose rtos frst nd then fnd the optmzed llocton bsed on the men vlues. Ths pproch tres to dentfy the power llocton tht mxmzes the power-rte equton ppled to the men chnnel gn-to-nose rtos. We refer to mxmzng ths s the sum-pseudo-rte over verged chnnels. We denote ths objectve by O 2, s shown n (5). mx f (, E[X ]) (5) We would lso lke to develop n onlne lernng polcy for O 2. Note tht the optml rm of O 2 s soluton to determnstc optmzton problem. So, we evlute the polces for O 2 wth respect to the expected totl number of tmes tht non-optml power llocton s selected. We denote by T (n) the number of tmes tht power llocton s pcked up to tme n. We denote r = f (, E[X ]). Let Tnon(n) π denote the totl number of tmes tht polcy π select power llocton r < r. Denote by 1 π t () the ndctor functon whch s equl to 1 f s selected under polcy π t tme t, nd 0 else. Then n E[Tnon π (n)] = n E[ 1 π t ( ) = 1] (6) = r <r t=1 E[T (n)]. IV. ONLINE LEARNING FOR MAXIMIZING THE SUM-RATE We frst present n ths secton n onlne lernng polcy for stochstc wter-fllng under object O 1. A. Polcy Desgn A strghtforwrd, nve wy to solve ths problem s to use the UCB1 polcy proposed [9]. For UCB1, ech power llocton s treted s n rm, nd the rm tht mxmzes 2ln n m k Ŷ k + wll be selected t ech tme slot, where Ŷk s the men observed rewrd on rm k, nd m k s the number of tmes tht rm k hs been plyed. Ths pproch essentlly gnores the underlyng dependences cross the dfferent rms, nd requres storge tht s lner n the number of rms nd yelds regret growng lnerly wth the number of rms. Snce there cn be n exponentl number of rms, the UCB1 lgorthm performs poorly on ths problem. We note tht for combntorl optmzton problems wth lner rewrd functons, n onlne lernng lgorthm LLR hs been proposed n [6] s n effcent soluton. LLR stores the men of observed vlues for every underlyng unknown rndom vrble, s well s the number of tmes ech hs been observed. So the storge of LLR s lner n the number of unknown rndom vrbles, nd the nlyss n [6] shows LLR cheves regret tht grows logrthmclly n tme, nd polynomlly n the number of unknown prmeters.

4 4 However, the chllenge wth stochstc wter-fllng wth objectve O 1, where the expectton s outsde the non-lner rewrd functon, drectly storng the men observtons of X wll not work. To del wth ths chllenge, we propose to store the nformton for ech, X combnton,.e., 1 N,, we defne new set of rndom vrbles Y, = f (, X ). So now the number of rndom vrbles Y, s N B, where B = { : 0}. Note tht N B PN. Then the rewrd functon cn be expressed s R = Y,, (7) Note tht (7) s n combntorl lner form. For ths redefned MAB problem wth N B unknown rndom vrbles nd lner rewrd functon (7), we propose the followng onlne lernng polcy CWF1 for stochstc wter-fllng s shown n Algorthm 1. Algorthm 1 Onlne Lernng for Stochstc Wter-Fllng: CWF1 1: // INITIALIZATION 2: If mx A s known, let L = mx A ; else, L = N; 3: for n = 1 to N do 4: Ply ny rm such tht n A ; 5: A, B, Y, := Y, m+f(,x) m +1 ; 6: A, m := m + 1; 7: end for 8: // MAIN LOOP 9: whle 1 do 10: n := n + 1; 11: Ply n rm whch solves the mxmzton problem (L + 1)lnn (Y, + ); (8) m 12: A, B, Y, := Y, m+f(,x) m +1 ; 13: A, m := m + 1; 14: end whle To hve tghter bound of regret, dfferent from the LLR lgorthm, nsted of storng the number of tmes tht ech unknown rndom vrbles Y, hs been observed, we use 1 by N vector, denoted s (m ) 1 N, to store the number of tmes tht X hs been observed up to the current tme slot. We use 1 by N B vector, denoted s (Y, )1 N B to store the nformton bsed on the observed vlues. (Y, s updted n s shown n lne 12. Ech tme )1 N B n rm (n) s plyed, A (n), the observed vlue of X s obtned. For every observed vlue of X, B vlues re updted: B, the verge vlue Y, of ll the vlues of Y, up to the current tme slot s updted. CWF1 polcy requres storge lner n N B. B. Anlyss of regret Theorem 1: The expected regret under the CWF1 polcy s t most [ ] 4L 2 (L + 1)N lnn ( mn ) 2 + N + π2 3 LN mx. (9) where mn = mn = R E[R ], mx = mx = R E[R ]. Note tht L N. The proof of Theorem 1 s omtted. Remrk 1: For CWF1 polcy, lthough there re N B rndom vrbles, the upper bound of regret remns O(N 4 log n), whch s the sme s LLR, s shown by Theorem 2 n [6]. Drectly pplyng LLR lgorthm to solve the redefned MAB problem n (7) wll result n regret tht grows s O(P 4 N 4 log n). Remrk 2: Algorthm 1 wll even work for rte functons tht do not stsfy subddtvty. Remrk 3: We cn develop smlr polces nd results when X re Mrkovn rewrds s n [19] nd [20]. V. ONLINE LEARNING FOR SUM-PSEUDO-RATE We now show our novel onlne lernng lgorthm CWF2 for stochstc wter-fllng wth object O 2. Unlke CWF1, CWF2 explots non-lner dependences between the choces of power lloctons nd requres lower storge. Under condton where the power llocton tht mxmze O 2 lso mxmze O 1, we wll see through smultons tht CWF2 hs better regret performnce. A. Polcy Desgn Our proposed polcy CWF2 for stochstc wter fllng wth objectve O 2 s shown n Algorthm 2. We use two 1 by N vectors to store the nformton fter we ply n rm t ech tme slot. One s (X ) 1 N n whch X s the verge (smple men) of ll the observed vlues of X up to the current tme slot (obtned through potentlly dfferent sets of rms over tme). The other one s (m ) 1 N n whch m s the number of tmes tht X hs been observed up to the current tme slot. So CWF2 polcy requres storge lner n N. B. Anlyss of regret For the nlyss of the upper bound for E[Tnon π (n)] of CWF2 polcy, we use the nequltes s stted n the Chernoff- Hoeffdng bound s follows: Lemm 1 (Chernoff-Hoeffdng bound [21]): X 1,...,X n re rndom vrbles wth rnge [0, 1], nd

