McMaster University. Advanced Optimization Laboratory. Title: How good are interior point methods?

Size: px
Start display at page:

Download "McMaster University. Advanced Optimization Laboratory. Title: How good are interior point methods?"

Transcription

1 McMaster Uiversity Advaced Optimizatio Laboratory Title: How good are iterior poit methods? Klee-Mity cubes tighte iteratio-complexity bouds Authors: Atoie Deza, Eissa Nematollahi ad Tamás Terlaky AdvOl-Report No. 2004/20 December 2004, Hamilto, Otario, Caada

2 How good are iterior poit methods? Klee-Mity cubes tighte iteratio-complexity bouds. Dedicated to Professor Emil Klafszky o the occasio of his 70th birthday Atoie Deza Eissa Nematollahi Tamás Terlaky December 3, 2004 Abstract By refiig a variat of the Klee-Mity example that forces the cetral path to visit all the vertices of the Klee-Mity -cube, we exhibit a early worst-case example for pathfollowig iterior poit methods. Namely, while the theoretical iteratio-complexity upper boud is O(2 5 2), we prove that solvig this -dimesioal liear optimizatio problem requires at least 2 1 iteratios. Key words: Liear programmig, iterior poit method, worst-case iteratio-complexity. MSC2000 Subject Classificatio: Primary: 90C05; Secodary: 90C51, 90C27, 52B12 1 Itroductio While the simplex method, itroduced by Datzig [1], works very well i practice for liear optimizatio problems, i 1972 Klee ad Mity [5] gave a example for which the simplex method takes a expoetial umber of iteratios. More precisely, they cosidered a maximizatio problem over a -dimesioal squashed cube ad proved that a variat of the simplex method visits all its 2 vertices. Thus, the time complexity is ot polyomial for the worst case, as 2 1 iteratios are ecessary for this -dimesioal liear optimizatio problem. The pivot rule used i the Klee-Mity example was the most egative reduced cost but variats of the Klee-Mity -cube allow to prove expoetial ruig time for most pivot rules; see [10] ad the refereces therei. The Klee-Mity worst-case example partially stimulated the search for a polyomial algorithm ad, i 1979, Khachiya s [4] ellipsoid method proved that liear programmig is ideed polyomially solvable. I 1984, Karmarkar [3] proposed a more efficiet polyomial algorithm that sparked the research o polyomial iterior poit methods. I short, while the simplex method goes alog the edges of the polyhedro correspodig to the feasible regio, iterior poit methods pass through the iterior of this polyhedro. Startig at the aalytic ceter, most iterior poit methods follow the so-called cetral path ad coverge to the aalytic ceter of the optimal face; see e.g., [6, 8, 9, 13, 14]. I 2004, Deza, Nematollahi, Peyghami ad Terlaky [2] showed that, by carefully addig a expoetial umber of redudat costraits to the Klee-Mity -cube, the cetral path ca be severely distorted. Specifically, they provided a example for which path-followig iterior poit methods have to take 2 2 sharp turs as the cetral path passes withi a arbitrarily small eighborhood of the correspodig vertices of 2

3 Atoie Deza, Eissa Nematollahi ad Tamás Terlaky 3 the Klee-Mity cube before covergig to the optimal solutio. This example yields a theoretical lower boud for the umber of iteratios eeded for path-followig iterior poit methods: The umber of iteratios is at least the umber of sharp turs; that is, the iteratio-complexity lower boud is Ω(2 ). O the other had, the theoretical iteratio-complexity upper boud is O( NL) where N ad L respectively deote the umber of costraits ad the bit-legth of the iput-data. The iteratio-complexity upper boud for the highly redudat Klee-Mity -cube of [2] is O(2 3 L) = O(2 9 4 ), as N = O(2 6 2 ) ad L = O(2 6 3 ) for this example. Therefore, these 2 1 sharp turs yield a Ω( 6 N ) iteratio-complexity lower boud. I this paper we show that a refied problem with the same Ω(2 ) iteratio-complexity lower boud exhibits a early worst-case iteratio-complexity as the complexity upper boud is O(2 2). 5 I other words, this ew example, with N = O(2 2 3 ), essetially closes the iteratio-complexity gap with a Ω( l 2 N N l 3 N ) lower boud ad a O( N l N) upper boud. 2 Notatios ad the Mai Results We cosider the followig Klee-Mity variat where ε is a small positive factor by which the uit cube [0,1] is squashed. mi x subject to 0 x 1 1 εx k 1 x k 1 εx k 1 for k = 2,...,. The above miimizatio problem has 2 costraits, variables ad the feasible regio is a -dimesioal cube deoted by C. Some variats of the simplex method take 2 1 iteratios to solve this problem as they visit all the vertices ordered by the decreasig value of the last coordiate x startig from v {} = (0,...,0,1) till the optimal value x = 0 is reached at the origi v. While addig a set h of redudat iequalities does ot chage the feasible regio, the aalytic ceter χ h ad the cetral path are affected by the additio of redudat costraits. We cosider redudat iequalities iduced by hyperplaes parallel to the facets of C cotaiig the origi. The costrait parallel to the facet H 1 : x 1 = 0 is added h 1 times at a distace ad the costrait parallel to the facet H k : x k = εx k 1 is added h k times at a distace d k for k = 2,...,. The set h is deoted by the iteger-vector h = (h 1,...,h ), d = (,...,d ), ad the redudat liear optimizatio problem is defied by mi x subject to 0 x 1 1 εx k 1 x k 1 εx k 1 for k = 2,..., 0 + x 1 repeated h 1 times εx 1 d 2 + x 2 repeated h 2 times. εx 1 d + x. repeated h times.

