Social Coordination in Unknown Price-Sensitive Populations
|
|
- Isabella Stewart
- 5 years ago
- Views:
Transcription
1 Socal Coordnaton n Unknown Prce-Senstve Populatons Phlp N. Brown and Jason R. Marden 2 Abstract In ths paper, we nvestgate the relatonshp between uncertanty and a desgner s ablty to nfluence socal behavor. Pgovan taxes are a common approach to socal coordnaton. However, guaranteeng effcent behavor typcally requres that the system desgner has complete knowledge of the user populaton s senstvty to taxaton. In ths paper, we explore the effect of relaxng ths requrement n the context of congeston games wth affne costs. Focusng on the class of scaled Pgovan taxes, we derve the optmal tollng scheme that mnmzes the worst-case effcency loss under uncertanty n user senstvty. Furthermore, we derve explct bounds whch hghlght how the level of uncertanty n senstvty degrades performance. I. INTRODUCTION As engneered systems and ther users become ncreasngly nterconnected, engneers must consder not only mere system desgn, but also effcent system utlzaton []. It s wdely known that unnfluenced socal systems often exhbt suboptmal behavor. Ths s a well-researched topc n many felds, ncludng electrc power markets [2], communcaton spectrum sharng [3], and traffc congeston [4]. Accordngly, a central desgn challenge s to nfluence socal behavor through admssble mechansms n these systems to promote effcent outcomes. To analyze the role of socal coordnaton n engneered systems, we model the behavor of a system s users as a game between the users, n whch each user chooses hs actons to maxmze hs personal beneft. Equlbrum behavor establshed by every user actng n ths way s known as a Nash equlbrum. It s wdely known that Nash equlbra need not be effcent n any global sense. It s common to quantfy the effcency losses nherent n Nash equlbra by a metrc known as the prce of anarchy of a system, defned as the worst-possble rato between the effcency of a Nash equlbrum behavor and that of an optmal behavor [5], [6]. The prce of anarchy has been analyzed for a varety of engneered systems [7], [8], [9]. Ths leads to the central queston: how may socal behavor be nfluenced to reduce the prce of anarchy? A common methodology for socal coordnaton s known as a Pgovan tax [0]. Under a Pgovan tax, users are charged a tax equal to the negatve effect they have on other users; thus, each user s ncentvzed to choose actons that Ths research was supported by AFOSR grants #FA and #FA and by ONR grant #N P. N. Brown s a graduate research assstant wth the Department of Electrcal, Computer, and Energy Engneerng, Unversty of Colorado, Boulder, CO 80309, phlp.brown@colorado.edu. J. R. Marden s wth the Department of Electrcal, Computer, and Energy Engneerng, Unversty of Colorado, Boulder, CO 80309, jason.marden@colorado.edu. Correspondng author. maxmze the publc beneft. In a general sense, Pgovan taxes are attractve snce they guarantee that Nash equlbrum behavor exactly equals system optmal behavor. Ths type of taxaton strategy has been suggested as a soluton to ndustral polluton [], harmful speculatve tradng n fnancal markets [2], and road congeston [3], to name a few. To llustrate a socal coordnaton settng n whch Pgovan taxes apply, we turn to the problem of road congeston. It s well-known that Nash equlbra n routng games are often neffcent [4], and a central research focus has been to nfluence routng behavor usng tolls to promote effcent routng. In the congeston settng, a Pgovan tax s known as a margnal-cost toll or congeston toll; each drver s charged a toll equvalent to the congeston caused by an addtonal unt of traffc [5], [6], [7]. An nherent challenge s that congeston s experenced n unts of tme, but tolls are charged n unts of money. If the desgner lacks perfect nformaton regardng drvers prce-senstvty, t wll not n general be possble for the desgner to levy a perfect toll. Recent research has sdestepped ths ssue by fxng all system varables and dervng a tollng strategy that optmzes a fxed nstance of the system [8]. It has been shown that under a general class of congeston games, ths type of fxed tollng can optmze Nash Equlbra [9]. Whle these results represent successes, they rely strongly on a complete characterzaton of the network structure, user capabltes, and prce senstvtes. Accordngly, t may be that small varatons n any of these parameters could lead to hghly suboptmal behavor; uncertanty may dramatcally reduce a desgner s ablty to nfluence socal behavor and could n general lead to a degradaton n effcency from the unnfluenced case. Lttle research has been done to characterze the effect of ths uncertanty on the ablty to coordnate a socal system. Wth these lmtatons n mnd, we ntate a study on the relatonshp between nformaton and a desgner s ablty to nfluence socal behavor. More concsely, we nvestgate uncertan user prce-senstvty for a routng problem n a class of congeston games: the set of network routng games such that each edge n a network possesses a congeston functon that s affne n the number of edge users. Congeston games of ths type have been extensvely studed n the exstng lterature [20], [2], [4]. For ths class of congeston games, we develop a dstrbuted tollng scheme that ncreases routng effcency for any bounded uncertanty of the user populaton s prcesenstvty. We show that under certan common utlzaton condtons, provded that the populaton s prce-senstvty s
2 range-restrcted, scaled margnal-cost tolls always ncrease effcency. Furthermore, we show that the unque margnalcost toll scale factor that mnmzes the prce of anarchy s the geometrc mean of the upper and lower bounds of the populaton s prce senstvty. That s, our toll on a road s a functon only of that road s congeston propertes and the rato of the populaton s senstvty bounds t s ndependent of the traffc flow rate and the network topology. We also provde a tght bound on the prce of anarchy for games usng our specfed tollng methodology. Ths prce of anarchy depends only on the rato of the user populaton s senstvty bounds; t s ndependent of network cost functons, network topology, and overall traffc rate. Thus, for a broad class of games, we concsely characterze the relatonshp between nformaton and effcency. A. Network Defnton II. MODEL Consder a network (V, E, whch conssts of a vertex set V and edge set E (V V. Each edge e E s assocated wth a latency functon of the affne form l e (f e = d e f e + c e, ( where f e desgnates the flow on edge e. In ths paper, we wll focus on characterzng the effcency of flows nduced by ndvduals strategc behavor. Specfcally, we focus on the settng n whch there s a populaton of users seekng to traverse the network from a common source to a common destnaton. Consequently, each user chooses ts path from a common set of paths P = {p, p 