5 5 Algorthm 2 Onlne Lernng for Stochstc Wter-Fllng: CWF2 1: // INITIALIZATION 2: If mx A s known, let L = mx A ; else, L = N; 3: for n = 1 to N do 4: Ply ny rm such tht n A ; 5: A, X := Xm+X m +1, m := m + 1; 6: end for 7: // MAIN LOOP 8: whle 1 do 9: n := n + 1; 10: Ply n rm whch solves the mxmzton problem mx f (, X ) + f (, (L + 1)lnn ) m 11: A (n), X := Xm+X m +1, m := m + 1; 12: end whle ; (10) E[X t X 1,..., X t 1 ] = µ, 1 t n. Denote S n = X. Then for ll 0 P{S n nµ + } e 22 /n P{S n nµ } e 22 /n (11) Theorem 2: Under the CWF2 polcy, the expected totl number of tmes tht non-optml power lloctons re selected s t most E[Tnon π N(L + 1)lnn (n)] Bmn 2 + N + π2 LN, (12) 3 where B mn s constnt defned by δ mn nd L; δ mn = mn :r <r (r r ). Proof: See [22]. Remrk 4: CWF2 cn be used to solve the stochstc wterfllng wth objectve O 1 s well f O, such tht / O, f (, θ )) > f j ( j, θ j ). (13) j A Then the regret of CWF2 s t most [ N(L + 1)lnn R CWF2 (n) B 2 mn REFERENCES + N + π2 3 LN ] mx, (14) [1] T. Cover nd J. Thoms, Elements of Informton Theory. New York: Wley, [2] A. J. Goldsmth nd P. P. Vry, Cpcty of Fdng MIMO Chnnels wth Chnnel Estmton Error, IEEE Interntonl Conference on Communctons (ICC), June, [3] A. J. Goldsmth, Wreless Communctons. New York: Cmbrdge Unversty Press, [4] X. Wng, D. Wng, H. Zhung, nd S. D. Morger, Energy- Effcent Resource Allocton n Wreless Sensor Networks over Fdng TDMA, vol. 28, no. 7, pp , [5] I. Zd nd V. Krshnmurthy, Stochstc Adptve Multlevel Wterfllng n MIMO-OFDM WLANs, the 39th Aslomr Conference on Sgnls, Systems nd Computers, October, [6] Y. G, B. Krshnmchr nd R. Jn, Combntorl Network Optmzton wth Unknown Vrbles: Mult-Armed Bndts wth Lner Rewrds nd Indvdul Observtons, to pper n IEEE/ACM Trnsctons on Networkng. [7] V. Annthrm, P. Vry, nd J. Wlrnd, Asymptotclly Effcent Allocton Rules for the Multrmed Bndt Problem wth Multple Plys-Prt I: IID Rewrds, IEEE Trnsctons on Automtc Control, vol. 32, no. 11, pp , [8] R. Agrwl, Smple Men Bsed Index Polces wth O(log n) Regret for the Mult-Armed Bndt Problem, Advnces n Appled Probblty, vol. 27, no. 4, pp , [9] P. Auer, N. Ces-Bnch, nd P. Fscher, Fnte-tme Anlyss of the Multrmed Bndt Problem, Mchne Lernng, vol. 47, no. 2, pp , [10] Y. G, B. Krshnmchr, nd R. Jn, Lernng Multuser Chnnel Alloctons n Cogntve Rdo Networks: A Combntorl Mult-rmed Bndt Formulton, IEEE Interntonl Dynmc Spectrum Access Networks (DySPAN) Symposum, Sngpore, Aprl, [11] A. Anndkumr, N. Mchel, nd A.K. Tng, Opportunstc Spectrum Access wth Multple Users: Lernng under Competton, IEEE Interntonl Conference on Computer Communctons (INFOCOM), Mrch, [12] A. Anndkumr, N. Mchel, A. Tng, nd A. Swm, Dstrbuted Lernng nd Allocton of Cogntve Users wth Logrthmc Regret, IEEE JSAC on Advnces n Cogntve Rdo Networkng nd Communctons, vol. 29, no. 4, pp , [13] K. Lu nd Q. Zho, Dstrbuted Lernng n Cogntve Rdo Networks: Mult-Armed Bndt wth Dstrbuted Multple Plyers, Proc. of IEEE Interntonl Conference on Acoustcs, Speech, nd Sgnl Processng (ICASSP), Mrch, [14] Y. G nd B. Krshnmchr, Decentrlzed Onlne Lernng Algorthms for Opportunstc Spectrum Access, IEEE Globl Communctons Conference (GLOBECOM), December, [15] C. Tekn nd M. Lu, Onlne Algorthms for the Mult-Armed Bndt Problem wth Mrkovn Rewrds, the 48th Annul Allerton Conference on Communcton, Control, nd Computng (Allerton), September, [16] C. Tekn nd M. Lu, Onlne Lernng n Opportunstc Spectrum Access: Restless Bndt Approch, IEEE Interntonl Conference on Computer Communctons (INFOCOM), Aprl, [17] H. Lu, K. Lu, nd Q. Zho, Lernng nd Shrng n Chngng World: Non-Byesn Restless Bndt wth Multple Plyers, Informton Theory nd Applctons Workshop (ITA), Jnury, [18] W. D, Y. G, B. Krshnmchr nd Q. Zho, The Non-Byesn Restless Mult-Armed Bndt: Cse of Ner- Logrthmc Regret, IEEE Interntonl Conference on Acoustcs, Speech, nd Sgnl Processng (ICASSP), My, [19] Y. G, B. Krshnmchr nd M. Lu, On the Combntorl Mult-Armed Bndt Problem wth Mrkovn Rewrds, IEEE Globl Communctons Conference (GLOBECOM), December, [20] Y. G, B. Krshnmchr nd M. Lu, Onlne lernng for combntorl network optmzton wth restless mrkovn rewrds, rxv: [21] D. Pollrd, Convergence of Stochstc Processes. Berln: Sprnger, [22] Y. G nd B. Krshnmchr, Onlne Lernng Algorthms for Stochstc Wter-Fllng, rxv:

Dennis Bricker, 2001 Dept of Industrial Engineering The University of Iowa. MDP: Taxi page 1

Dennis Bricker, 2001 Dept of Industrial Engineering The University of Iowa. MDP: Taxi page 1 Denns Brcker, 2001 Dept of Industrl Engneerng The Unversty of Iow MDP: Tx pge 1 A tx serves three djcent towns: A, B, nd C. Ech tme the tx dschrges pssenger, the drver must choose from three possble ctons:

More information

UNIVERSITY OF IOANNINA DEPARTMENT OF ECONOMICS. M.Sc. in Economics MICROECONOMIC THEORY I. Problem Set II

