Algorithmica 2001 Springer-Verlag New York Inc.

Size: px
Start display at page:

Download "Algorithmica 2001 Springer-Verlag New York Inc."

Transcription

1 Algorthmca (2001) 29: DOI: /s Algorthmca 2001 Sprnger-Verlag New York Inc. Effcent Algorthms for Integer Programs wth Two Varables per Constrant 1 R. Bar-Yehuda 2 and D. Rawtz 2 Abstract. Gven a bounded nteger program wth n varables and m constrants, each wth two varables, we present an O(mU) tme and O(m) space feasblty algorthm, where U s the maxmal varable range sze. We show that wth the same complexty we can fnd an optmal soluton for the postvely weghted mnmzaton problem for monotone systems. Usng the local-rato technque we develop an O(nmU) tme and O(m) space 2-approxmaton algorthm for the postvely weghted mnmzaton problem for the general case. We further generalze all results to nonlnear constrants (called axs-convex constrants) and to nonlnear (but monotone) weght functons. Our algorthms are not only better n complexty than other known algorthms, but also consderably smpler, and they contrbute to the understandng of these very fundamental problems. Key Words. Combnatoral optmzaton, Integer programmng, Approxmaton algorthm, Local-rato technque, 2SAT, Vertex cover. 1. Introducton. Ths paper s motvated by a paper by Hochbaum et al. [11] whch dscusses nteger programs wth two varables per constrant. The problem s defned as follows: (2VIP) mn n =1 w x s.t. a k x k + b k x k c k, k {1,...,m}, l x u, {1,...,n}, x N, {1,...,n}, where 1 k, k n, w 0, a, b, c Z m, and l, u N n. Obvously, ths problem s a generalzaton of the well-known mnmum weght vertex cover problem (VC) and the mnmum weght 2 satsfablty problem (2SAT). Both problems are known to be NP-hard [7], and the best known approxmaton rato for VC [3], [10], [14] and 2SAT [9] s asymptotcally 2. Both results are best vewed va the local-rato technque (see [2] and [4]). A 2VIP system s called monotone f each constrant s an nequalty on two varables wth coeffcents of opposte sgns. The problem of checkng whether an nteger monotone system has a feasble soluton was shown to be NP-complete by Lagaras [13]. Therefore, even for the 2VIP feasblty problem, t s natural to consder pseudopolynomal algorthms,.e., algorthms wth a runnng tme whch s polynomal n the sze of the nput and n U, where U = max {u l }. 1 A prelmnary verson of ths paper appeared n the Proceedngs of the 7th Annual European Symposum on Algorthms (1999). 2 Computer Scence Department, Technon - IIT, Hafa Israel. {reuven,rawtz}@cs.technon.ac.l. Receved June 21, 1996; revsed December 5, Communcated by N. Megddo. Onlne publcaton December 15, 2000.

2 596 R. Bar-Yehuda and D. Rawtz Hochbaum and Naor [12] were the frst to consder effcent algorthms for nteger programs wth two varables per nequalty. They presented an O(mn 2 log m + nmu) tme feasblty algorthm for monotone 2VIP systems, and an optmzaton algorthm for monotone systems wth general (possbly negatve) weghts. Ths optmzaton algorthm constructs a graph representng the monotone system n queston and then uses a maxmum flow algorthm. Consequently, the tme complexty of ther algorthm s relatvely hgh,.e., when usng Goldberg and Taran s maxmum-flow algorthm [8] t s O(nmU 2 log(n 2 U/m)). Usng ths optmzaton algorthm, Hochbaum et al. [11] were able to construct an O(nmU 2 log(n 2 U/m)) tme 2-approxmaton algorthm for the 2VIP problem. Ther algorthm also uses an O(mU) tme and space 2VIP feasblty algorthm based on a transformaton to 2SAT whch they attrbute to Feder. By usng the local-rato technque, we present an O(nmU) tme and O(m) space 2- approxmaton algorthm. Ths algorthm s not only more effcent, but also more natural and smpler. In order to develop an approxmaton algorthm, t seems natural to study the feasblty problem frst. 3 Indeed, our 2-approxmaton algorthm s n fact a specfc mplementaton of a feasblty algorthm presented here. The remander of ths paper s organzed as follows: In Secton 2 we present an O(mU) tme and O(m) space feasblty algorthm for 2VIP systems. The 2-approxmaton algorthm for 2VIP systems s presented n Secton 3. In Secton 4 we show that the feasblty algorthm and the approxmaton algorthm presented n ths paper can be generalzed to some nonlnear systems wth the same tme and space complexty. We defne a generalzaton of lnear nequaltes, called axs-convex constrants, and show that the algorthms can be generalzed to work wth such constrants. We also generalze the 2-approxmaton algorthm to obectve functons of the form n =1 w (x ), where all the w s are monotone weght functons. An optmalty algorthm for monotone lnear systems appears n Secton 5. We show that ths algorthm can work wth some nonlnear constrants, and we generalze the algorthm to monotone weght functons as well. Table 1 summarzes the results for 2VIP systems. Table 1. Summary of results. Problem Prevous results (tme, space) Our results (tme, space) 2SAT feasblty O(m), O(m) [6] 2VIP feasblty O(mU), O(mU) O(mU), O(m) by usng reducton to 2SAT [11] 2SAT 2-approxmaton O(nm), O(n 2 + m) [9] O(nm), O(m) 2VIP 2-approxmaton O(nmU 2 log(n 2 U/m)), O(mU) [11] O(nmU), O(m) Monotone 2VIP O(mU), O(mU) O(mU), O(m) optmzaton by usng reducton to 2SAT 3 Ths was done by Hochbaum and Naor [12] for the monotone 2VIP problem, by Hochbaum et al. [11] for the 2VIP problem, and by Gusfeld and Ptt [9] for the 2SAT problem.

3 Effcent Algorthms for Integer Programs wth Two Varables per Constrant Feasblty Algorthm. Gven a 2VIP system, we are nterested n developng an algorthm whch fnds a feasble soluton, f such a soluton exsts. Snce the specal case where l = 0 n and u = 1 n s the known 2SAT feasblty problem, t s natural to try to extend the well known O(m) tme and space algorthm of Even et al. [6]. It s possble to transform the gven 2VIP system nto an equvalent 2SAT nstance wth nu varables and (m + n)u constrants (ths transformaton due to Feder appears n [11]). By combnng ths transformaton wth the lnear tme and space algorthm of Even et al. we get an O(mU) tme and space feasblty algorthm. In ths secton we present an O(mU) tme and O(m) space feasblty algorthm whch generalzes the algorthm by Even et al. The man dea of the algorthm from [6] s as follows: choose a varable x t, frst dscover the forced values of other varables by assgnng x t = 0, and then do the same for the assgnment x t = 1. If one of these assgnments does not lead to a contradcton, assgn ths value to x t and make the correspondng forced assgnments. The correctness of the approach of Even et al. s shown by provng that a noncontradctory assgnment preserves the feasblty property. The effcency of ther algorthm s acheved by dscoverng the forced assgnments of x t = 0 and those of x t = 1 n parallel. Our 2VIP feasblty algorthm works wth bounds, as opposed to assgnments. We choose a varable x t and an nteger α [l t, u t ] and dscover the forced bounds of other varables by consderng n turn the bound x t α and the bound x t >α. For ths purpose we use constrant propagaton. 4 Much lke Even et al. [6], we prove that a noncontradctory bound preserves the feasblty property, and use ths to show the correctness of our algorthm. Also, the effcency of our algorthm s acheved by dscoverng the forced bounds by both new bounds of x t n parallel. The purpose of ths secton s not only to show the factor (U) mprovement n space complexty, but also to lay the foundatons for the 2-approxmaton algorthm presented n the next secton. DEFINITION 1. For a gven 2VIP nstance, sat(l, u) ={x: l x u and x satsfes all 2VIP constrants}. DEFINITION 2. For the kth constrant (on the varables x k, x k ) of a gven 2VIP nstance, constrant(k) ={(α, β): x k = α, x k = β satsfy constrant k}. DEFINITION 3. Gven α, β Z we defne [α, β] ={z Z: α z β}. A lnear constrant s convex, thus the followng hold for the kth constrant on the varables x and x : OBSERVATION 1. δ [β,γ]. If (α, β), (α, γ ) constrant(k), then (α, δ) constrant(k) for all 4 Constrant propagaton was prevously used for the LP verson of the problem (e.g., see [1] and [15]), and n [12] for nteger feasblty over monotone nequaltes.