2,..., p n } where each path p P conssts of a set of edges,.e., P 2 E. We defne the total mass of the populaton to be r 0. Usng these defntons, let a game G be defned by the tuple G = (V, E, P, {l e } e E, r, and let G represent the class of all such games G. B. Network Flows For a game G, the set of feasble flows f(g = (f p, f p2,..., f pn s characterzed by F (G = {f p} p P : f p 0 and f p = r, p P where f p desgnates the flow on path p. For a gven flow f(g F (G, the latency l p (f(g of any path p P s the sum of the latences of edges n that path: l p (f(g = e p l e (f e, (2 where f e s the flow on edge e gven f(g,.e., f e = f p. (3 p P:e p Thus, the total latency on the network s the flow-weghted sum of all path latences: L(f(G = p P f p l p (f p. (4 Users are concerned wth mnmzng ther latences, so the cost J p ( to a user of any path p P s the path latency: J p (f(g = e p l e (f e. (5 We focus n partcular on the settng n whch each user chooses the path wth the lowest cost. A routng obtaned n ths way s known as a Nash flow [4], whch corresponds to a feasble flow f ne (G F (G such that for all p, p j P > 0, j > 0 = J p (f ne (G = J pj (f ne (G, (6 > 0, j = 0 = J p (f ne (G J pj (f ne (G. (7 We defne an optmal flow on G as f opt (G = nf L(f(G. f(g F (G Then we defne the prce of anarchy of G as L(f ne (G : G PoA(G = sup G G L(f opt (G C. Tollng for Known Homogeneous Populatons The network owner, wshng to enforce optmal routng, charges a toll τ e (f e on each lnk n the network, addng to the lnk user s cost. A Pgovan toll, also known as a margnal-cost toll, s of the form τ e (f e = d e f e. In general, these margnal-cost tolls nduce optmal routng. Homogenety s a common assumpton when analyzng worst-case behavor [2]; t s a smple approach that can often be used to bound the behavor of heterogeneous populatons. The clear beneft of tollng functons of the Pgovan form s that they can be assgned n a dstrbuted manner; each edge toll s a functon only of the congeston propertes of that edge. For latency functons of the form n (, Pgovan tolls modfy (5 to gve a path cost of J p (f(g = e p (8 [2d e (f e + c e ]. (9 In [4], t s shown that cost functons of ths form always nduce optmal behavor. D. Tollng for Unknown Homogeneous Populatons In real systems, t s often true that users exhbt prcesenstvtes that are unknown or that vary from day to day. We model ths prce senstvty wth a parameter β [β L, β U ]. Intutvely, hgh β mples hgh averson to toll. Incorporatng prce senstvty, we represent the cost of a path generally as J p (f(g = e p [l e (f e + βτ e (f e ]. (0 In the case when β s known precsely, an optmal margnal-cost toll s of the form τ e (f e = d e f e /β; wth ths toll, (0 clearly smplfes to (9. We desre to explot the dstrbuted nature of margnal-cost tolls, so we nvestgate tollng functons of the form τ e (f e = µd e f e, where µ 0 s a common factor for every edge, and scales for our estmate of the true β of the populaton. We call tolls of ths form
3 scaled margnal-cost tolls. These tolls further modfy (5 to gve the fnal form of our path cost functon J p (f(g = e p [( + βµd e f e + c e ]. ( These modfed cost functons nduce a new Nash flow whch we denote f ne (µ : G, β. For smplcty of notaton, we defne the total latency of ths Nash flow under scaled margnal-cost tolls as L ne (µ : G, β = L(f ne (µ : G, β, (2 Where L(f( s defned as n (4. Note that µ s our control parameter; G and β represent the system and populaton we are attemptng to control. Therefore, gven a game G, we seek to characterze the scaled margnal-cost tolls whch mnmze the worst-case total latency for G gven a range of user prce senstvtes. That s, we seek to characterze solutons µ to the optmzaton problem ( L ne (µ : G, β = nf µ 0 sup (L ne (µ : G, β β [β L,β U] III. MAIN RESULT. (3 We show that under reasonable utlzaton constrants, the soluton to the optmzaton problem defned n (3 s ndependent of the underlyng game G. Defnton 3.: A game G = (V, E, P, {l e } e E, r s fully-utlzed f r r, p P, (µ = 0 : G, β > 0. That s, all paths are n use n an un-tolled Nash flow, and ncreasng r wll not cause any path flow to decrease to zero. In Secton V-A we show that ths defnton apples to a broad class of games. Theorem 3.2: Consder any game G that satsfes Defnton 3.. Let µ =. (4 βl β U Then µ satsfes L ne (µ : G, β = nf µ 0 ( sup (L ne (µ : G, β β [β L,β U] Furthermore, let G represent the class of all games G satsfyng Defnton 3.. For β [β L, β U ], 4 ( + β L /β U + β L /β U PoA (µ : G, β = 3 ( + 2. (5 β L /β U Thus, µ ensures that the worst-case total latency of a game s lmted to a mnmum for any range-restrcted β. Note that when β L / = 0, meanng that t s not possble to nfluence behavor wth tolls, we obtan the well-known prce of anarchy of 4/3 derved n [2]. On the other hand, f β L = β U, meanng that the prce-senstvty of all users s known precsely, we obtan a prce of anarchy of, mplyng that complete nformaton allows complete control. Note also that µ depends only on β L and β U ; t s ndependent of G. Ths allows us to defne a dstrbuted rule. for our edge tolls so that we can set each edge s toll wthout knowledge of the overall network topology, and guarantees a reducton n total latency from the un-tolled network. IV. PROOF OF RESULTS The proof proceeds as follows: In Lemma 4. we show that n games satsfyng Defnton 3., the flow on any path can be decomposed as the sum of a term dependng on r and a term dependng on µ. 2 In Lemma 4.2 we show that for such games, mples that scaled margnal-cost tolls always decrease latency by a quantty that s ndependent of the traffc rate r. 3 Lastly, we show that the propertes proved n 2 lead to a characterzaton of the unque margnal-cost toll scale factor that mnmzes worst-case total latency for uncertan senstvty. We then derve the prce of anarchy under our optmal margnal-cost tolls. For smplcty, we wll henceforth express (µ : G, β as merely fp ne when the dependence on µ, G, and β s clear. Our proof utlzes the followng defnton: for any p, p j P, { e p d j = p j d e, p p j 0, p p j =. (6 Intutvely, d j s the amount that affects J pj (f ne. Note that d j = d j. Lemma 4.: For any game G = (V, E, P, {l e } e E, r satsfyng Defnton 3., for any µ 0, the Nash flow on each path p P s of the form = rr + + µβ M, (7 where the coeffcents R R and M R have no dependence on µ, r, or β, and satsfy the followng equatons: R =, (8 M = 0, (9 R k d k = R k d jk, (20 M k d k + c e = M k d jk + c e, (2 e p e p j where p, p j, p k P. Proof: Frst, we the decomposton of the path flows. Then we show that due to the propertes nduced by Defnton 3., the coeffcents defned n Lemma 4. satsfy (8- (2. Combnng ( and (3, the cost of path p s J p (f ne = ( ( ( + βµd e + c e. (22 e p Defne d e (p n the followng way: k p k :e p k d e (p = { de, e p 0, e / p.