UNIVERSITY OF IOANNINA DEPARTMENT OF ECONOMICS. M.Sc. in Economics MICROECONOMIC THEORY I. Problem Set II Mcroeconomc Theory I UNIVERSITY OF IOANNINA DEPARTMENT OF ECONOMICS MSc n Economcs MICROECONOMIC THEORY I Techng: A Lptns (Note: The number of ndctes exercse s dffculty level) ()True or flse? If V( y )

More information

Partially Observable Systems. 1 Partially Observable Markov Decision Process (POMDP) Formalism

Partially Observable Systems. 1 Partially Observable Markov Decision Process (POMDP) Formalism CS294-40 Lernng for Rootcs nd Control Lecture 10-9/30/2008 Lecturer: Peter Aeel Prtlly Oservle Systems Scre: Dvd Nchum Lecture outlne POMDP formlsm Pont-sed vlue terton Glol methods: polytree, enumerton,

More information

Principle Component Analysis

Principle Component Analysis Prncple Component Anlyss Jng Go SUNY Bufflo Why Dmensonlty Reducton? We hve too mny dmensons o reson bout or obtn nsghts from o vsulze oo much nose n the dt Need to reduce them to smller set of fctors

More information

Rank One Update And the Google Matrix by Al Bernstein Signal Science, LLC

Rank One Update And the Google Matrix by Al Bernstein Signal Science, LLC Introducton Rnk One Updte And the Google Mtrx y Al Bernsten Sgnl Scence, LLC www.sgnlscence.net here re two dfferent wys to perform mtrx multplctons. he frst uses dot product formulton nd the second uses

More information

Applied Statistics Qualifier Examination

Applied Statistics Qualifier Examination Appled Sttstcs Qulfer Exmnton Qul_june_8 Fll 8 Instructons: () The exmnton contns 4 Questons. You re to nswer 3 out of 4 of them. () You my use ny books nd clss notes tht you mght fnd helpful n solvng

More information

Remember: Project Proposals are due April 11.

Remember: Project Proposals are due April 11. Bonformtcs ecture Notes Announcements Remember: Project Proposls re due Aprl. Clss 22 Aprl 4, 2002 A. Hdden Mrov Models. Defntons Emple - Consder the emple we tled bout n clss lst tme wth the cons. However,

More information

Lecture 4: Piecewise Cubic Interpolation

Lecture 4: Piecewise Cubic Interpolation Lecture notes on Vrtonl nd Approxmte Methods n Appled Mthemtcs - A Perce UBC Lecture 4: Pecewse Cubc Interpolton Compled 6 August 7 In ths lecture we consder pecewse cubc nterpolton n whch cubc polynoml

More information

Two Coefficients of the Dyson Product

Two Coefficients of the Dyson Product Two Coeffcents of the Dyson Product rxv:07.460v mth.co 7 Nov 007 Lun Lv, Guoce Xn, nd Yue Zhou 3,,3 Center for Combntorcs, LPMC TJKLC Nnk Unversty, Tnjn 30007, P.R. Chn lvlun@cfc.nnk.edu.cn gn@nnk.edu.cn

More information

Statistics and Probability Letters

Statistics and Probability Letters Sttstcs nd Probblty Letters 79 (2009) 105 111 Contents lsts vlble t ScenceDrect Sttstcs nd Probblty Letters journl homepge: www.elsever.com/locte/stpro Lmtng behvour of movng verge processes under ϕ-mxng

More information

CONTEXTUAL MULTI-ARMED BANDIT ALGORITHMS FOR PERSONALIZED LEARNING ACTION SELECTION. Indu Manickam, Andrew S. Lan, and Richard G.

CONTEXTUAL MULTI-ARMED BANDIT ALGORITHMS FOR PERSONALIZED LEARNING ACTION SELECTION. Indu Manickam, Andrew S. Lan, and Richard G. CONTEXTUAL MULTI-ARMED BANDIT ALGORITHMS FOR PERSONALIZED LEARNING ACTION SELECTION Indu Mnckm, Andrew S. Ln, nd Rchrd G. Brnuk Rce Unversty ABSTRACT Optmzng the selecton of lernng resources nd prctce

More information

Variable time amplitude amplification and quantum algorithms for linear algebra. Andris Ambainis University of Latvia

Variable time amplitude amplification and quantum algorithms for linear algebra. Andris Ambainis University of Latvia Vrble tme mpltude mplfcton nd quntum lgorthms for lner lgebr Andrs Ambns Unversty of Ltv Tlk outlne. ew verson of mpltude mplfcton;. Quntum lgorthm for testng f A s sngulr; 3. Quntum lgorthm for solvng

More information

Using Predictions in Online Optimization: Looking Forward with an Eye on the Past

Using Predictions in Online Optimization: Looking Forward with an Eye on the Past Usng Predctons n Onlne Optmzton: Lookng Forwrd wth n Eye on the Pst Nngjun Chen Jont work wth Joshu Comden, Zhenhu Lu, Anshul Gndh, nd Adm Wermn 1 Predctons re crucl for decson mkng 2 Predctons re crucl

More information

Definition of Tracking

Definition of Tracking Trckng Defnton of Trckng Trckng: Generte some conclusons bout the moton of the scene, objects, or the cmer, gven sequence of mges. Knowng ths moton, predct where thngs re gong to project n the net mge,

More information

The Schur-Cohn Algorithm

The Schur-Cohn Algorithm Modelng, Estmton nd Otml Flterng n Sgnl Processng Mohmed Njm Coyrght 8, ISTE Ltd. Aendx F The Schur-Cohn Algorthm In ths endx, our m s to resent the Schur-Cohn lgorthm [] whch s often used s crteron for

More information

PERFORMANCE OF OPPORTUNISTIC SPECTRUM ACCESS WITH SENSING ERROR IN COGNITIVE RADIO AD HOC NETWORKS

PERFORMANCE OF OPPORTUNISTIC SPECTRUM ACCESS WITH SENSING ERROR IN COGNITIVE RADIO AD HOC NETWORKS Journl of Engneerng Scence nd Technology Vol. 7, No. 2 (202) 42-55 School of Engneerng, Tylor s Unversty PERFORMANCE OF OPPORTUNISTIC SPECTRUM ACCESS WITH SENSING ERROR IN COGNITIVE RADIO AD HOC NETWORKS

More information

CHI-SQUARE DIVERGENCE AND MINIMIZATION PROBLEM

CHI-SQUARE DIVERGENCE AND MINIMIZATION PROBLEM CHI-SQUARE DIVERGENCE AND MINIMIZATION PROBLEM PRANESH KUMAR AND INDER JEET TANEJA Abstrct The mnmum dcrmnton nformton prncple for the Kullbck-Lebler cross-entropy well known n the lterture In th pper