4 598 R. Bar-Yehuda and D. Rawtz Fg. 1. Routne OneOnOneImpact. OBSERVATION 2. If (α 1,α 2 ), (β 1,β 2 ), (γ 1,γ 2 ) constrant(k), then all ponts nsde the trangle nduced by (α 1,α 2 ), (β 1,β 2 ), and (γ 1,γ 2 ) satsfy constrant k. We present a routne n Fgure 1 whch wll be repeatedly used for constrant propagaton. It receves as nput two arrays l and u of sze n (passed by reference), two varables ndces,, and a constrant ndex k on these two varables. The obectve of ths routne s to fnd the mpact of constrant k and the bounds l, u on the bounds l, u. We denote by l after and u after the values of the bounds l and u after callng OneOnOne- Impact(l, u,,, k). OBSERVATION 3. OBSERVATION 4. constrant(k). sat(l after, u after ) = sat(l, u). If β [l after, u after ] there exsts α [l after, u after ] such that (α, β) The routne n Fgure 2, whch s called OneOnAllImpact, receves as nput two arrays l and u of sze n (passed by reference) and a varable ndex t, and changes l and u accordng to the mpact of l t and u t on all the ntervals. We now prove that we do not lose feasble solutons after actvatng OneOnAllImpact. Fg. 2. Routne OneOnAllImpact.

5 Effcent Algorthms for Integer Programs wth Two Varables per Constrant 599 LEMMA 1. If l after and u after are the values of l and u after callng OneOnAllImpact, then sat(l after, u after ) = sat(l, u). PROOF. All changes to l and u are made by routne OneOnOneImpact. It s easy to prove the lemma by nducton usng Observaton 3. LEMMA 2. If OneOnAllImpact(l, u, t) termnates wthout falure wth the bounds l after and u after and sat((l 1,...,l t 1,,l t+1,...,l n ), (u 1,...,u t 1,, u t+1,...,u n )), then sat(l after, u after ). PROOF. Let y sat((l 1,...,l t 1,,l t+1,...,l n ), (u 1,...,u t 1,, u t+1,..., u n )). We defne a vector y as y t, y t [l after y t t, u after t ], = l after t, y t <l after t, u after t, y t > u after t. Consder constrant k on x and x. We need to show that y, y constrant(k). Case 1: y [l after, u after ] and y [l after, u after ]. (y, y ) = (y, y ) constrant(k). Case 2: y <l after and y [l after, u after ]. y s a feasble soluton, thus (y, y ) constrant(k). When we changed the lower bound of x to l after we called OneOnOneImpact for all constrants nvolvng x ncludng constrant k. By Observaton 4 there exsts α [l after, u after ] for whch (α, y ) constrant(k). Thus, by Observaton 1 we get that, y ) constrant(k). (l after Case 3: y <l after and y <l after. y s a feasble soluton, thus (y, y ) constrant(k). When we changed the lower bound of x to l after we called OneOnOneImpact for all constrant nvolvng x ncludng constrant k. By Observaton 4 there exsts α [l after, u after ] for whch (α, l after ) constrant(k). From the same arguments we get that there exsts β [l after, u after ] for whch (l after, β) constrant(k) as well. Thus, by Observaton 2 we get that (l after,l after ) constrant(k). Other cases are smlar to Cases 2 and 3. After provng that OneOnAllImpact preserves the feasblty property t s possble to use ths routne as part of the feasblty algorthm from Fgure 3. THEOREM 1. Algorthm Feasblty returns a feasble soluton f such a soluton exsts. PROOF. Each recursve call reduces at least one of the ranges (the t th), thus the executon of the algorthm must termnate. By Lemma 1 f sat(l, u) = the algorthm wll return fal. On the other hand, for the case sat(l, u) we can prove by nducton on n =1 (u l ) that the algorthm fnds a feasble soluton. Base. n =1 (u l ) = 0 mples l = u, thus x = l s a feasble soluton.

6 600 R. Bar-Yehuda and D. Rawtz Fg. 3. Feasblty algorthm. Step. By Lemma 1 at least one of the calls to OneOnAllImpact termnates wthout falure. If call left was chosen, then by Lemma 2 we know that sat(l left, u left ). Therefore, by the nducton hypothess we can fnd a feasble soluton for l left, u left. Obvously, a feasble soluton x sat(l left, u left ) satsfes x sat(l, u). The same apples for call rght. Ths concludes the proof. THEOREM 2. O(m). Algorthm Feasblty can be mplemented n tme O(mU) and space PROOF. To acheve a tme complexty of O(mU), we run both calls to OneOnAllImpact n parallel (ths approach was used for 2SAT by Even et al. [6]), and prefer the faster opton of the two, f such a choce exsts. After every change n the range of a varable x, we need to check the m constrants nvolvng ths varable, n order to dscover the mpact of the change. To perform ths task effcently we can store the nput n an ncdence lst, where every varable has ts constrants lst. As l and u can change up to (u l ) tmes, we conclude that the total tme complexty of the changes s O( n =1 m (u l )) = O(mU) (the tme wasted on uncompleted trals s bounded by the tme complexty of the chosen trals). The algorthm uses O(m) space for the nput and a constant number of arrays of sze n, thus uses lnear space. 3. From Feasblty to Approxmaton. Before presentng our approxmaton algorthm, we frst dscuss the specal case where U = 2, whch s the mnmum 2SAT problem, and ts specal case, the vertex cover problem. The man dea of Bar-Yehuda and Even s 2-approxmaton algorthm for the vertex cover problem [2] [4] s as follows: choose an edge e = (u,v) and subtract ε =