4 Now, (22 can be wrtten as J p (f ne = ( + βµ e p = ( + βµ = ( + βµ f ne p k d e (p k + de (p + e p c e c e f ne p k d k + e p c e. (23 Snce f ne s a Nash flow and every path flow s postve, by (6 all path costs are equal, so p, p j P, k (d k d jk = c e c e. (24 + βµ e pj e p Thus, each {, j} defnes an equaton of the form f p ne k a j = b j. We take any dstnct P of these = r, and we have a P -dmensonal system of lnear equatons, whch can be descrbed by the followng matrx equaton: equatons and combne them wth f ne p k Af ne = b. 2 (25 A soluton to (25 represents a Nash flow for (G, β, µ, and n [22] t s shown Nash flows always exst for congeston games of the type consdered n ths paper. Snce b s of the form b /( + βµ b =. b P /( + βµ r, (26 Any soluton f ne must be a lnear combnaton of /(+βµ and r. 3 That s, p P, can be represented as shown n (7. Note that {R } and {M } have no dependence on µ, r, or β. Clearly, snce G satsfes Defnton 3., R must be nonnegatve for every path p P. Note also that µ > 0, (µ : G, β > 0 snce (0 : G, β > 0. That s, (7 s robust to ncreases n tolls. Next, we derve two mportant propertes of R and M. Substtutng (7 nto p f ne P p = r, we obtan r R + M = 0. (27 + µβ Snce (27 must hold for any arbtrarly large r and for all µ 0, we obtan the proof of (8 and (9: R =, M = 0. Note that {, j} s not dstnct from {j, }. 2 Wth abuse of notaton, here we represent the tuple of path flows as a column vector. 3 To see ths, perform Gaussan elmnaton on the augmented matrx [A b] to solve for f. Note that matrx row operatons only nvolve sums of elements of b, never products. Now, substtutng (7 nto (24 and smplfyng, we obtan r R k (d k d jk + + µβ M k (d k d jk + c e c e = 0. e p j e p (28 Smlarly to above, snce (28 must hold for any arbtrarly large r and for all µ 0, we obtan the proof of (20 and (2: R k d k = R k d j, M k d k + c e = M k d jk + c e, e p e p j Whch completes the proof. Lemma 4.2: For any game G = (V, E, P, {l e } e E, r satsfyng Defnton 3., there exsts a functon L r (r dependng only on r, and a constant K µ 0, such that L ne µβ (µ : G, β = L r (r ( + µβ 2 K µ. (29 Here, we show for total latency a smlar qualty to that expressed for flows n Lemma 4.; namely, rate and tolls have ndependent effects on the total latency. Ths s a crucallymportant property of games satsfyng Defnton 3., snce t allows us to defne tollng rules wthout knowledge of r. Proof: Frst, we show the decomposton of the total latency of a Nash flow. We then show that the total latency of every Nash flow s of the specfc form shown n Lemma 4.2. Substtute (7 nto (4 and smplfy to obtan L ne (µ : G, β = L r (r + r ζ + η, (30 + µβ where L r (r = r 2 d j R R j + r c e R, (3 e p ζ = η = d j (R j M + R M j, d j ( + µβ 2 M M j + c e e p + µβ M. (32 Note that the second term n (30 depends on both r and µ. The followng argument shows that ζ = 0. By factorng and reversng sum order, we obtan ζ = M R j d j + M j R d j. Choose any p k P. By Lemma 4., p, p k P, R jd j = R jd kj, and analogously, p j, p k
5 P, p R P d j = p R P d k. Then ζ can be wrtten as ζ = R j d kj + R d k. M M j By Lemma 4., we know that each term n parentheses equals 0, so ζ = 0. Ths allows us to express the total latency as the sum of a functon dependent on r, and a functon dependent on µ and β. Thus, we see here the decouplng of the effects of rate and tolls. Consder now the /( + µβ 2 term of (32 and denote t K µ : K µ = M d j M j. (33 By Lemma 4., p, p k P, we know that d j M j = d kj M j + c e c e. e p k e p Accordngly, we can wrte K µ as M d jk M j + c e M c e. e p k e p By Lemma 4. we know that p M P = 0. Therefore K µ s smply K µ = c e M. e We can then express (30 as ( L ne (µ : G, β = L r (r + ( + µβ 2 ( + µβ = L r (r K µ µβ ( + µβ 2 K µ. (34 To complete the proof of the lemma, we show that K µ 0. Equaton (34 clearly shows that L ne (0 : G, β = L r (r. (35 That s, L r (r represents the total latency of the un-tolled system. Equatons (34 and (35 show that for any game G, f some toll decreases latency, then all tolls must decrease latency. Snce µ = /β always results n optmal routng 4 (.e., a reducton n latency, t must be true that for any G, K µ 0. Thus, scaled margnal-cost tolls always decrease total latency. Ths property of the total latency of a Nash flow lets us concsely quantfy the effect of tolls, and leads drectly to our proof of the optmal margnal-cost toll scale factor. Proof of Theorem 3.2. Consder the partal dervatve of (29 wth respect to β: L ne (µ : G, β β 4 See (9 and ( n Secton II. ( = µβ β ( + µβ 2 K µ ( µ(µβ = ( + µβ 3 K µ. (36 From (36, t can be seen that L ne (µ : G, β has a mnmum at β = /µ, so the end-ponts of the β-range gve the worst possble latences: sup {L ne (µ:g,β}=max{l ne (µ:g,β L,L ne (µ:g,β U }. β [β L,β U] Consder now the partal dervatve of L ne (G, β, µ wth respect to µ: L ne ( (µ : G, β β(µβ = µ ( + µβ 3 K µ. (37 Clearly, for any β > 0, L ne (µ : G, β has a mnmum at µ = /β. Snce L ne (/β : G, β = L(f opt (G, t must be true that L ne (/β L : G, β L = L ne (/β U : G, β U. Thus, snce L ne (µ : G, β