More information

Jean Fernand Nguema LAMETA UFR Sciences Economiques Montpellier. Abstract

Jean Fernand Nguema LAMETA UFR Sciences Economiques Montpellier. Abstract Stochstc domnnce on optml portfolo wth one rsk less nd two rsky ssets Jen Fernnd Nguem LAMETA UFR Scences Economques Montpeller Abstrct The pper provdes restrctons on the nvestor's utlty functon whch re

More information

Katholieke Universiteit Leuven Department of Computer Science

Katholieke Universiteit Leuven Department of Computer Science Updte Rules for Weghted Non-negtve FH*G Fctorzton Peter Peers Phlp Dutré Report CW 440, Aprl 006 Ktholeke Unverstet Leuven Deprtment of Computer Scence Celestjnenln 00A B-3001 Heverlee (Belgum) Updte Rules

More information

4. Eccentric axial loading, cross-section core

4. Eccentric axial loading, cross-section core . Eccentrc xl lodng, cross-secton core Introducton We re strtng to consder more generl cse when the xl force nd bxl bendng ct smultneousl n the cross-secton of the br. B vrtue of Snt-Vennt s prncple we

More information

Online Stochastic Matching: New Algorithms with Better Bounds

Online Stochastic Matching: New Algorithms with Better Bounds Onlne Stochstc Mtchng: New Algorthms wth Better Bounds Ptrck Jllet Xn Lu My 202; revsed Jnury 203, June 203 Abstrct We consder vrnts of the onlne stochstc bprtte mtchng problem motvted by Internet dvertsng

More information

M/G/1/GD/ / System. ! Pollaczek-Khinchin (PK) Equation. ! Steady-state probabilities. ! Finding L, W q, W. ! π 0 = 1 ρ

M/G/1/GD/ / System. ! Pollaczek-Khinchin (PK) Equation. ! Steady-state probabilities. ! Finding L, W q, W. ! π 0 = 1 ρ M/G//GD/ / System! Pollcze-Khnchn (PK) Equton L q 2 2 λ σ s 2( + ρ ρ! Stedy-stte probbltes! π 0 ρ! Fndng L, q, ) 2 2 M/M/R/GD/K/K System! Drw the trnston dgrm! Derve the stedy-stte probbltes:! Fnd L,L

More information

GAUSS ELIMINATION. Consider the following system of algebraic linear equations

GAUSS ELIMINATION. Consider the following system of algebraic linear equations Numercl Anlyss for Engneers Germn Jordnn Unversty GAUSS ELIMINATION Consder the followng system of lgebrc lner equtons To solve the bove system usng clsscl methods, equton () s subtrcted from equton ()

More information

Computing a complete histogram of an image in Log(n) steps and minimum expected memory requirements using hypercubes

Computing a complete histogram of an image in Log(n) steps and minimum expected memory requirements using hypercubes Computng complete hstogrm of n mge n Log(n) steps nd mnmum expected memory requrements usng hypercubes TAREK M. SOBH School of Engneerng, Unversty of Brdgeport, Connectcut, USA. Abstrct Ths work frst revews

More information

Online Appendix to. Mandating Behavioral Conformity in Social Groups with Conformist Members

Online Appendix to. Mandating Behavioral Conformity in Social Groups with Conformist Members Onlne Appendx to Mndtng Behvorl Conformty n Socl Groups wth Conformst Members Peter Grzl Andrze Bnk (Correspondng uthor) Deprtment of Economcs, The Wllms School, Wshngton nd Lee Unversty, Lexngton, 4450

More information

An Introduction to Support Vector Machines

An Introduction to Support Vector Machines An Introducton to Support Vector Mchnes Wht s good Decson Boundry? Consder two-clss, lnerly seprble clssfcton problem Clss How to fnd the lne (or hyperplne n n-dmensons, n>)? Any de? Clss Per Lug Mrtell

More information

Chapter 5 Supplemental Text Material R S T. ij i j ij ijk

Chapter 5 Supplemental Text Material R S T. ij i j ij ijk Chpter 5 Supplementl Text Mterl 5-. Expected Men Squres n the Two-fctor Fctorl Consder the two-fctor fxed effects model y = µ + τ + β + ( τβ) + ε k R S T =,,, =,,, k =,,, n gven s Equton (5-) n the textook.

More information

Dynamic Power Management in a Mobile Multimedia System with Guaranteed Quality-of-Service

Dynamic Power Management in a Mobile Multimedia System with Guaranteed Quality-of-Service Dynmc Power Mngement n Moble Multmed System wth Gurnteed Qulty-of-Servce Qnru Qu, Qng Wu, nd Mssoud Pedrm Dept. of Electrcl Engneerng-Systems Unversty of Southern Clforn Los Angeles CA 90089 Outlne! Introducton

More information

DCDM BUSINESS SCHOOL NUMERICAL METHODS (COS 233-8) Solutions to Assignment 3. x f(x)

DCDM BUSINESS SCHOOL NUMERICAL METHODS (COS 233-8) Solutions to Assignment 3. x f(x) DCDM BUSINESS SCHOOL NUMEICAL METHODS (COS -8) Solutons to Assgnment Queston Consder the followng dt: 5 f() 8 7 5 () Set up dfference tble through fourth dfferences. (b) Wht s the mnmum degree tht n nterpoltng

More information

LOCAL FRACTIONAL LAPLACE SERIES EXPANSION METHOD FOR DIFFUSION EQUATION ARISING IN FRACTAL HEAT TRANSFER

LOCAL FRACTIONAL LAPLACE SERIES EXPANSION METHOD FOR DIFFUSION EQUATION ARISING IN FRACTAL HEAT TRANSFER Yn, S.-P.: Locl Frctonl Lplce Seres Expnson Method for Dffuson THERMAL SCIENCE, Yer 25, Vol. 9, Suppl., pp. S3-S35 S3 LOCAL FRACTIONAL LAPLACE SERIES EXPANSION METHOD FOR DIFFUSION EQUATION ARISING IN

More information

A Family of Multivariate Abel Series Distributions. of Order k

A Family of Multivariate Abel Series Distributions. of Order k Appled Mthemtcl Scences, Vol. 2, 2008, no. 45, 2239-2246 A Fmly of Multvrte Abel Seres Dstrbutons of Order k Rupk Gupt & Kshore K. Ds 2 Fculty of Scence & Technology, The Icf Unversty, Agrtl, Trpur, Ind

More information

B-CaB: Optimizing the SP s Bidding for Cache and Band Resources in Virtualized Wireless Networks

B-CaB: Optimizing the SP s Bidding for Cache and Band Resources in Virtualized Wireless Networks 26 Interntonl Conference on Network nd Informton Systems for Computers B-CB: Optmzng the SP s Bddng for Cche nd Bnd Resources n Vrtulzed Wreless Networks Yuwe YAN Pengbo SI nd Ynhu ZHANG Beng Advnced Innovton