7 Effcent Algorthms for Integer Programs wth Two Varables per Constrant 601 mn{w(u), w(v)} from both w(u) and w(v). Every vertex cover C V must cover e, therefore the subtracton of ε from w(u), w(v), whch may cost up to 2 ε, reduces the optmum by at least ε. After repeatng ths process untl such subtractons are no longer possble, the resultng cover s C ={v: w(v) = 0}. When consderng a 2CNF formula, 5 wth respect to the mnmum 2SAT problem, monotone clauses (e.g., x x ) can be treated as n the vertex cover case. However, what about clauses wth negatve lterals (of the form x x or x x )? Also, even f we knew how to make weght subtractons, how would we choose a truth assgnment? Thus, the above algorthm should be modfed n order to be employed for the mnmum 2SAT problem. The dea s that for a gven 2CNF formula, f x 1 x 2 and x 1 x 3 we can subtract ε = mn{w 2,w 3 } from w 2 and w 3. Ths leaves us wth the task of choosng a feasble truth assgnment: when t s possble, assgn a zero cost partal assgnment, whle relyng upon the consstency property (Lemma 2). Ths approach was used by Gusfeld and Ptt [9] for approxmatng the mnmum 2SAT problem. The 2CNF formula can be presented as a dgraph where each vertex represents a boolean varable or ts negaton, and an edge represents a OneOnOneImpact propagaton (logcal ). A propagaton of an assgnment can be vewed as a traversal (e.g., BFS, DFS) of the dgraph. In order to be able to propagate forced assgnments, Gusfeld and Ptt s algorthm starts wth a preprocess phase of constructng a transtve closure. Ths phase uses (n 2 ) extra memory, whch s expensve. It s much more crtcal when tryng to approxmate 2VIP by usng the transformaton to 2SAT from [11], and then Gusfeld and Ptt s [9] algorthm. In ths case the preprocess uses (n 2 U 2 ) extra memory. 6 It seems only natural to try to extend the prevous deas when approxmatng the 2VIP problem. In the prevous secton we have seen that t s possble to generalze the 2CNF assgnment propagaton to a 2VIP bound propagaton, but how should we mplement weght subtractons? We can vew a boolean assgnment as a bound change, thus the cost of the assgnment x = 1 s actually the cost of rasng the lower bound of x from 0 to 1. For a 2VIP nstance, every unt ncrease of the varable x would cost w,orn other words, an ncrease of the lower bound l to l costs w (l l ). Bearng that n mnd, we defne an array called ˆl, whch holds the values (of the varables), for whch we have already pad. Ths means that nstead of reducng a weght due to a rse of l we ncrease ˆl. Note that, unlke l, ˆl can hold nonntegral values. Ths allows us to avod usng drect weght reductons and the consequent reducton to 2SAT. We present an O(nmU) tme and O(m) space 2-approxmaton algorthm. Ths approxmaton algorthm s a specfc mplementaton of our feasblty algorthm (namely, the algorthm chooses varables and bounds n a specfc order to get a 2-approxmaton). Not only does ths algorthm seem natural, but also ts complexty, O(nm) tme and O(m) space, n the case of 2SAT s lower than that of Gusfeld and Ptt s 2SAT algorthm (O(nm) and O(n 2 + m) correspondngly). In order to use the local-rato technque [2] we extend the 2VIP defnton. 5 Conunctve Normal Form, see, e.g., [5]. 6 Ths s n addton to the (mu 2 ) extra memory needed for the transformaton. As far as we know, every algorthm whch reles upon drect 2SAT transformaton suffers from ths drawback.

8 602 R. Bar-Yehuda and D. Rawtz DEFINITION 4. Gven a, b R n, we wrte a b f a b. Also, we defne max{a, b} =(max{a 1, b 1 },...,max{a n, b n }), mn{a, b} =(mn{a 1, b 1 },...,mn{a n, b n }). Gven l, u N n and ˆl, û R n for whch l ˆl û u, we defne the followng Extended 2VIP problem: (E2VIP) mn n =1 (x, ˆl, û )w s.t. a k x k + b k x k c k, k {1,...,m}, x [l, u ], {1,...,n}, where 0, x < ˆl, (x, ˆl, û ) = (x ˆl ), x [ ˆl, û ], (û ˆl ), x > û, and 1 k, k n, w 0, a, b, c Z m, and l, u N n. We defne W (x, ˆl, û) = n =1 (x, ˆl, û )w. A feasble soluton x s called an optmal soluton f, for every feasble soluton x, W (x, ˆl, û) W (x, ˆl, û). Wedenote W ( ˆl, û) = W (x, ˆl, û). A feasble soluton x s called an r-approxmaton f W (x, ˆl, û) r W ( ˆl, û). OBSERVATION 5. Gven ˆl, û, ˆm R n for whch ˆl ˆm û, we get W (x, ˆl, û) = W (x, ˆl, ˆm) + W (x, ˆm, û). Smlarly to the Decomposton Observaton from [2] we have: OBSERVATION 6 (Decomposton Observaton). ˆm û, W ( ˆl, ˆm) + W ( ˆm, û) W ( ˆl, û). Gven ˆl, û, ˆm R n such that ˆl PROOF. Let x, y, and z be optmal solutons for the system wth respect to ˆl, ˆm; ˆm, û; and ˆl, û correspondngly. W ( ˆl, ˆm) + W ( ˆm, û) = W (x, ˆl, ˆm) + W (y, ˆm, û) (by defnton) W (z, ˆl, ˆm) + W (z, ˆm, û) (optmalty of x, y ) W (z, ˆl, û) (observaton 5) = W ( ˆl, û) (by defnton). The followng s ths paper s Local-Rato Theorem (see [2] and [4]): THEOREM 3. If x s an r-approxmaton wth respect to ˆl, ˆm and wth respect to ˆm, û, then x s an r-approxmaton wth respect to ˆl, û.

9 Effcent Algorthms for Integer Programs wth Two Varables per Constrant 603 Fg Approxmaton algorthm. PROOF. W (x, ˆl, û) = W (x, ˆl, ˆm) + W (x, ˆm, û) (Observaton 5) r W ( ˆl, ˆm) + r W ( ˆl, û) (gven) r W ( ˆl, û) (Decomposton Observaton). We are ready to present the 2-approxmaton algorthm see Fgure 4. OBSERVATION 7. Feasblty. Algorthm Approxmate s a specfc mplementaton of Algorthm THEOREM 4. systems. Algorthm Approxmate s a 2-approxmaton algorthm for E2VIP PROOF. By Observaton 7 Algorthm Approxmate returns a feasble soluton. We prove by nducton on the depth of the recurson that the algorthm fnds a 2-approxmaton.

10 604 R. Bar-Yehuda and D. Rawtz Base. u = l mples W (l, ˆl, û) = 0. Step. There are several cases: Case 1: ˆl l. A 2-approxmaton wth respect to max{ ˆl, l} s obvously a 2- approxmaton soluton wth respect to ˆl. Case 2: û u. Trval. Case 3: Call rght faled. By Lemma 1 there s no feasble soluton whch satsfes x t α + 1, therefore we do not change the problem by addng the constrant x y α. By Lemma 1 callng OneOnAllImpact(l left, u left, t) does not change the problem ether. Case 4: Call left faled. Smlar to Case 3. Case 5: Both calls succeeded and W (l left, ˆl, û) W (l rght, ˆl, û). We frst show that every feasble soluton s a 2-approxmaton wth respect to ˆl and ˆm. We examne an optmal soluton x wth respect to ˆl and ˆm. By the constructon of ˆm we know that ˆm l left. Thus, f l left x u left, then the monotoncty of W (, ˆl, ˆm) mples If l rght x u rght, then W (x, ˆl, ˆm) W (l left, ˆl, ˆm). W (x, ˆl, ˆm) W (l rght, ˆl, ˆm) (x l rght ) W (l left, ˆl, û) (by the constructon of ˆm) = W (l left, ˆl, ˆm) (l left ˆm). On the other hand, by the constructon of ˆm: W ( ˆm, ˆl, ˆm) W (l left, ˆl, ˆm) + W (l rght, ˆl, ˆm) 2 W (l left, ˆl, ˆm). Thus, for every feasble soluton x: W (x, ˆl, ˆm) 2 W (l left, ˆl, ˆm). Therefore, by Theorem 3 a 2-approxmaton wth respect to ˆm and û s a 2-approxmaton wth respect to ˆl and û. We need to show that there exsts an optmal soluton x wth respect to ˆm and û for whch x α. For every feasble soluton y such that y α + 1 we defne y as y, y = l left, y <l left u left, y > u left y [l left, u left By Lemma 2 y s a feasble soluton. l left ˆm mples W (y, ˆm, û) W (y, ˆm, û), thus there s an optmal soluton wth respect to ˆm and û wthn the bounds l left, u left. Therefore, a 2-approxmaton wthn the bounds l left, u left s a 2-approxmaton wth respect to ˆm and û. Case 6: Both calls succeeded and W (l left, ˆl, û) >W (l rght, ˆl, û). Smlar to Case 5. Ths concludes the proof.,. ],