s contnuous n µ, there must exst some µ [/β U, /β L ] such that L ne (µ : G, β L = L ne (µ : G, β U. It can easly be verfed from (29 that µ = βl β U. (38 Now we bound the neffcency of a game G under tolls as defned n (38. Snce we know that an un-tolled latency can never be more than 4/3 tmes an optmal latency, from (29 we can wrte L ne (0 : G, β L r (r L opt = (G L r (r 4 K µ Ths allows us to compute a bound on K µ : 4 3. (39 K µ L r (r. (40 It follows algebracally that for µ as defned n (38, β [β L, β U ], and G as defned n the statement of the theorem, 4 ( + β L /β U + β L /β U PoA(µ : G, β = 3 ( + 2. β L /β U We show that ths bound s tght n Secton V-B wth Pgou s Example. A. A Note on Full Utlzaton V. DISCUSSION To see that Defnton 3. apples to a broad class of networks, consder a smple network of n parallel lnks, wth latency functons of the form specfed n (. Due to the smplcty of the network, P = E. Let Ḡ = (V, E, P, {l e } e E, r. Suppose that every path p carres postve rate (Ḡ n a Nash flow. That s, all lnks have equal cost. Now, let G = (V, E, P, {l e } e E, r,
6 where r > r. That s, our only change from Ḡ to G s that we ncreased the traffc rate for G. By ncreasng the overall traffc rate, there must be some path p j such that j (G > f ne (Ḡ. Thus, p j l f = J pj (f ne p j (G J pj (f ne p j (Ḡ. Ths mples that for any path p p j, J p (f ne p (G J pj (f ne p j (Ḡ = J p (f ne p (Ḡ. Snce cost functons are nondecreasng, ths mples that (G (Ḡ. That s, no path flow decreased; thus, all networks of n parallel lnks such that all lnks carry postve flow satsfy Defnton 3.. There are many ways to extend ths class of networks; for example, smple arguments can show that any combnaton of parallel networks n seres wth other parallel networks also satsfes Defnton 3.. B. Pgou s Example To demonstrate these results, let us turn to a smple canoncal network routng problem known as Pgou s Example [20], llustrated n Fgure. In ths example, traffc has access to two routes: the frst, a costly constant-latency route; the second, a congeston-senstve route. Let r =. Wthout tolls, all traffc prefers route 2 snce no user can mprove her latency by usng route, and the total latency L ne (0 : G, β =. 5 However, consder the optmal routng: the traffc splts evenly between the routes so that f p = f p2 = 0.5; now half the populaton experences a latency of 0.5 and half, so the total latency s L(f opt (G = 3/4, a 25% mprovement. We know that chargng a toll equal to f 2 (correspondng to µ = /β; see (9 and ( n Secton II on route 2 wll enforce ths optmal flow. Now assume that all users share some value of β [β L, β U ]. Consder for example the case when β = β L. If our toll s f 2 /β L, the routng wll stll be optmal. However, for β > β L, a dsproportonate number of users wll dvert to route. Now, the Nash flow nduced by µ > 0 s such that f p = β/(µ + β and f p2 = µ/(µ + β. Thus the total latency s ( 2 L ne β µ (µ : G, β = (µ + β + µ + β = (µ2 + β(µ + β (µ + β 2. (4 It s easy to verfy that for µ = µ = (β L β U /2 as prescrbed by Theorem 3.2 and β = β L, the total latency s L ne (µ : G, β L = ( + β L /β U + β L /β U ( + β L /β U 2. (42 Snce L(f opt (G = 3/4, the result n (42 proves the tghtness of the prce of anarchy specfed n Theorem 3.2. l 2 f 2 = f 2 Fg.. Pgou s Example. Network used n Secton V-B to demonstrate the prce of anarchy for scaled margnal-cost tolls and uncertan user populatons. β µ τ 2 f p f p2 L ne (µ : G, β β L = f p β U = 9 9f p β L = /3 (/3f p β U = 9 /3 3f p TABLE I TOLLS AND TOTAL LATENCY IN PIGOU S EXAMPLE Table I presents some numercal examples. In partcular, f β [, 9], then for µ = µ = /3, the total latency wll never be worse than 0.825, nearly a 9% mprovement over the un-tolled case. C. Braess Topology To llustrate a network that may not satsfy Defnton 3., we turn to a second canoncal example: that of Braess s Paradox [23], [4]. The topology n Braess s Paradox has been used to show that addng lnks to a graph can ncrease the cost of a Nash flow; we use t to show that ncreasng the total rate of flow can cause an ndvdual path flow to decrease. Consder the network n Fgure 2. Note that n the canoncal Braess s Paradox network, d 2 = d 3 = 0. We consder two varants of the network: n Varant, d 2 = d 3 = 0; n Varant 2, d 2 =, d 3 = 2. Traffc routes from node A to node D, and has a choce of three paths. Defne p = {e, e 3 }, p 2 = {e, e 5, e 4 }, and p 3 = {e 2, e 4 }. The values of R (refer to Lemma 4. n the Proof sectons for each varant appear n Table II. For Varant, R 2 =. Ths means that ncreasng the total traffc rate wll actually decrease the amount of traffc usng path 2, whch s to say edge 5 (refer to Lemma 4.. Thus ths network never satsfes Defnton 3., and our optmal toll analyss s not necessarly vald. 5 The astute reader wll notce that ths example does not strctly satsfy Defnton 3.. Ths wll not adversely mpact our analyss, because we may consder the case n whch r = +ɛ for ɛ > 0. Then f p = ɛ and f p2 =, so Defnton 3. s satsfed. We take the lmt as ɛ 0, and all analyss proceeds as before.