More information

Chapter Newton-Raphson Method of Solving a Nonlinear Equation

Chapter Newton-Raphson Method of Solving a Nonlinear Equation Chpter.4 Newton-Rphson Method of Solvng Nonlner Equton After redng ths chpter, you should be ble to:. derve the Newton-Rphson method formul,. develop the lgorthm of the Newton-Rphson method,. use the Newton-Rphson

More information

6 Roots of Equations: Open Methods

6 Roots of Equations: Open Methods HK Km Slghtly modfed 3//9, /8/6 Frstly wrtten t Mrch 5 6 Roots of Equtons: Open Methods Smple Fed-Pont Iterton Newton-Rphson Secnt Methods MATLAB Functon: fzero Polynomls Cse Study: Ppe Frcton Brcketng

More information

International Journal of Pure and Applied Sciences and Technology

International Journal of Pure and Applied Sciences and Technology Int. J. Pure Appl. Sc. Technol., () (), pp. 44-49 Interntonl Journl of Pure nd Appled Scences nd Technolog ISSN 9-67 Avlle onlne t www.jopst.n Reserch Pper Numercl Soluton for Non-Lner Fredholm Integrl

More information

Linear and Nonlinear Optimization

Linear and Nonlinear Optimization Lner nd Nonlner Optmzton Ynyu Ye Deprtment of Mngement Scence nd Engneerng Stnford Unversty Stnford, CA 9430, U.S.A. http://www.stnford.edu/~yyye http://www.stnford.edu/clss/msnde/ Ynyu Ye, Stnford, MS&E

More information

Quiz: Experimental Physics Lab-I

Quiz: Experimental Physics Lab-I Mxmum Mrks: 18 Totl tme llowed: 35 mn Quz: Expermentl Physcs Lb-I Nme: Roll no: Attempt ll questons. 1. In n experment, bll of mss 100 g s dropped from heght of 65 cm nto the snd contner, the mpct s clled

More information

Model Fitting and Robust Regression Methods

Model Fitting and Robust Regression Methods Dertment o Comuter Engneerng Unverst o Clorn t Snt Cruz Model Fttng nd Robust Regresson Methods CMPE 64: Imge Anlss nd Comuter Vson H o Fttng lnes nd ellses to mge dt Dertment o Comuter Engneerng Unverst

More information

In this Chapter. Chap. 3 Markov chains and hidden Markov models. Probabilistic Models. Example: CpG Islands

In this Chapter. Chap. 3 Markov chains and hidden Markov models. Probabilistic Models. Example: CpG Islands In ths Chpter Chp. 3 Mrov chns nd hdden Mrov models Bontellgence bortory School of Computer Sc. & Eng. Seoul Ntonl Unversty Seoul 5-74, Kore The probblstc model for sequence nlyss HMM (hdden Mrov model)

More information

Decentralized Rate Assignments in a Multi-Sector CDMA Network

Decentralized Rate Assignments in a Multi-Sector CDMA Network 1 Decentrlzed Rte Assgnments n Mult-ector CDMA Network Tr Jvd lectrcl ngneerng, Unversty of Wshngton, ettle tr@ee.wshngton.edu Abstrct We consder wdebnd CDMA network wth rbtrry but known lyout of sectors

More information

Utility function estimation: The entropy approach

Utility function estimation: The entropy approach Physc A 387 (28) 3862 3867 www.elsever.com/locte/phys Utlty functon estmton: The entropy pproch Andre Donso,, A. Hetor Res b,c, Lus Coelho Unversty of Evor, Center of Busness Studes, CEFAGE-UE, Lrgo Colegs,

More information

CS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 9

CS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 9 CS434/541: Pttern Recognton Prof. Olg Veksler Lecture 9 Announcements Fnl project proposl due Nov. 1 1-2 prgrph descrpton Lte Penlt: s 1 pont off for ech d lte Assgnment 3 due November 10 Dt for fnl project

More information

Introduction to Numerical Integration Part II

Introduction to Numerical Integration Part II Introducton to umercl Integrton Prt II CS 75/Mth 75 Brn T. Smth, UM, CS Dept. Sprng, 998 4/9/998 qud_ Intro to Gussn Qudrture s eore, the generl tretment chnges the ntegrton prolem to ndng the ntegrl w

More information

Decoupling Coupled Constraints Through Utility Design

Decoupling Coupled Constraints Through Utility Design IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 59, NO. 8, AUGUST 2014 2289 Decouplng Coupled Constrnts Through Utlty Desgn N L nd Json R. Mrden Abstrct Severl multgent systems exemplfy the need for estblshng

More information

Jens Siebel (University of Applied Sciences Kaiserslautern) An Interactive Introduction to Complex Numbers

Jens Siebel (University of Applied Sciences Kaiserslautern) An Interactive Introduction to Complex Numbers Jens Sebel (Unversty of Appled Scences Kserslutern) An Interctve Introducton to Complex Numbers 1. Introducton We know tht some polynoml equtons do not hve ny solutons on R/. Exmple 1.1: Solve x + 1= for

More information

Multiple view geometry

Multiple view geometry EECS 442 Computer vson Multple vew geometry Perspectve Structure from Moton - Perspectve structure from moton prolem - mgutes - lgerc methods - Fctorzton methods - Bundle djustment - Self-clrton Redng:

More information

INTRODUCTION TO COMPLEX NUMBERS

INTRODUCTION TO COMPLEX NUMBERS INTRODUCTION TO COMPLEX NUMBERS The numers -4, -3, -, -1, 0, 1,, 3, 4 represent the negtve nd postve rel numers termed ntegers. As one frst lerns n mddle school they cn e thought of s unt dstnce spced

More information

Math 497C Sep 17, Curves and Surfaces Fall 2004, PSU

Math 497C Sep 17, Curves and Surfaces Fall 2004, PSU Mth 497C Sep 17, 004 1 Curves nd Surfces Fll 004, PSU Lecture Notes 3 1.8 The generl defnton of curvture; Fox-Mlnor s Theorem Let α: [, b] R n be curve nd P = {t 0,...,t n } be prtton of [, b], then the

More information

523 P a g e. is measured through p. should be slower for lesser values of p and faster for greater values of p. If we set p*

523 P a g e. is measured through p. should be slower for lesser values of p and faster for greater values of p. If we set p* R. Smpth Kumr, R. Kruthk, R. Rdhkrshnn / Interntonl Journl of Engneerng Reserch nd Applctons (IJERA) ISSN: 48-96 www.jer.com Vol., Issue 4, July-August 0, pp.5-58 Constructon Of Mxed Smplng Plns Indexed