11 Effcent Algorthms for Integer Programs wth Two Varables per Constrant 605 COROLLARY 5. systems. THEOREM 6. O(m). Algorthm Approxmate s a 2-approxmaton algorthm for 2VIP Algorthm Approxmate can be mplemented n tme O(nmU) and space PROOF. In order to get the requred tme complexty, we must choose the x t s carefully. One possblty s to choose the varables n an ncreasng order,.e., x 1, x 2,...,x n, and to restart from the begnnng after reachng x n. We call such n teratons on all n varables a pass. As stated before, changng the range of x mght cause changes n the ranges of other varables. The exstence of a constrant on x and another varable x makes x a canddate for a range update. Ths means that we have to check the m constrants nvolvng x n order to dscover the consequences of changng ts range each tme ths range changes. l and u can change up to (u l ) tmes, therefore we get that the tme complexty of a sngle teraton s O(mU + n) = O(mU). One pass may nvolve all n varables, so the tme complexty of one pass s O(nmU). By choosng α = 1 2 (l t + u t ), we reduce the possble range for x t at least by half. Therefore, n a sngle pass we reduce the possble ranges for all varables at least by half. Thus, we get that the total tme complexty s log U k=1 O(mn(U/2k )) = O(mnU). As before, the algorthm uses an ncdence lst data structure and a constant number of arrays of sze n, thus uses lnear space. 4. Generalzatons. What f the constrants are not lnear? Also, what f the weght functons of the varables are not lnear? Can we avod the expensve transformaton to the 2SAT problem? In ths secton we try to extend our approach to a wder famly of problems. In order to do so, we had to dentfy the propertes of lnear constrants and lnear weght functons whch are suffcent for the correctness of our clams Generalzed Constrants. In ths subsecton we are nterested n problems of the form: (G2VIP) mn n =1 w x s.t. (x k, x k ) constrant(k), k {1,...,m}, x [l, u ], {1,...,n}, where 1 k, k n, w 0, and l, u N n. Also each constrant(k) satsfes the followng: f (α 1,α 2 ), (β 1,β 2 ) constrant(k), then there exsts a shortest path n N 2 (lattce) connectng (α 1,α 2 ) and (β 1,β 2 ) such that all ts ponts satsfy constrant k. A constrant whch satsfes ths property s called an axs-convex constrant. We assume that for an axs-convex constrant we have an O(1) tme oracle OneOnOneImpact whch returns a tght range on x when gven a range for x. The followng s mpled by the defnton of an axs-convex constrant: OBSERVATION 8. Gven an axs-convex constrant C, f (α 1,α 2 ), (β 1,β 2 ) C, then for all β [β 1,β 2 ] there exsts α [α 1,α 2 ] such that (α, β) C.

12 606 R. Bar-Yehuda and D. Rawtz Observatons 1 and 4 hold for axs-convex constrants. A weaker verson of Observaton 2 also holds for axs-convex constrants. The proof uses Observaton 8. OBSERVATION 9. Gven an axs-convex constrant C and (α 1,α 2 ), (β 1,β 2 ), (γ 1,γ 2 ) C, f (α 1,β 2 ) s nsde the trangle nduced by (α 1,α 2 ), (β 1,β 2 ), and (γ 1,γ 2 ), then (α 1,β 2 ) C. It s easy to see that Lemmas 1 and 2 reman vald wth axs-convex constrants. Thus, Algorthm Feasblty and Algorthm Approxmate can be appled to G2VIP systems. COROLLARY 7. A feasble soluton, f such a soluton exsts, can be found for G2VIP systems n tme O(mU) and space O(m). COROLLARY 8. A 2-approxmaton, f a feasble soluton exsts, can be found for G2VIP systems n tme O(nmU) and space O(m) Generalzed Weght Functons. In the orgnal defnton of a 2VIP system we used lnear weght functons (w x for every varable x ). The followng defnton generalzes the weght functon of a varable x : DEFINITION 5. A nonnegatve weght functon ω s called a monotone weght functon wth respect to an nterval f ω(α) ω(α + 1) for all α Z n the nterval. We assume, wthout loss of generalty, that a monotone weght functon s monotone over the real numbers and s nvertble. If t s not we can always defne ω (z) = ω( z )+ (ω( z ) ω( z )) (z z ). For a lnear weght functon ω we get that α 0 ω (α+1) ω (α) = w. For a general monotone weght functon ths s not necessarly the case. Therefore, n order to make Algorthm Approxmate applcable to such weght functons we replace (x, ˆl, û )w n the defnton of E2VIP systems by ω (x ) ω ( ˆl ), x [ ˆl, û ]. (x, ˆl, û ) = ω (û ) ω ( ˆl ), x > û, 0, x < ˆl. Now Algorthm Approxmate s applcable as s. COROLLARY 9. A 2-approxmaton, f a feasble soluton exsts, can be found for systems wth monotone weght functons n tme O(nmU) and space O(m). REMARK 1. If x { s,1,...,s,n }, we can defne a new varable x {0,...,n 1} and a new monotone weght functon w (α) = w (s,α+1 ). Thus, we get that the same results hold for l = 0 and u = n 1.

13 Effcent Algorthms for Integer Programs wth Two Varables per Constrant Optmzaton Algorthm for Monotone Systems. In ths secton we show how to mprove the complexty of Hochbaum and Naor s [12] optmzaton algorthm for a specfc case of 2VIP systems called monotone systems: DEFINITION 6. A monotone nequalty s an nequalty on two varables wth coeffcents of opposte sgns. A system wth two varables per nequalty s called monotone f all the constrants n the system are monotone. We formulate the followng monotone system: (M2VIP) mn n =1 w x s.t. a k x k b k x k c k, k {1,...,m}, x [l, u ], {1,...,n}, where 1 k, k n, w 0, a, b N n, c Z n, and l, u N. Lagaras [13] proved that decdng whether a monotone nteger program has a feasble soluton s NP-complete. Hochbaum and Naor [12] showed that the set of all feasble solutons of a monotone system form a dstrbuton lattce (ths has been observed before by Venott [16]). Usng ths property they presented an O(mn 2 log m + nmu) tme feasblty algorthm for monotone systems, whch fnds the top (or bottom) of ths lattce. They also presented an O(nmU 2 log(n 2 U/m)) tme optmzaton algorthm for such systems wth general (possbly negatve) weghts. By lmtng the dscusson to nonnegatve weghts, one can construct an O(mU) tme optmzaton algorthm: use the transformaton to a 2SAT nstance [11], then assgn 0 to all boolean varables whch dd not get an assgnment by the transformaton. As before, ths transformaton uses O(mU) space. In Fgure 5 we present an O(mU) tme and O(m) space optmalty algorthm for monotone systems wth nonnegatve weghts. Ths algorthm s actually a feasblty algorthm whch fnds the bottom of the feasble solutons lattce. 7 Due to the nonnegatve weghts, the bottom of the lattce s also an optmal soluton. THEOREM 10. An optmal soluton can be found for nonnegatve monotone systems n tme O(mU) and space O(m). Fg. 5. Optmzaton algorthm for monotone systems. 7 Note that when gven l as the ntal value, the feasblty algorthm from [12] can be mplemented n tme O(mU) and space O(m).