7 .4.35 Prce of Anarchy Prce of Anarchy 4/3.05 Fg. 2. Braess s Topology. For d 2 = d 3 = 0, no game usng ths topology satsfes Defnton 3.. However, for d 2 =, d 3 = 2, games usng ths topology can satsfy Defnton 3.. Varant R R 2 R 3 2 /3 /6 /2 TABLE II VALUES OF R FOR BRAESS TOPOLOGY For Varant 2, the topology s unchanged, but the modfed latency functons have changed routng behavor dramatcally. Now R 0 for every path, so all path flows are ncreasng n r, and thus f all paths are n use, the game satsfes Defnton 3.. Thus, scaled margnal-cost tolls can never degrade network performance. D. Prce of Anarchy Our expresson for the prce of anarchy n Theorem 3.2 succnctly characterzes the relatonshp between effcency and nformaton about the user populaton. It llustrates the ntutve prncple that wth perfect nformaton, a desgner can acheve perfect effcency. Fgure 3 shows how the prce of anarchy vares wth β L /β U. Note that for any value of ths rato, the prce of anarchy s less than 4/3, that s, less than that of an un-tolled network. Though mperfect nformaton does degrade effcency, the fgure shows that hgh levels of effcency are mantaned even for a relatvely hgh level of uncertanty. Note that even when β s only known wthn a factor of 0, the effcency loss s less than 0%. Our prce of anarchy result s encouragng for two reasons. Frst, ths worst-case analyss was obtaned from assumng that the entre populaton les at one of the extremes of the range. Ths assumpton, though mportant to establsh the bound, s lkely to be unrealstc for most systems. Any β closer to our geometrc mean tollng parameter wll yeld even better performance than the prce of anarchy suggests. Ths s llustrated n Fgure 4; the fgure was generated usng values from Pgou s Example n (see Secton V-B wth r =, β L =, β U = 9, and µ = /3. The horzontal axs represents the true β of the populaton; thus t s easy to see Rato of mnmum and maxmum betas Fg. 3. Prce of anarchy for games usng our scaled margnal-cost tollng rule, as a functon of the precson wth whch we know the user populaton s prce-senstvty. Effcency Loss User Populaton s True Beta Fg. 4. Effcency Loss n Pgou s Example wth respect to true beta for µ = /3 and r =. Note that the effcency loss s equal at the endponts of the β-range, and perfect effcency s acheved at the geometrc mean of the endponts. that values close to the geometrc mean allow only very low effcency loss. Second, the analyss contans a strong assumpton of homogenety; all users are requred to have dentcal values of β. Ths s also unlkely n real systems. Consder a populaton whose β-values are dstrbuted over some range. More rgorous analyss s needed, but our smulatons have suggested that f a greater mass of users are closer to the geometrc mean than to the endponts of the range, tolls would acheve consderably mproved performance. VI. CONCLUSIONS Effectve desgn for socal coordnaton requres nformaton about a system s user populaton; we have provded ntal results on the relatonshp between the qualty of ths nformaton and the effcency of Nash equlbra resultng from a tollng scheme. In our model, mperfect nformaton
8 caused a bounded degradaton of effcency, but relatvely hgh effcency was mantaned even for hgh levels of uncertanty. Future research wll focus on extendng ths result to broader classes of games. In the model addressed n ths paper, we have shown that even under uncertanty, t s possble to acheve sgnfcant gans n the effcency of socal behavor wth a smple dstrbuted taxaton scheme. Our model strongly assumes homogenety of the populaton s prce senstvtes; as research progresses, we hope to relax ths assumpton. Homogenety s rare n practce, and ncorporatng the effects of heterogenety nto our models wll be a crucal step. In partcular, uncertan heterogeneous populatons may be well-suted to heterogeneous tolls or a level of prce dscrmnaton. A crucal queston s: what gans n effcency are possble f we charge dfferent tolls for dfferent users? Ths work has found the optmal toll among a specfc class of tollng functons: Pgovan margnal-cost tolls scaled by a unform factor. It remans to be shown how our tolls compare wth tolls of some other form. Are these tolls the best of any possble toll? A more comprehensve revew of general taxaton functons for uncertan populatons s requred to answer ths queston. [7] M. J. Smth, The Margnal Cost Taxaton of a Transportaton Network, Transportaton Research Part B: Methodologcal, vol. 3, no. 3, pp , 979. [8] R. Cole, Y. Dods, and T. Roughgarden, Prcng network edges for heterogeneous selfsh users, Proc. of the thrty-ffth ACM symp. on Theory of computng - STOC 03, p. 52, [9] L. Flescher, K. Jan, and M. Mahdan, Tolls for heterogeneous selfsh users n multcommodty networks and generalzed congeston games, 45th Annu. IEEE Symp. on Foundatons of Computer Scence, pp , [20] A. C. Pgou, The Economcs of Welfare [2] T. Roughgarden, How Bad Is Selfsh Routng?, Journal of the ACM JACM (2002, vol. 49, no. 2, pp , [22] J. Wardrop, Some theoretcal aspects of road traffc research, n Proc. of the Insttuton of Cvl Engneers, Part II, Vol., No. 36, pp , 952. [23] D. Braess, Uber en Paradoxon aus der Verkehrsplanung, Unternehmensforschung, pp , 968. REFERENCES [] G. E. Bolton and A. Ockenfels, Behavoral economc engneerng, Journal of Economc Psychology, vol. 33, pp , June 202. [2] C. Slva, Applcaton of mechansm desgn to electrc power markets (Republshed, IEEE Transactons on Power Systems, vol. 6, no. 4, pp , 200. [3] D. Nyato and E. Hossan, Compettve Prcng for Spectrum Sharng n Cogntve Rado Networks : Dynamc Game, Ineffcency of Nash Equlbrum, and Colluson, IEEE Journal on Selected Areas n Communcatons, vol. 26, no., pp , [4] S. Morrson, A survey of road prcng, Transportaton Research Part A: General, no. 2, pp , 986. [5] C. Papadmtrou, Algorthms, Games, and the Internet, n Proc. of the 28th Internatonal Colloquum on Automata, Languages and Programmng, 200. [6] E. Koutsoupas and C. Papadmtrou, Worst-case equlbra, Computer Scence Revew, vol. 3, pp , May [7] R. Johar and J. N. Tstskls, Effcency Loss n a Network Resource Allocaton Game, Mathematcs of Operatons Research, vol. 29, pp , Aug [8] G. Peraks and G. Roels, The Prce of Anarchy n Supply Chans: Quantfyng the Effcency of Prce-Only Contracts, Management Scence, vol. 53, pp , Aug [9] H. Youn, M. Gastner, and H. Jeong, Prce of Anarchy n Transportaton Networks: Effcency and Optmalty Control, Physcal Revew Letters, vol. 0, p. 2870, Sept [0] N. G. Mankw, Smart Taxes: An Open Invtaton to Jon the Pgou Club, Eastern Economc Journal, vol. 35, no., pp. 4 23, [] W. Baumol, On taxaton and the control of externaltes, The Amercan Economc Revew, vol. 62, no. 3, pp , 972. [2] J. Stgltz, Usng tax polcy to curb speculatve short-term tradng, Journal of Fnancal Servces Research, vol. 5, no. May, pp. 0 5, 989. [3] I. W. H. Parry, M. Walls, and W. Harrngton, Automoble externaltes and polces, SSRN Electronc Journal, [4] T. Roughgarden, Selfsh routng and the prce of anarchy. MIT Press, [5] R. Cole, Y. Dods, and T. Roughgarden, How much can taxes help selfsh routng?, Journal of Computer and System Scences, vol. 72, pp , May [6] H. Yang and H.-J. Huang, Prncple of margnal-cost prcng: how does t work n a general road network?, Transportaton Research Part A: Polcy and Practce, vol. 32, pp , Jan. 998.