More information

CISE 301: Numerical Methods Lecture 5, Topic 4 Least Squares, Curve Fitting

CISE 301: Numerical Methods Lecture 5, Topic 4 Least Squares, Curve Fitting CISE 3: umercl Methods Lecture 5 Topc 4 Lest Squres Curve Fttng Dr. Amr Khouh Term Red Chpter 7 of the tetoo c Khouh CISE3_Topc4_Lest Squre Motvton Gven set of epermentl dt 3 5. 5.9 6.3 The reltonshp etween

More information

Machine Learning Support Vector Machines SVM

Machine Learning Support Vector Machines SVM Mchne Lernng Support Vector Mchnes SVM Lesson 6 Dt Clssfcton problem rnng set:, D,,, : nput dt smple {,, K}: clss or lbel of nput rget: Construct functon f : X Y f, D Predcton of clss for n unknon nput

More information

A Tri-Valued Belief Network Model for Information Retrieval

A Tri-Valued Belief Network Model for Information Retrieval December 200 A Tr-Vlued Belef Networ Model for Informton Retrevl Fernndo Ds-Neves Computer Scence Dept. Vrgn Polytechnc Insttute nd Stte Unversty Blcsburg, VA 24060. IR models t Combnng Evdence Grphcl

More information

Group-based active query selection. for rapid diagnosis in time-critical situations

Group-based active query selection. for rapid diagnosis in time-critical situations Group-bsed ctve query selecton for rpd dgnoss n tme-crtcl stutons *Gowthm Belll, Student Member, IEEE, Suresh K. Bhvnn, nd Clyton Scott, Member, IEEE Abstrct In pplctons such s ctve lernng or dsese/fult

More information

Lecture 36. Finite Element Methods

Lecture 36. Finite Element Methods CE 60: Numercl Methods Lecture 36 Fnte Element Methods Course Coordntor: Dr. Suresh A. Krth, Assocte Professor, Deprtment of Cvl Engneerng, IIT Guwht. In the lst clss, we dscussed on the ppromte methods

More information

CHAPTER - 7. Firefly Algorithm based Strategic Bidding to Maximize Profit of IPPs in Competitive Electricity Market

CHAPTER - 7. Firefly Algorithm based Strategic Bidding to Maximize Profit of IPPs in Competitive Electricity Market CHAPTER - 7 Frefly Algorthm sed Strtegc Bddng to Mxmze Proft of IPPs n Compettve Electrcty Mrket 7. Introducton The renovton of electrc power systems plys mjor role on economc nd relle operton of power

More information

Chapter Newton-Raphson Method of Solving a Nonlinear Equation

Chapter Newton-Raphson Method of Solving a Nonlinear Equation Chpter 0.04 Newton-Rphson Method o Solvng Nonlner Equton Ater redng ths chpter, you should be ble to:. derve the Newton-Rphson method ormul,. develop the lgorthm o the Newton-Rphson method,. use the Newton-Rphson

More information

A New Markov Chain Based Acceptance Sampling Policy via the Minimum Angle Method

A New Markov Chain Based Acceptance Sampling Policy via the Minimum Angle Method Irnn Journl of Opertons Reserch Vol. 3, No., 202, pp. 04- A New Mrkov Chn Bsed Acceptnce Smplng Polcy v the Mnmum Angle Method M. S. Fllh Nezhd * We develop n optmzton model bsed on Mrkovn pproch to determne

More information

CALIBRATION OF SMALL AREA ESTIMATES IN BUSINESS SURVEYS

CALIBRATION OF SMALL AREA ESTIMATES IN BUSINESS SURVEYS CALIBRATION OF SMALL AREA ESTIMATES IN BUSINESS SURVES Rodolphe Prm, Ntle Shlomo Southmpton Sttstcl Scences Reserch Insttute Unverst of Southmpton Unted Kngdom SAE, August 20 The BLUE-ETS Project s fnnced

More information

ON SIMPSON S INEQUALITY AND APPLICATIONS. 1. Introduction The following inequality is well known in the literature as Simpson s inequality : 2 1 f (4)

ON SIMPSON S INEQUALITY AND APPLICATIONS. 1. Introduction The following inequality is well known in the literature as Simpson s inequality : 2 1 f (4) ON SIMPSON S INEQUALITY AND APPLICATIONS SS DRAGOMIR, RP AGARWAL, AND P CERONE Abstrct New neultes of Smpson type nd ther pplcton to udrture formule n Numercl Anlyss re gven Introducton The followng neulty

More information

THE COMBINED SHEPARD ABEL GONCHAROV UNIVARIATE OPERATOR

THE COMBINED SHEPARD ABEL GONCHAROV UNIVARIATE OPERATOR REVUE D ANALYSE NUMÉRIQUE ET DE THÉORIE DE L APPROXIMATION Tome 32, N o 1, 2003, pp 11 20 THE COMBINED SHEPARD ABEL GONCHAROV UNIVARIATE OPERATOR TEODORA CĂTINAŞ Abstrct We extend the Sheprd opertor by

More information

18.7 Artificial Neural Networks

18.7 Artificial Neural Networks 310 18.7 Artfcl Neurl Networks Neuroscence hs hypotheszed tht mentl ctvty conssts prmrly of electrochemcl ctvty n networks of brn cells clled neurons Ths led McCulloch nd Ptts to devse ther mthemtcl model

More information

Improving Anytime Point-Based Value Iteration Using Principled Point Selections

Improving Anytime Point-Based Value Iteration Using Principled Point Selections In In Proceedngs of the Twenteth Interntonl Jont Conference on Artfcl Intellgence (IJCAI-7) Improvng Anytme Pont-Bsed Vlue Iterton Usng Prncpled Pont Selectons Mchel R. Jmes, Mchel E. Smples, nd Dmtr A.

More information

We consider a finite-state, finite-action, infinite-horizon, discounted reward Markov decision process and

We consider a finite-state, finite-action, infinite-horizon, discounted reward Markov decision process and MANAGEMENT SCIENCE Vol. 53, No. 2, Februry 2007, pp. 308 322 ssn 0025-1909 essn 1526-5501 07 5302 0308 nforms do 10.1287/mnsc.1060.0614 2007 INFORMS Bs nd Vrnce Approxmton n Vlue Functon Estmtes She Mnnor

More information

Research Article On the Upper Bounds of Eigenvalues for a Class of Systems of Ordinary Differential Equations with Higher Order

Research Article On the Upper Bounds of Eigenvalues for a Class of Systems of Ordinary Differential Equations with Higher Order Hndw Publshng Corporton Interntonl Journl of Dfferentl Equtons Volume 0, Artcle ID 7703, pges do:055/0/7703 Reserch Artcle On the Upper Bounds of Egenvlues for Clss of Systems of Ordnry Dfferentl Equtons