14 608 R. Bar-Yehuda and D. Rawtz PROOF. We prove that Algorthm Monotone fnds an optmal soluton for M2VIP systems wth nonnegatve weghts n tme O(mU) and space O(m). It s easy to prove by nducton usng Lemma 1 that sat(l after, u after ) = sat(l, u), where l after and u after are the values of l and u after the loop. Thus, f one of the calls to OneOnAllImpact fals, then no feasble soluton exsts. The lemma also mples that f the nstance of M2VIP s satsfable, then the above algorthm termnates wthout falure. We now prove that x = l s a feasble soluton by showng that every constrant k: a k x k b k x k c k s satsfed by x = l. We examne the value of the lower bound of x k after the last call of OneOnOneImpact(l, u, k, k, k) durng the algorthm (the call s made at least once). After ths call we get l k (c k + b k l k )/a k. l k wll not change from ths pont untl the algorthm termnates (as a change of l k mples another call), and other lower bounds only ncrease durng the executon algorthm, thus l k,l k must satsfy constrant k. Due to Lemma 1, we get that x = l s an optmal soluton (and t s the only optmal soluton f w > 0). Based on the same arguments as n Theorem 6 we get that the tme complexty of fndng an optmal soluton for a monotone system s O( n =1 m (u l )) = O(mU). The algorthm uses an ncdence lst data structure and therefore uses lnear space. As n the prevous secton we try to extend our approach to some nonlnear systems. We use an mportant property of monotone constrants: DEFINITION 7. Let constrant k be an axs-convex constrant on x and x. Constrant k s called a monotone axs-convex constrant f for every (α 1,α 2 ), (β 1,β 2 ) constrant(k) also (mn{α 1,β 1 }, mn{α 2,β 2 }) constrant(k). A system of monotone axs-convex constrants s called a generalzed monotone system. By usng smlar arguments to those used n the prevous secton we get: COROLLARY 11. An optmal soluton can be found for generalzed monotone systems wth monotone weght functons n tme O(mU) and space O(m). Acknowledgments. We thank Ar Freund, Nv Glboa, Avgal Orn, and especally Hadas Heer for ther careful readng and suggestons. References [1] B. Aspvall and Y. Shloach. A polynomal tme algorthm for solvng systems of lnear nequaltes wth two varables per nequalty. SIAM Journal on Computng, 9: , [2] R. Bar-Yehuda. One for the prce of two: a unfed approach for approxmatng coverng problems. In Proceedngs of the 1st Internatonal Workshop on Approxmaton Algorthms for Combnatoral Optmzaton Problems, pages 49 62, July Also n Algorthmca, 27: , [3] R. Bar-Yehuda and S. Even. A lnear tme approxmaton algorthm for the weghted vertex cover problem. Journal of Algorthms, 2: , 1981.

15 Effcent Algorthms for Integer Programs wth Two Varables per Constrant 609 [4] R. Bar-Yehuda and S. Even. A local-rato theorem for approxmatng the weghted vertex cover problem. Annals of Dscrete Mathematcs, 25:27 46, [5] T. H. Cormen, C. E. Leserson, and R. L. Rvest. Introducton to Algorthms. The MIT Press, Cambrdge, MA, [6] S. Even, A. Ita, and A. Shamr. On the complexty of tmetable and mult-commodty flow problems. SIAM Journal on Computng, 5(4): , [7] M. R. Garey and D. S. Johnson. Computers and Intractablty; A Gude to the Theory of NP- Completeness. Freeman, San Francsco, CA, [8] A. V. Goldberg and R. E. Taran. A new approach to the maxmum flow problem. Journal of the ACM, 35: , [9] D. Gusfeld and L. Ptt. A bounded approxmaton for the mnmum cost 2-SAT problem. Algorthmca, 8: , [10] D. S. Hochbaum. Approxmaton algorthms for the set coverng and vertex cover problems. SIAM Journal on Computng, 11(3): , [11] D. S. Hochbaum, N. Megddo, J. Naor, and A. Tamr. Tght bounds and 2-approxmaton algorthms for nteger programs wth two varables per nequalty. Mathematcal Programmng, 62:69 83, [12] D. S. Hochbaum and J. Naor. Smple and fast algorthms for lnear and nteger programs wth two varables per nequalty. SIAM Journal on Computng, 23: , [13] J. C. Lagaras. The computatonal complexty of smultaneous dophantne approxmaton problems. SIAM Journal on Computng, 14: , [14] G. L. Nemhauser and L. E. Trotter. Vertex packngs: structural propertes and algorthms. Mathematcal Programmng, 8: , [15] R. Shostak. Decdng lnear nequaltes by computng loop resdues. Journal of the ACM, 28: , [16] A. F. Venott. Representaton of general and polyhedral subsemlattces and sublattces of product spaces. Lnear Algebra and ts Applcatons, 114/115: , 1989.

Problem Set 9 Solutions

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

More information

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

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

More information

The Minimum Universal Cost Flow in an Infeasible Flow Network

The 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 information

Calculation of time complexity (3%)

Calculation of time complexity (3%) Problem 1. (30%) Calculaton of tme complexty (3%) Gven n ctes, usng exhaust search to see every result takes O(n!). Calculaton of tme needed to solve the problem (2%) 40 ctes:40! dfferent tours 40 add

More information

MMA and GCMMA two methods for nonlinear optimization

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

More information

NP-Completeness : Proofs

NP-Completeness : Proofs NP-Completeness : Proofs Proof Methods A method to show a decson problem Π NP-complete s as follows. (1) Show Π NP. (2) Choose an NP-complete problem Π. (3) Show Π Π. A method to show an optmzaton problem

More information

Some modelling aspects for the Matlab implementation of MMA

Some modelling aspects for the Matlab implementation of MMA Some modellng aspects for the Matlab mplementaton of MMA Krster Svanberg krlle@math.kth.se Optmzaton and Systems Theory Department of Mathematcs KTH, SE 10044 Stockholm September 2004 1. Consdered optmzaton

More information

Errors for Linear Systems

Errors 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 information

CS : Algorithms and Uncertainty Lecture 17 Date: October 26, 2016

CS : Algorithms and Uncertainty Lecture 17 Date: October 26, 2016 CS 29-128: Algorthms and Uncertanty Lecture 17 Date: October 26, 2016 Instructor: Nkhl Bansal Scrbe: Mchael Denns 1 Introducton In ths lecture we wll be lookng nto the secretary problem, and an nterestng

More information

Generalized Linear Methods

Generalized Linear Methods Generalzed Lnear Methods 1 Introducton In the Ensemble Methods the general dea s that usng a combnaton of several weak learner one could make a better learner. More formally, assume that we have a set

More information

Lecture 20: Lift and Project, SDP Duality. Today we will study the Lift and Project method. Then we will prove the SDP duality theorem.