Social Coordination in Unknown Price-Sensitive Populations
Socal Coordnaton n Unnown Prce-Senstve Populatons Phlp N. Brown 1 and Jason R. Marden 2 Abstract In ths paper, we nvestgate the relatonshp between uncertanty and a desgner s ablty to nfluence socal behavor.
More informationPricing and Resource Allocation Game Theoretic Models
Prcng and Resource Allocaton Game Theoretc Models Zhy Huang Changbn Lu Q Zhang Computer and Informaton Scence December 8, 2009 Z. Huang, C. Lu, and Q. Zhang (CIS) Game Theoretc Models December 8, 2009
More informationCOS 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 informationPerfect Competition and the Nash Bargaining Solution
Perfect Competton and the Nash Barganng Soluton Renhard John Department of Economcs Unversty of Bonn Adenauerallee 24-42 53113 Bonn, Germany emal: rohn@un-bonn.de May 2005 Abstract For a lnear exchange
More informationModule 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 informationThe 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 informationAssortment 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 informationEconomics 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 informationWelfare Properties of General Equilibrium. What can be said about optimality properties of resource allocation implied by general equilibrium?
APPLIED WELFARE ECONOMICS AND POLICY ANALYSIS Welfare Propertes of General Equlbrum What can be sad about optmalty propertes of resource allocaton mpled by general equlbrum? Any crteron used to compare
More informationGlobal 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 informationInner 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 information3.1 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 informationPHYS 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 informationCompetition and Efficiency in Congested Markets
Competton and Effcency n Congested Markets Daron Acemoglu Department of Economcs Massachusetts Insttute of Technology Asuman E. Ozdaglar Department of Electrcal Engneerng and Computer Scence Massachusetts
More informationLinear 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 informationThe 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 informationMarket structure and Innovation
Market structure and Innovaton Ths presentaton s based on the paper Market structure and Innovaton authored by Glenn C. Loury, publshed n The Quarterly Journal of Economcs, Vol. 93, No.3 ( Aug 1979) I.
More informationProblem 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 informationLectures - 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 informationCOMPARISON 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 informationOnline Appendix: Reciprocity with Many Goods
T D T A : O A Kyle Bagwell Stanford Unversty and NBER Robert W. Stager Dartmouth College and NBER March 2016 Abstract Ths onlne Appendx extends to a many-good settng the man features of recprocty emphaszed
More informationModule 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 informationComputing Correlated Equilibria in Multi-Player Games
Computng Correlated Equlbra n Mult-Player Games Chrstos H. Papadmtrou Presented by Zhanxang Huang December 7th, 2005 1 The Author Dr. Chrstos H. Papadmtrou CS professor at UC Berkley (taught at Harvard,
More informationThe Minimum Universal Cost Flow in an Infeasible Flow Network
Journal of Scences, Islamc Republc of Iran 17(2): 175-180 (2006) Unversty of Tehran, ISSN 1016-1104 http://jscencesutacr The Mnmum Unversal Cost Flow n an Infeasble Flow Network H Saleh Fathabad * M Bagheran
More informationStructure 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 informationIndeterminate pin-jointed frames (trusses)
Indetermnate pn-jonted frames (trusses) Calculaton of member forces usng force method I. Statcal determnacy. The degree of freedom of any truss can be derved as: w= k d a =, where k s the number of all
More informationAffine 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 informationCollege 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 informationErrors for Linear Systems
Errors for Lnear Systems When we solve a lnear system Ax b we often do not know A and b exactly, but have only approxmatons  and ˆb avalable. Then the best thng we can do s to solve ˆx ˆb exactly whch
More informationn 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 informationCopyright (C) 2008 David K. Levine This document is an open textbook; you can redistribute it and/or modify it under the terms of the Creative
Copyrght (C) 008 Davd K. Levne Ths document s an open textbook; you can redstrbute t and/or modfy t under the terms of the Creatve Commons Attrbuton Lcense. Compettve Equlbrum wth Pure Exchange n traders
More informationMixed Taxation and Production Efficiency
Floran Scheuer 2/23/2016 Mxed Taxaton and Producton Effcency 1 Overvew 1. Unform commodty taxaton under non-lnear ncome taxaton Atknson-Stgltz (JPubE 1976) Theorem Applcaton to captal taxaton 2. Unform
More informationPROBLEM SET 7 GENERAL EQUILIBRIUM
PROBLEM SET 7 GENERAL EQUILIBRIUM Queston a Defnton: An Arrow-Debreu Compettve Equlbrum s a vector of prces {p t } and allocatons {c t, c 2 t } whch satsfes ( Gven {p t }, c t maxmzes βt ln c t subject
More informationNUMERICAL 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(1 ) (1 ) 0 (1 ) (1 ) 0
Appendx A Appendx A contans proofs for resubmsson "Contractng Informaton Securty n the Presence of Double oral Hazard" Proof of Lemma 1: Assume that, to the contrary, BS efforts are achevable under a blateral
More informationFeature Selection: Part 1
CSE 546: Machne Learnng Lecture 5 Feature Selecton: Part 1 Instructor: Sham Kakade 1 Regresson n the hgh dmensonal settng How do we learn when the number of features d s greater than the sample sze n?