More information

Least squares. Václav Hlaváč. Czech Technical University in Prague

Least squares. Václav Hlaváč. Czech Technical University in Prague Lest squres Václv Hlváč Czech echncl Unversty n Prgue hlvc@fel.cvut.cz http://cmp.felk.cvut.cz/~hlvc Courtesy: Fred Pghn nd J.P. Lews, SIGGRAPH 2007 Course; Outlne 2 Lner regresson Geometry of lest-squres

More information

arxiv: v2 [cs.lg] 9 Nov 2017

arxiv: v2 [cs.lg] 9 Nov 2017 Renforcement Lernng under Model Msmtch Aurko Roy 1, Hun Xu 2, nd Sebstn Pokutt 2 rxv:1706.04711v2 cs.lg 9 Nov 2017 1 Google Eml: urkor@google.com 2 ISyE, Georg Insttute of Technology, Atlnt, GA, USA. Eml:

More information

Demand. Demand and Comparative Statics. Graphically. Marshallian Demand. ECON 370: Microeconomic Theory Summer 2004 Rice University Stanley Gilbert

Demand. Demand and Comparative Statics. Graphically. Marshallian Demand. ECON 370: Microeconomic Theory Summer 2004 Rice University Stanley Gilbert Demnd Demnd nd Comrtve Sttcs ECON 370: Mcroeconomc Theory Summer 004 Rce Unversty Stnley Glbert Usng the tools we hve develoed u to ths ont, we cn now determne demnd for n ndvdul consumer We seek demnd

More information

Review of linear algebra. Nuno Vasconcelos UCSD

Review of linear algebra. Nuno Vasconcelos UCSD Revew of lner lgebr Nuno Vsconcelos UCSD Vector spces Defnton: vector spce s set H where ddton nd sclr multplcton re defned nd stsf: ) +( + ) (+ )+ 5) λ H 2) + + H 6) 3) H, + 7) λ(λ ) (λλ ) 4) H, - + 8)

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

Formulated Algorithm for Computing Dominant Eigenvalue. and the Corresponding Eigenvector

Formulated Algorithm for Computing Dominant Eigenvalue. and the Corresponding Eigenvector Int. J. Contemp. Mth. Scences Vol. 8 23 no. 9 899-9 HIKARI Ltd www.m-hkr.com http://dx.do.org/.2988/jcms.23.3674 Formulted Algorthm for Computng Domnnt Egenlue nd the Correspondng Egenector Igob Dod Knu

More information

The solution of transport problems by the method of structural optimization

The solution of transport problems by the method of structural optimization Scentfc Journls Mrtme Unversty of Szczecn Zeszyty Nukowe Akdem Morsk w Szczecne 2013, 34(106) pp. 59 64 2013, 34(106) s. 59 64 ISSN 1733-8670 The soluton of trnsport problems by the method of structurl

More information

Simultaneous estimation of rewards and dynamics from noisy expert demonstrations

Simultaneous estimation of rewards and dynamics from noisy expert demonstrations Smultneous estmton of rewrds nd dynmcs from nosy expert demonstrtons Mchel Hermn,2, Tobs Gndele, Jo rg Wgner, Felx Schmtt, nd Wolfrm Burgrd2 - Robert Bosch GmbH - 70442 Stuttgrt - Germny 2- Unversty of

More information

Two Activation Function Wavelet Network for the Identification of Functions with High Nonlinearity

Two Activation Function Wavelet Network for the Identification of Functions with High Nonlinearity Interntonl Journl of Engneerng & Computer Scence IJECS-IJENS Vol:1 No:04 81 Two Actvton Functon Wvelet Network for the Identfcton of Functons wth Hgh Nonlnerty Wsm Khld Abdulkder Abstrct-- The ntegrton

More information

The Dynamic Multi-Task Supply Chain Principal-Agent Analysis

The Dynamic Multi-Task Supply Chain Principal-Agent Analysis J. Servce Scence & Mngement 009 : 9- do:0.46/jssm.009.409 Publshed Onlne December 009 www.scp.org/journl/jssm) 9 he Dynmc Mult-sk Supply Chn Prncpl-Agent Anlyss Shnlng LI Chunhu WANG Dol ZHU Mngement School

More information

Multi-Armed Bandits: Non-adaptive and Adaptive Sampling

Multi-Armed Bandits: Non-adaptive and Adaptive Sampling CSE 547/Stt 548: Mchine Lerning for Big Dt Lecture Multi-Armed Bndits: Non-dptive nd Adptive Smpling Instructor: Shm Kkde 1 The (stochstic) multi-rmed bndit problem The bsic prdigm is s follows: K Independent

More information

Optimal Resource Allocation and Policy Formulation in Loosely-Coupled Markov Decision Processes

Optimal Resource Allocation and Policy Formulation in Loosely-Coupled Markov Decision Processes Optml Resource Allocton nd Polcy Formulton n Loosely-Coupled Mrkov Decson Processes Dmtr A. Dolgov nd Edmund H. Durfee Deprtment of Electrcl Engneerng nd Computer Scence Unversty of Mchgn Ann Arbor, MI

More information

Investigation phase in case of Bragg coupling

Investigation phase in case of Bragg coupling Journl of Th-Qr Unversty No.3 Vol.4 December/008 Investgton phse n cse of Brgg couplng Hder K. Mouhmd Deprtment of Physcs, College of Scence, Th-Qr, Unv. Mouhmd H. Abdullh Deprtment of Physcs, College

More information

LAPLACE TRANSFORM SOLUTION OF THE PROBLEM OF TIME-FRACTIONAL HEAT CONDUCTION IN A TWO-LAYERED SLAB

LAPLACE TRANSFORM SOLUTION OF THE PROBLEM OF TIME-FRACTIONAL HEAT CONDUCTION IN A TWO-LAYERED SLAB Journl of Appled Mthemtcs nd Computtonl Mechncs 5, 4(4), 5-3 www.mcm.pcz.pl p-issn 99-9965 DOI:.75/jmcm.5.4. e-issn 353-588 LAPLACE TRANSFORM SOLUTION OF THE PROBLEM OF TIME-FRACTIONAL HEAT CONDUCTION

More information

A Reinforcement Learning System with Chaotic Neural Networks-Based Adaptive Hierarchical Memory Structure for Autonomous Robots

A Reinforcement Learning System with Chaotic Neural Networks-Based Adaptive Hierarchical Memory Structure for Autonomous Robots Interntonl Conference on Control, Automton nd ystems 008 Oct. 4-7, 008 n COEX, eoul, Kore A Renforcement ernng ystem wth Chotc Neurl Networs-Bsed Adptve Herrchcl Memory tructure for Autonomous Robots Msno

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

Dynamic Power Management in a Mobile Multimedia System with Guaranteed Quality-of-Service

Dynamic Power Management in a Mobile Multimedia System with Guaranteed Quality-of-Service Dynmc Power Mngement n Moble Multmed System wth Gurnteed Qulty-of-Servce Abstrct In ths pper we ddress the problem of dynmc power mngement n dstrbuted multmed system wth requred qulty of servce (QoS).