Lecture 20: Lift and Project, SDP Duality. Today we will study the Lift and Project method. Then we will prove the SDP duality theorem. prnceton u. sp 02 cos 598B: algorthms and complexty Lecture 20: Lft and Project, SDP Dualty Lecturer: Sanjeev Arora Scrbe:Yury Makarychev Today we wll study the Lft and Project method. Then we wll prove

More information

On the Multicriteria Integer Network Flow Problem

On 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 information

princeton univ. F 17 cos 521: Advanced Algorithm Design Lecture 7: LP Duality Lecturer: Matt Weinberg

princeton 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 information

Module 9. Lecture 6. Duality in Assignment Problems

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

More information

Min Cut, Fast Cut, Polynomial Identities

Min Cut, Fast Cut, Polynomial Identities Randomzed Algorthms, Summer 016 Mn Cut, Fast Cut, Polynomal Identtes Instructor: Thomas Kesselhem and Kurt Mehlhorn 1 Mn Cuts n Graphs Lecture (5 pages) Throughout ths secton, G = (V, E) s a mult-graph.

More information

O-line Temporary Tasks Assignment. Abstract. In this paper we consider the temporary tasks assignment

O-line Temporary Tasks Assignment. Abstract. In this paper we consider the temporary tasks assignment O-lne Temporary Tasks Assgnment Yoss Azar and Oded Regev Dept. of Computer Scence, Tel-Avv Unversty, Tel-Avv, 69978, Israel. azar@math.tau.ac.l??? Dept. of Computer Scence, Tel-Avv Unversty, Tel-Avv, 69978,

More information

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

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

More information

U.C. Berkeley CS294: Beyond Worst-Case Analysis Luca Trevisan September 5, 2017

U.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 information

Lecture 11. minimize. c j x j. j=1. 1 x j 0 j. +, b R m + and c R n +

Lecture 11. minimize. c j x j. j=1. 1 x j 0 j. +, b R m + and c R n + Topcs n Theoretcal Computer Scence May 4, 2015 Lecturer: Ola Svensson Lecture 11 Scrbes: Vncent Eggerlng, Smon Rodrguez 1 Introducton In the last lecture we covered the ellpsod method and ts applcaton

More information

Lecture 4: Constant Time SVD Approximation

Lecture 4: Constant Time SVD Approximation Spectral Algorthms and Representatons eb. 17, Mar. 3 and 8, 005 Lecture 4: Constant Tme SVD Approxmaton Lecturer: Santosh Vempala Scrbe: Jangzhuo Chen Ths topc conssts of three lectures 0/17, 03/03, 03/08),

More information

Maximizing the number of nonnegative subsets

Maximizing 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 information

Speeding up Computation of Scalar Multiplication in Elliptic Curve Cryptosystem

Speeding up Computation of Scalar Multiplication in Elliptic Curve Cryptosystem H.K. Pathak et. al. / (IJCSE) Internatonal Journal on Computer Scence and Engneerng Speedng up Computaton of Scalar Multplcaton n Ellptc Curve Cryptosystem H. K. Pathak Manju Sangh S.o.S n Computer scence

More information

Computing Correlated Equilibria in Multi-Player Games

Computing 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 information

Assortment Optimization under MNL

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

More information

Affine transformations and convexity

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

More information

Single-Facility Scheduling over Long Time Horizons by Logic-based Benders Decomposition

Single-Facility Scheduling over Long Time Horizons by Logic-based Benders Decomposition Sngle-Faclty Schedulng over Long Tme Horzons by Logc-based Benders Decomposton Elvn Coban and J. N. Hooker Tepper School of Busness, Carnege Mellon Unversty ecoban@andrew.cmu.edu, john@hooker.tepper.cmu.edu

More information

Combining Constraint Programming and Integer Programming

Combining Constraint Programming and Integer Programming Combnng Constrant Programmng and Integer Programmng GLOBAL CONSTRAINT OPTIMIZATION COMPONENT Specal Purpose Algorthm mn c T x +(x- 0 ) x( + ()) =1 x( - ()) =1 FILTERING ALGORITHM COST-BASED FILTERING ALGORITHM

More information

EEE 241: Linear Systems

EEE 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 information

VARIATION OF CONSTANT SUM CONSTRAINT FOR INTEGER MODEL WITH NON UNIFORM VARIABLES

VARIATION OF CONSTANT SUM CONSTRAINT FOR INTEGER MODEL WITH NON UNIFORM VARIABLES VARIATION OF CONSTANT SUM CONSTRAINT FOR INTEGER MODEL WITH NON UNIFORM VARIABLES BÂRZĂ, Slvu Faculty of Mathematcs-Informatcs Spru Haret Unversty barza_slvu@yahoo.com Abstract Ths paper wants to contnue

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

Yong Joon Ryang. 1. Introduction Consider the multicommodity transportation problem with convex quadratic cost function. 1 2 (x x0 ) T Q(x x 0 )

Yong 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 information

Finding Primitive Roots Pseudo-Deterministically

Finding Primitive Roots Pseudo-Deterministically Electronc Colloquum on Computatonal Complexty, Report No 207 (205) Fndng Prmtve Roots Pseudo-Determnstcally Ofer Grossman December 22, 205 Abstract Pseudo-determnstc algorthms are randomzed search algorthms

More information

A New Refinement of Jacobi Method for Solution of Linear System Equations AX=b

A New Refinement of Jacobi Method for Solution of Linear System Equations AX=b Int J Contemp Math Scences, Vol 3, 28, no 17, 819-827 A New Refnement of Jacob Method for Soluton of Lnear System Equatons AX=b F Naem Dafchah Department of Mathematcs, Faculty of Scences Unversty of Gulan,

More information

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

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

More information

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

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

More information

a b a In case b 0, a being divisible by b is the same as to say that

a b a In case b 0, a being divisible by b is the same as to say that Secton 6.2 Dvsblty among the ntegers An nteger a ε s dvsble by b ε f there s an nteger c ε such that a = bc. Note that s dvsble by any nteger b, snce = b. On the other hand, a s dvsble by only f a = :

More information

Lecture 2: Gram-Schmidt Vectors and the LLL Algorithm

Lecture 2: Gram-Schmidt Vectors and the LLL Algorithm NYU, Fall 2016 Lattces Mn Course Lecture 2: Gram-Schmdt Vectors and the LLL Algorthm Lecturer: Noah Stephens-Davdowtz 2.1 The Shortest Vector Problem In our last lecture, we consdered short solutons to

More information

Chapter Newton s Method

Chapter Newton s Method Chapter 9. Newton s Method After readng ths chapter, you should be able to:. Understand how Newton s method s dfferent from the Golden Secton Search method. Understand how Newton s method works 3. Solve

More information

The Number of Ways to Write n as a Sum of ` Regular Figurate Numbers

The Number of Ways to Write n as a Sum of ` Regular Figurate Numbers Syracuse Unversty SURFACE Syracuse Unversty Honors Program Capstone Projects Syracuse Unversty Honors Program Capstone Projects Sprng 5-1-01 The Number of Ways to Wrte n as a Sum of ` Regular Fgurate Numbers

More information

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

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

More information

COS 521: Advanced Algorithms Game Theory and Linear Programming

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

More information

Foundations of Arithmetic

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

More information

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

Difference Equations

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

More information

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

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

More information

Lecture 4: Universal Hash Functions/Streaming Cont d

Lecture 4: Universal Hash Functions/Streaming Cont d CSE 5: Desgn and Analyss of Algorthms I Sprng 06 Lecture 4: Unversal Hash Functons/Streamng Cont d Lecturer: Shayan Oves Gharan Aprl 6th Scrbe: Jacob Schreber Dsclamer: These notes have not been subjected

More information

6.854J / J Advanced Algorithms Fall 2008

6.854J / J Advanced Algorithms Fall 2008 MIT OpenCourseWare http://ocw.mt.edu 6.854J / 18.415J Advanced Algorthms Fall 2008 For nformaton about ctng these materals or our Terms of Use, vst: http://ocw.mt.edu/terms. 18.415/6.854 Advanced Algorthms