More informationEstimation: Part 2. Chapter GREG estimation
Chapter 9 Estmaton: Part 2 9. GREG estmaton In Chapter 8, we have seen that the regresson estmator s an effcent estmator when there s a lnear relatonshp between y and x. In ths chapter, we generalzed the
More informationU.C. Berkeley CS294: Beyond Worst-Case Analysis Luca Trevisan September 5, 2017
U.C. Berkeley CS94: Beyond Worst-Case Analyss Handout 4s Luca Trevsan September 5, 07 Summary of Lecture 4 In whch we ntroduce semdefnte programmng and apply t to Max Cut. Semdefnte Programmng Recall that
More informationLecture 4. Instructor: Haipeng Luo
Lecture 4 Instructor: Hapeng Luo In the followng lectures, we focus on the expert problem and study more adaptve algorthms. Although Hedge s proven to be worst-case optmal, one may wonder how well t would
More information2E 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 informationMMA 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 informationAppendix for Causal Interaction in Factorial Experiments: Application to Conjoint Analysis
A Appendx for Causal Interacton n Factoral Experments: Applcaton to Conjont Analyss Mathematcal Appendx: Proofs of Theorems A. Lemmas Below, we descrbe all the lemmas, whch are used to prove the man theorems
More informationKernel 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 informationNotes 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 informationOpen Systems: Chemical Potential and Partial Molar Quantities Chemical Potential
Open Systems: Chemcal Potental and Partal Molar Quanttes Chemcal Potental For closed systems, we have derved the followng relatonshps: du = TdS pdv dh = TdS + Vdp da = SdT pdv dg = VdP SdT For open systems,
More informationDifference 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 informationOnline Appendix. t=1 (p t w)q t. Then the first order condition shows that
Artcle forthcomng to ; manuscrpt no (Please, provde the manuscrpt number!) 1 Onlne Appendx Appendx E: Proofs Proof of Proposton 1 Frst we derve the equlbrum when the manufacturer does not vertcally ntegrate
More informationprinceton univ. F 17 cos 521: Advanced Algorithm Design Lecture 7: LP Duality Lecturer: Matt Weinberg
prnceton unv. F 17 cos 521: Advanced Algorthm Desgn Lecture 7: LP Dualty Lecturer: Matt Wenberg Scrbe: LP Dualty s an extremely useful tool for analyzng structural propertes of lnear programs. Whle there
More informationLecture 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 informationTransfer 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 informationPsychology 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 informationCHAPTER 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 informationEquilibrium with Complete Markets. Instructor: Dmytro Hryshko
Equlbrum wth Complete Markets Instructor: Dmytro Hryshko 1 / 33 Readngs Ljungqvst and Sargent. Recursve Macroeconomc Theory. MIT Press. Chapter 8. 2 / 33 Equlbrum n pure exchange, nfnte horzon economes,
More informationEEL 6266 Power System Operation and Control. Chapter 3 Economic Dispatch Using Dynamic Programming
EEL 6266 Power System Operaton and Control Chapter 3 Economc Dspatch Usng Dynamc Programmng Pecewse Lnear Cost Functons Common practce many utltes prefer to represent ther generator cost functons as sngle-
More informationAPPROXIMATE PRICES OF BASKET AND ASIAN OPTIONS DUPONT OLIVIER. Premia 14
APPROXIMAE PRICES OF BASKE AND ASIAN OPIONS DUPON OLIVIER Prema 14 Contents Introducton 1 1. Framewor 1 1.1. Baset optons 1.. Asan optons. Computng the prce 3. Lower bound 3.1. Closed formula for the prce
More informationLecture 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 informationSupplementary 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 informationUniversity of California, Davis Date: June 22, 2009 Department of Agricultural and Resource Economics. PRELIMINARY EXAMINATION FOR THE Ph.D.
Unversty of Calforna, Davs Date: June 22, 29 Department of Agrcultural and Resource Economcs Department of Economcs Tme: 5 hours Mcroeconomcs Readng Tme: 2 mnutes PRELIMIARY EXAMIATIO FOR THE Ph.D. DEGREE
More informationMaximizing the number of nonnegative subsets
Maxmzng the number of nonnegatve subsets Noga Alon Hao Huang December 1, 213 Abstract Gven a set of n real numbers, f the sum of elements of every subset of sze larger than k s negatve, what s the maxmum
More informationx i1 =1 for all i (the constant ).