More information

Audio De-noising Analysis Using Diagonal and Non-Diagonal Estimation Techniques

Audio De-noising Analysis Using Diagonal and Non-Diagonal Estimation Techniques Audo De-nosng Anlyss Usng Dgonl nd Non-Dgonl Estmton Technques Sugt R. Pwr 1, Vshl U. Gdero 2, nd Rhul N. Jdhv 3 1 AISSMS, IOIT, Pune, Ind Eml: sugtpwr@gml.com 2 Govt Polytechnque, Pune, Ind Eml: vshl.gdero@gml.com

More information

Bi-level models for OD matrix estimation

Bi-level models for OD matrix estimation TNK084 Trffc Theory seres Vol.4, number. My 2008 B-level models for OD mtrx estmton Hn Zhng, Quyng Meng Abstrct- Ths pper ntroduces two types of O/D mtrx estmton model: ME2 nd Grdent. ME2 s mxmum-entropy

More information

Lecture Notes on Linear Regression

Lecture Notes on Linear Regression Lecture Notes on Lnear Regresson Feng L fl@sdueducn Shandong Unversty, Chna Lnear Regresson Problem In regresson problem, we am at predct a contnuous target value gven an nput feature vector We assume

More information

Reinforcement Learning with a Gaussian Mixture Model

Reinforcement Learning with a Gaussian Mixture Model Renforcement Lernng wth Gussn Mxture Model Alejndro Agostn, Member, IEEE nd Enrc Cely Abstrct Recent pproches to Renforcement Lernng (RL) wth functon pproxmton nclude Neurl Ftted Q Iterton nd the use of

More information

8. INVERSE Z-TRANSFORM

8. INVERSE Z-TRANSFORM 8. INVERSE Z-TRANSFORM The proce by whch Z-trnform of tme ere, nmely X(), returned to the tme domn clled the nvere Z-trnform. The nvere Z-trnform defned by: Computer tudy Z X M-fle trn.m ued to fnd nvere

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

The Number of Rows which Equal Certain Row

The Number of Rows which Equal Certain Row Interntonl Journl of Algebr, Vol 5, 011, no 30, 1481-1488 he Number of Rows whch Equl Certn Row Ahmd Hbl Deprtment of mthemtcs Fcult of Scences Dmscus unverst Dmscus, Sr hblhmd1@gmlcom Abstrct Let be X

More information

4.4 Areas, Integrals and Antiderivatives

4.4 Areas, Integrals and Antiderivatives . res, integrls nd ntiderivtives 333. Ares, Integrls nd Antiderivtives This section explores properties of functions defined s res nd exmines some connections mong res, integrls nd ntiderivtives. In order

More information

Averaging Level Control of Multiple Tanks: A Passivity Based Approach

Averaging Level Control of Multiple Tanks: A Passivity Based Approach Avergng Level Control of Mltple Tnks: A Pssvty Bse Approch Dnel Sbrbro Deprtment of Electrcl Engneerng, Unvers e Concepcón, Concepcón,, Chle. In collborton wth Romeo Orteg Lbortore es Sgn et Systèmes mes,

More information

Chemical Reaction Engineering

Chemical Reaction Engineering Lecture 20 hemcl Recton Engneerng (RE) s the feld tht studes the rtes nd mechnsms of chemcl rectons nd the desgn of the rectors n whch they tke plce. Lst Lecture Energy Blnce Fundmentls F 0 E 0 F E Q W

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

PLEASE SCROLL DOWN FOR ARTICLE

PLEASE SCROLL DOWN FOR ARTICLE Ths rtcle ws downloded by:ntonl Cheng Kung Unversty] On: 1 September 7 Access Detls: subscrpton number 7765748] Publsher: Tylor & Frncs Inform Ltd Regstered n Englnd nd Wles Regstered Number: 17954 Regstered

More information

HAMILTON-JACOBI TREATMENT OF LAGRANGIAN WITH FERMIONIC AND SCALAR FIELD

HAMILTON-JACOBI TREATMENT OF LAGRANGIAN WITH FERMIONIC AND SCALAR FIELD AMION-JACOBI REAMEN OF AGRANGIAN WI FERMIONIC AND SCAAR FIED W. I. ESRAIM 1, N. I. FARAA Dertment of Physcs, Islmc Unversty of Gz, P.O. Box 18, Gz, Plestne 1 wbrhm 7@hotml.com nfrht@ugz.edu.s Receved November,

More information

Proof that if Voting is Perfect in One Dimension, then the First. Eigenvector Extracted from the Double-Centered Transformed

Proof that if Voting is Perfect in One Dimension, then the First. Eigenvector Extracted from the Double-Centered Transformed Proof tht f Votng s Perfect n One Dmenson, then the Frst Egenvector Extrcted from the Doule-Centered Trnsformed Agreement Score Mtrx hs the Sme Rn Orderng s the True Dt Keth T Poole Unversty of Houston

More information

Solution of Tutorial 5 Drive dynamics & control

Solution of Tutorial 5 Drive dynamics & control ELEC463 Unversty of New South Wles School of Electrcl Engneerng & elecommunctons ELEC463 Electrc Drve Systems Queston Motor Soluton of utorl 5 Drve dynmcs & control 500 rev/mn = 5.3 rd/s 750 rted 4.3 Nm

More information

Pyramid Algorithms for Barycentric Rational Interpolation

Pyramid Algorithms for Barycentric Rational Interpolation Pyrmd Algorthms for Brycentrc Rtonl Interpolton K Hormnn Scott Schefer Astrct We present new perspectve on the Floter Hormnn nterpolnt. Ths nterpolnt s rtonl of degree (n, d), reproduces polynomls of degree

More information

SVMs for regression Non-parametric/instance based classification method

SVMs for regression Non-parametric/instance based classification method S 75 Mchne ernng ecture Mos Huskrecht mos@cs.ptt.edu 539 Sennott Squre SVMs for regresson Non-prmetrc/nstnce sed cssfcton method S 75 Mchne ernng Soft-mrgn SVM Aos some fet on crossng the seprtng hperpne

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

Electrochemical Thermodynamics. Interfaces and Energy Conversion

Electrochemical Thermodynamics. Interfaces and Energy Conversion CHE465/865, 2006-3, Lecture 6, 18 th Sep., 2006 Electrochemcl Thermodynmcs Interfces nd Energy Converson Where does the energy contrbuton F zϕ dn come from? Frst lw of thermodynmcs (conservton of energy):

More information