More information

A new construction of 3-separable matrices via an improved decoding of Macula s construction

A new construction of 3-separable matrices via an improved decoding of Macula s construction Dscrete Optmzaton 5 008 700 704 Contents lsts avalable at ScenceDrect Dscrete Optmzaton journal homepage: wwwelsevercom/locate/dsopt A new constructon of 3-separable matrces va an mproved decodng of Macula

More information

find (x): given element x, return the canonical element of the set containing x;

find (x): given element x, return the canonical element of the set containing x; COS 43 Sprng, 009 Dsjont Set Unon Problem: Mantan a collecton of dsjont sets. Two operatons: fnd the set contanng a gven element; unte two sets nto one (destructvely). Approach: Canoncal element method:

More information

Simultaneous Optimization of Berth Allocation, Quay Crane Assignment and Quay Crane Scheduling Problems in Container Terminals

Simultaneous 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 information

The Second Anti-Mathima on Game Theory

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

More information

The L(2, 1)-Labeling on -Product of Graphs

The L(2, 1)-Labeling on -Product of Graphs Annals of Pure and Appled Mathematcs Vol 0, No, 05, 9-39 ISSN: 79-087X (P, 79-0888(onlne Publshed on 7 Aprl 05 wwwresearchmathscorg Annals of The L(, -Labelng on -Product of Graphs P Pradhan and Kamesh

More information

The Order Relation and Trace Inequalities for. Hermitian Operators

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

More information

Complete subgraphs in multipartite graphs

Complete subgraphs in multipartite graphs Complete subgraphs n multpartte graphs FLORIAN PFENDER Unverstät Rostock, Insttut für Mathematk D-18057 Rostock, Germany Floran.Pfender@un-rostock.de Abstract Turán s Theorem states that every graph G

More information

More metrics on cartesian products

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

More information

Kernel Methods and SVMs Extension

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

More information

Perfect Competition and the Nash Bargaining Solution

Perfect 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 information

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

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

More information

Feature Selection: Part 1

Feature 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 information

Vickrey Auction VCG Combinatorial Auctions. Mechanism Design. Algorithms and Data Structures. Winter 2016

Vickrey Auction VCG Combinatorial Auctions. Mechanism Design. Algorithms and Data Structures. Winter 2016 Mechansm Desgn Algorthms and Data Structures Wnter 2016 1 / 39 Vckrey Aucton Vckrey-Clarke-Groves Mechansms Sngle-Mnded Combnatoral Auctons 2 / 39 Mechansm Desgn (wth Money) Set A of outcomes to choose

More information

Outline and Reading. Dynamic Programming. Dynamic Programming revealed. Computing Fibonacci. The General Dynamic Programming Technique

Outline and Reading. Dynamic Programming. Dynamic Programming revealed. Computing Fibonacci. The General Dynamic Programming Technique Outlne and Readng Dynamc Programmng The General Technque ( 5.3.2) -1 Knapsac Problem ( 5.3.3) Matrx Chan-Product ( 5.3.1) Dynamc Programmng verson 1.4 1 Dynamc Programmng verson 1.4 2 Dynamc Programmng

More information

Dynamic Programming. Preview. Dynamic Programming. Dynamic Programming. Dynamic Programming (Example: Fibonacci Sequence)

Dynamic Programming. Preview. Dynamic Programming. Dynamic Programming. Dynamic Programming (Example: Fibonacci Sequence) /24/27 Prevew Fbonacc Sequence Longest Common Subsequence Dynamc programmng s a method for solvng complex problems by breakng them down nto smpler sub-problems. It s applcable to problems exhbtng the propertes

More information

Introductory Cardinality Theory Alan Kaylor Cline

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

More information

Additional Codes using Finite Difference Method. 1 HJB Equation for Consumption-Saving Problem Without Uncertainty

Additional Codes using Finite Difference Method. 1 HJB Equation for Consumption-Saving Problem Without Uncertainty Addtonal Codes usng Fnte Dfference Method Benamn Moll 1 HJB Equaton for Consumpton-Savng Problem Wthout Uncertanty Before consderng the case wth stochastc ncome n http://www.prnceton.edu/~moll/ HACTproect/HACT_Numercal_Appendx.pdf,

More information

Approximate Smallest Enclosing Balls

Approximate Smallest Enclosing Balls Chapter 5 Approxmate Smallest Enclosng Balls 5. Boundng Volumes A boundng volume for a set S R d s a superset of S wth a smple shape, for example a box, a ball, or an ellpsod. Fgure 5.: Boundng boxes Q(P

More information

A PROBABILITY-DRIVEN SEARCH ALGORITHM FOR SOLVING MULTI-OBJECTIVE OPTIMIZATION PROBLEMS

A PROBABILITY-DRIVEN SEARCH ALGORITHM FOR SOLVING MULTI-OBJECTIVE OPTIMIZATION PROBLEMS HCMC Unversty of Pedagogy Thong Nguyen Huu et al. A PROBABILITY-DRIVEN SEARCH ALGORITHM FOR SOLVING MULTI-OBJECTIVE OPTIMIZATION PROBLEMS Thong Nguyen Huu and Hao Tran Van Department of mathematcs-nformaton,

More information

Exercises. 18 Algorithms

Exercises. 18 Algorithms 18 Algorthms Exercses 0.1. In each of the followng stuatons, ndcate whether f = O(g), or f = Ω(g), or both (n whch case f = Θ(g)). f(n) g(n) (a) n 100 n 200 (b) n 1/2 n 2/3 (c) 100n + log n n + (log n)

More information

For now, let us focus on a specific model of neurons. These are simplified from reality but can achieve remarkable results.

For now, let us focus on a specific model of neurons. These are simplified from reality but can achieve remarkable results. Neural Networks : Dervaton compled by Alvn Wan from Professor Jtendra Malk s lecture Ths type of computaton s called deep learnng and s the most popular method for many problems, such as computer vson

More information

Lecture 17: Lee-Sidford Barrier

Lecture 17: Lee-Sidford Barrier CSE 599: Interplay between Convex Optmzaton and Geometry Wnter 2018 Lecturer: Yn Tat Lee Lecture 17: Lee-Sdford Barrer Dsclamer: Please tell me any mstake you notced. In ths lecture, we talk about the

More information

Common loop optimizations. Example to improve locality. Why Dependence Analysis. Data Dependence in Loops. Goal is to find best schedule:

Common loop optimizations. Example to improve locality. Why Dependence Analysis. Data Dependence in Loops. Goal is to find best schedule: 15-745 Lecture 6 Data Dependence n Loops Copyrght Seth Goldsten, 2008 Based on sldes from Allen&Kennedy Lecture 6 15-745 2005-8 1 Common loop optmzatons Hostng of loop-nvarant computatons pre-compute before

More information

Global Optimization of Truss. Structure Design INFORMS J. N. Hooker. Tallys Yunes. Slide 1

Global Optimization of Truss. Structure Design INFORMS J. N. Hooker. Tallys Yunes. Slide 1 Slde 1 Global Optmzaton of Truss Structure Desgn J. N. Hooker Tallys Yunes INFORMS 2010 Truss Structure Desgn Select sze of each bar (possbly zero) to support the load whle mnmzng weght. Bar szes are dscrete.