Chapter 5 The Multple Regresson Model Consder an economc model where the dependent varable s a functon of K explanatory varables. The economc model has the form: y = f ( x,x,..., ) xk Approxmate ths by
More informationResource 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 informationThe Study of Teaching-learning-based Optimization Algorithm
Advanced Scence and Technology Letters Vol. (AST 06), pp.05- http://dx.do.org/0.57/astl.06. The Study of Teachng-learnng-based Optmzaton Algorthm u Sun, Yan fu, Lele Kong, Haolang Q,, Helongang Insttute
More informationEcon107 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 informationThe 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 informationChapter 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 informationSolutions to exam in SF1811 Optimization, Jan 14, 2015
Solutons to exam n SF8 Optmzaton, Jan 4, 25 3 3 O------O -4 \ / \ / The network: \/ where all lnks go from left to rght. /\ / \ / \ 6 O------O -5 2 4.(a) Let x = ( x 3, x 4, x 23, x 24 ) T, where the varable
More informationHow Strong Are Weak Patents? Joseph Farrell and Carl Shapiro. Supplementary Material Licensing Probabilistic Patents to Cournot Oligopolists *
How Strong Are Weak Patents? Joseph Farrell and Carl Shapro Supplementary Materal Lcensng Probablstc Patents to Cournot Olgopolsts * September 007 We study here the specal case n whch downstream competton
More informationYong Joon Ryang. 1. Introduction Consider the multicommodity transportation problem with convex quadratic cost function. 1 2 (x x0 ) T Q(x x 0 )
Kangweon-Kyungk Math. Jour. 4 1996), No. 1, pp. 7 16 AN ITERATIVE ROW-ACTION METHOD FOR MULTICOMMODITY TRANSPORTATION PROBLEMS Yong Joon Ryang Abstract. The optmzaton problems wth quadratc constrants often
More informationCSci 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 informationSimultaneous Optimization of Berth Allocation, Quay Crane Assignment and Quay Crane Scheduling Problems in Container Terminals
Smultaneous Optmzaton of Berth Allocaton, Quay Crane Assgnment and Quay Crane Schedulng Problems n Contaner Termnals Necat Aras, Yavuz Türkoğulları, Z. Caner Taşkın, Kuban Altınel Abstract In ths work,
More informationCanonical 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 informationSmooth Games, Price of Anarchy and Composability of Auctions - a Quick Tutorial
Smooth Games, Prce of Anarchy and Composablty of Auctons - a Quck Tutoral Abhshek Snha Laboratory for Informaton and Decson Systems, Massachusetts Insttute of Technology, Cambrdge, MA 02139 Emal: snhaa@mt.edu
More information2.3 Nilpotent endomorphisms
s a block dagonal matrx, wth A Mat dm U (C) In fact, we can assume that B = B 1 B k, wth B an ordered bass of U, and that A = [f U ] B, where f U : U U s the restrcton of f to U 40 23 Nlpotent endomorphsms
More informationECE559VV 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 informationEEE 241: Linear Systems
EEE : Lnear Systems Summary #: Backpropagaton BACKPROPAGATION The perceptron rule as well as the Wdrow Hoff learnng were desgned to tran sngle layer networks. They suffer from the same dsadvantage: they
More informationSalmon: Lectures on partial differential equations. Consider the general linear, second-order PDE in the form. ,x 2
Salmon: Lectures on partal dfferental equatons 5. Classfcaton of second-order equatons There are general methods for classfyng hgher-order partal dfferental equatons. One s very general (applyng even to
More informationSection 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 informationComposite Hypotheses testing
Composte ypotheses testng In many hypothess testng problems there are many possble dstrbutons that can occur under each of the hypotheses. The output of the source s a set of parameters (ponts n a parameter
More informationChapter 2 - The Simple Linear Regression Model S =0. e i is a random error. S β2 β. This is a minimization problem. Solution is a calculus exercise.
Chapter - The Smple Lnear Regresson Model The lnear regresson equaton s: where y + = β + β e for =,..., y and are observable varables e s a random error How can an estmaton rule be constructed for the
More informationWeek3, 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 informationNumerical 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 informationOn the Multicriteria Integer Network Flow Problem
BULGARIAN ACADEMY OF SCIENCES CYBERNETICS AND INFORMATION TECHNOLOGIES Volume 5, No 2 Sofa 2005 On the Multcrtera Integer Network Flow Problem Vassl Vasslev, Marana Nkolova, Maryana Vassleva Insttute of
More informationU.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 information2016 Wiley. Study Session 2: Ethical and Professional Standards Application
6 Wley Study Sesson : Ethcal and Professonal Standards Applcaton LESSON : CORRECTION ANALYSIS Readng 9: Correlaton and Regresson LOS 9a: Calculate and nterpret a sample covarance and a sample correlaton
More informationSolution Thermodynamics
Soluton hermodynamcs usng Wagner Notaton by Stanley. Howard Department of aterals and etallurgcal Engneerng South Dakota School of nes and echnology Rapd Cty, SD 57701 January 7, 001 Soluton hermodynamcs
More informationOperating conditions of a mine fan under conditions of variable resistance
Paper No. 11 ISMS 216 Operatng condtons of a mne fan under condtons of varable resstance Zhang Ynghua a, Chen L a, b, Huang Zhan a, *, Gao Yukun a a State Key Laboratory of Hgh-Effcent Mnng and Safety
More informationLecture 3. Ax x i a i. i i
18.409 The Behavor of Algorthms n Practce 2/14/2 Lecturer: Dan Spelman Lecture 3 Scrbe: Arvnd Sankar 1 Largest sngular value In order to bound the condton number, we need an upper bound on the largest
More informationMore 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 informationSIO 224. m(r) =(ρ(r),k s (r),µ(r))
SIO 224 1. A bref look at resoluton analyss Here s some background for the Masters and Gubbns resoluton paper. Global Earth models are usually found teratvely by assumng a startng model and fndng small
More informationThe 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 informationIII. Econometric Methodology Regression Analysis
Page Econ07 Appled Econometrcs Topc : An Overvew of Regresson Analyss (Studenmund, Chapter ) I. The Nature and Scope of Econometrcs. Lot s of defntons of econometrcs. Nobel Prze Commttee Paul Samuelson,
More informationFundamental loop-current method using virtual voltage sources technique for special cases
Fundamental loop-current method usng vrtual voltage sources technque for specal cases George E. Chatzaraks, 1 Marna D. Tortorel 1 and Anastasos D. Tzolas 1 Electrcal and Electroncs Engneerng Departments,
More informationCS286r Assign One. Answer Key
CS286r Assgn One Answer Key 1 Game theory 1.1 1.1.1 Let off-equlbrum strateges also be that people contnue to play n Nash equlbrum. Devatng from any Nash equlbrum s a weakly domnated strategy. That s,
More informationELASTIC 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 informationGames 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 informationIterative General Dynamic Model for Serial-Link Manipulators
EEL6667: Knematcs, Dynamcs and Control of Robot Manpulators 1. Introducton Iteratve General Dynamc Model for Seral-Lnk Manpulators In ths set of notes, we are gong to develop a method for computng a general
More informationSingular Value Decomposition: Theory and Applications
Sngular Value Decomposton: Theory and Applcatons Danel Khashab Sprng 2015 Last Update: March 2, 2015 1 Introducton A = UDV where columns of U and V are orthonormal and matrx D s dagonal wth postve real
More informationLinear 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 informationSolutions HW #2. minimize. Ax = b. Give the dual problem, and make the implicit equality constraints explicit. Solution.
Solutons HW #2 Dual of general LP. Fnd the dual functon of the LP mnmze subject to c T x Gx h Ax = b. Gve the dual problem, and make the mplct equalty constrants explct. Soluton. 1. The Lagrangan s L(x,
More informationPhysics 5153 Classical Mechanics. D Alembert s Principle and The Lagrangian-1
P. Guterrez Physcs 5153 Classcal Mechancs D Alembert s Prncple and The Lagrangan 1 Introducton The prncple of vrtual work provdes a method of solvng problems of statc equlbrum wthout havng to consder the
More information