More information

Lecture 4: November 17, Part 1 Single Buffer Management

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

More information

On the correction of the h-index for career length

On the correction of the h-index for career length 1 On the correcton of the h-ndex for career length by L. Egghe Unverstet Hasselt (UHasselt), Campus Depenbeek, Agoralaan, B-3590 Depenbeek, Belgum 1 and Unverstet Antwerpen (UA), IBW, Stadscampus, Venusstraat

More information

The internal structure of natural numbers and one method for the definition of large prime numbers

The internal structure of natural numbers and one method for the definition of large prime numbers The nternal structure of natural numbers and one method for the defnton of large prme numbers Emmanul Manousos APM Insttute for the Advancement of Physcs and Mathematcs 3 Poulou str. 53 Athens Greece Abstract

More information

Communication Complexity 16:198: February Lecture 4. x ij y ij

Communication Complexity 16:198: February Lecture 4. x ij y ij Communcaton Complexty 16:198:671 09 February 2010 Lecture 4 Lecturer: Troy Lee Scrbe: Rajat Mttal 1 Homework problem : Trbes We wll solve the thrd queston n the homework. The goal s to show that the nondetermnstc

More information

LOW BIAS INTEGRATED PATH ESTIMATORS. James M. Calvin

LOW BIAS INTEGRATED PATH ESTIMATORS. James M. Calvin Proceedngs of the 007 Wnter Smulaton Conference S G Henderson, B Bller, M-H Hseh, J Shortle, J D Tew, and R R Barton, eds LOW BIAS INTEGRATED PATH ESTIMATORS James M Calvn Department of Computer Scence

More information

Hongyi Miao, College of Science, Nanjing Forestry University, Nanjing ,China. (Received 20 June 2013, accepted 11 March 2014) I)ϕ (k)

Hongyi Miao, College of Science, Nanjing Forestry University, Nanjing ,China. (Received 20 June 2013, accepted 11 March 2014) I)ϕ (k) ISSN 1749-3889 (prnt), 1749-3897 (onlne) Internatonal Journal of Nonlnear Scence Vol.17(2014) No.2,pp.188-192 Modfed Block Jacob-Davdson Method for Solvng Large Sparse Egenproblems Hongy Mao, College of

More information

Annexes. EC.1. Cycle-base move illustration. EC.2. Problem Instances

Annexes. EC.1. Cycle-base move illustration. EC.2. Problem Instances ec Annexes Ths Annex frst llustrates a cycle-based move n the dynamc-block generaton tabu search. It then dsplays the characterstcs of the nstance sets, followed by detaled results of the parametercalbraton

More information

NUMERICAL DIFFERENTIATION

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

More information

arxiv: v1 [math.co] 12 Sep 2014

arxiv: v1 [math.co] 12 Sep 2014 arxv:1409.3707v1 [math.co] 12 Sep 2014 On the bnomal sums of Horadam sequence Nazmye Ylmaz and Necat Taskara Department of Mathematcs, Scence Faculty, Selcuk Unversty, 42075, Campus, Konya, Turkey March

More information

e - c o m p a n i o n

e - c o m p a n i o n OPERATIONS RESEARCH http://dxdoorg/0287/opre007ec e - c o m p a n o n ONLY AVAILABLE IN ELECTRONIC FORM 202 INFORMS Electronc Companon Generalzed Quantty Competton for Multple Products and Loss of Effcency

More information

Beyond Zudilin s Conjectured q-analog of Schmidt s problem

Beyond Zudilin s Conjectured q-analog of Schmidt s problem Beyond Zudln s Conectured q-analog of Schmdt s problem Thotsaporn Ae Thanatpanonda thotsaporn@gmalcom Mathematcs Subect Classfcaton: 11B65 33B99 Abstract Usng the methodology of (rgorous expermental mathematcs

More information

Welfare Properties of General Equilibrium. What can be said about optimality properties of resource allocation implied by general equilibrium?

Welfare 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 information

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

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

More information

arxiv: v1 [math.ho] 18 May 2008

arxiv: v1 [math.ho] 18 May 2008 Recurrence Formulas for Fbonacc Sums Adlson J. V. Brandão, João L. Martns 2 arxv:0805.2707v [math.ho] 8 May 2008 Abstract. In ths artcle we present a new recurrence formula for a fnte sum nvolvng the Fbonacc

More information

Global Sensitivity. Tuesday 20 th February, 2018

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

More information

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

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

More information

Stanford University CS359G: Graph Partitioning and Expanders Handout 4 Luca Trevisan January 13, 2011

Stanford University CS359G: Graph Partitioning and Expanders Handout 4 Luca Trevisan January 13, 2011 Stanford Unversty CS359G: Graph Parttonng and Expanders Handout 4 Luca Trevsan January 3, 0 Lecture 4 In whch we prove the dffcult drecton of Cheeger s nequalty. As n the past lectures, consder an undrected

More information

First day August 1, Problems and Solutions

First day August 1, Problems and Solutions FOURTH INTERNATIONAL COMPETITION FOR UNIVERSITY STUDENTS IN MATHEMATICS July 30 August 4, 997, Plovdv, BULGARIA Frst day August, 997 Problems and Solutons Problem. Let {ε n } n= be a sequence of postve

More information

Lecture 10: May 6, 2013

Lecture 10: May 6, 2013 TTIC/CMSC 31150 Mathematcal Toolkt Sprng 013 Madhur Tulsan Lecture 10: May 6, 013 Scrbe: Wenje Luo In today s lecture, we manly talked about random walk on graphs and ntroduce the concept of graph expander,

More information

Xiangwen Li. March 8th and March 13th, 2001

Xiangwen Li. March 8th and March 13th, 2001 CS49I Approxaton Algorths The Vertex-Cover Proble Lecture Notes Xangwen L March 8th and March 3th, 00 Absolute Approxaton Gven an optzaton proble P, an algorth A s an approxaton algorth for P f, for an

More information

Lecture Space-Bounded Derandomization

Lecture Space-Bounded Derandomization Notes on Complexty Theory Last updated: October, 2008 Jonathan Katz Lecture Space-Bounded Derandomzaton 1 Space-Bounded Derandomzaton We now dscuss derandomzaton of space-bounded algorthms. Here non-trval

More information

Spectral Graph Theory and its Applications September 16, Lecture 5

Spectral Graph Theory and its Applications September 16, Lecture 5 Spectral Graph Theory and ts Applcatons September 16, 2004 Lecturer: Danel A. Spelman Lecture 5 5.1 Introducton In ths lecture, we wll prove the followng theorem: Theorem 5.1.1. Let G be a planar graph

More information

The Expectation-Maximization Algorithm

The Expectation-Maximization Algorithm The Expectaton-Maxmaton Algorthm Charles Elan elan@cs.ucsd.edu November 16, 2007 Ths chapter explans the EM algorthm at multple levels of generalty. Secton 1 gves the standard hgh-level verson of the algorthm.

More information

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

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

More information

Lecture 12: Discrete Laplacian

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

More information

Greedy can also beat pure dynamic programming

Greedy can also beat pure dynamic programming Electronc Colloquum on Computatonal Complexty, Report No. 49 (208) Greedy can also beat pure dynamc programmng Stasys Jukna Hannes Sewert Insttut für Informatk, Goethe Unverstät Frankfurt am Man, Germany

More information

Section 8.3 Polar Form of Complex Numbers

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

More information

On the Repeating Group Finding Problem

On the Repeating Group Finding Problem The 9th Workshop on Combnatoral Mathematcs and Computaton Theory On the Repeatng Group Fndng Problem Bo-Ren Kung, Wen-Hsen Chen, R.C.T Lee Graduate Insttute of Informaton Technology and Management Takmng

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