4 4 How good are iterior poit methods? By aalogy with the uit cube [0,1], we deote the vertices of the Klee-Mity cube C by usig a subset S of {1,...,}. For S {1,...,}, a vertex v S of C is defied by v S 1 = v S k = { 1, if 1 S 0, otherwise { 1 εv S k 1, if k S εvk 1 S, otherwise k = 2,...,. The δ-eighborhood N δ (v S ) of a vertex v S is defied, with the covetio x 0 = 0, by N δ (v S ) = { { 1 xk εx x C : k 1 ε k 1 δ, if k S x k εx k 1 ε k 1 δ, otherwise } k = 1,...,. v {2} v {1,2} v {1} v Figure 1: The δ-eighborhoods of the 4 vertices of the Klee-Mity 2-cube. I this paper we focus o the followig problem C δ defied by ε = 2(+1), d = (2 +4,...,2 k+5,...,2 5 ), h = ( 22+8 (+1) δ 1 where 0 < δ 1 4(+1). 2+7 (+1) δ,..., 22+8 (+1) +k 1 δ +k 2 2+k+6 (+1) 2k 1 δ 2k 2 ),..., (+1) 2 1 δ, 2 2 Note that we have: ε+δ < 1 2 ; that is, the δ-eighborhoods of the 2 vertices are o-overlappig, ad that h is, up to a floor operatio, liearly depedet o δ 1. Propositio 2.1 states that, for C δ, the cetral path takes at least 2 2 turs before covergig to the origi as it passes through the δ-eighborhood of all the 2 vertices of the Klee-Mity -cube; See Sectio 3.2 for the proof. Note that the proof give i Sectio 3.2 yields a slightly stroger result tha Propositio 2.1: I additio to pass through the δ-eighborhood of all the vertices, the cetral path is bet alog the edge-path followed by the simplex method. We set δ = 1 4(+1) i Propositios 2.3 ad 2.4 i order to exhibit the sharpest bouds. The correspodig liear optimizatio problem C 1/4(+1) depeds oly o the dimesio.

5 Atoie Deza, Eissa Nematollahi ad Tamás Terlaky 5 Propositio 2.1. The cetral path P of C δ -cube. itersects the δ-eighborhood of each vertex of the Sice the umber of iteratios required by path-followig iterior poit methods is at least the umber of sharp turs, Propositio 2.1 yields a theoretical lower boud for the iteratiocomplexity for solvig this -dimesioal liear optimizatio problem. Corollary 2.2. For C δ, the iteratio-complexity lower boud of path-followig iterior poit methods is Ω(2 ). Sice the theoretical iteratio-complexity upper boud for path-followig iterior poit methods is O( NL), where N ad L respectively deote the umber of costraits ad the bit-legth of the iput-data, we have: Propositio 2.3. For C 1/4(+1), the iteratio-complexity upper boud of path-followig iterior poit methods is O(2 2L); 3 that is, O( ). Proof. We have N = 2 + k=1 h k = 2 + ( k= ( ) +1 +k 2 +k+8 ( ) ) +1 2k ad, sice ) +k k=1 e 2, we have N = O(2 2 3 ) ad L N l = O(2 2 4 ). ( +1 Noticig that the oly two vertices with last coordiates smaller tha or equal to ε 1 are v ad v {1}, with v = 0 ad v {1} = ε 1, the stoppig criterio ca be replaced by: stoppig duality gap smaller tha ε with the correspodig cetral path parameter at the stoppig poit beig µ = ε N. Additioally, oe ca check that by settig the cetral path parameter to µ0 = 1, we obtai a startig poit which belogs to the iterior of the δ-eighborhood of the highest vertex v {}, see Sectio 3.3 for a detailed proof. I other words, a path-followig algorithm usig a stadard ǫ-precisio as stoppig criterio ca stop whe the duality gap is smaller tha ε as the optimal vertex is idetified, see [8]. The correspodig iteratio-complexity boud O( N log N ǫ ) yields, for our costructio, a precisio-idepedet iteratio-complexity O( N l Nµ0 Nµ ) = O( N) ad Propositio 2.3 ca therefore be stregtheed to: Propositio 2.4. For C 1/4(+1), the iteratio-complexity upper boud of path-followig iterior poit methods is O(2 5 2). Remark 2.5. (i) For C 1/4(+1), by Corollary 2.2 ad Propositio 2.4, the order of the iteratio-complexity of path-followig iterior poit methods is betwee 2 ad or, equivaletly, betwee N l 3 N ad N l N. (ii) The k-th coordiate of the vector d correspods to the scalar d defied i [2] for dimesio k + 3. (iii) Other settigs for d ad h esurig that the cetral path visits all the vertices of the Klee-Mity -cube are possible. For example, d ca be set to (1.1,22) i dimesio 2. (iv) Our results apply to path-followig iterior poit methods but ot to other iterior poit methods such as Karmarkar s origial projective algorithm [3].

6 6 How good are iterior poit methods? Remark 2.6. (i) Megiddo ad Schub [7] proved, for affie scalig trajectories, a result with a similar flavor as our result for the cetral path, ad oted that their approach does ot exted to projective scalig. They cosidered the o-redudat Klee-Mity cube. (ii) Todd ad Ye [11] gave a Ω( 3 N) iteratio-complexity lower boud betwee two updates of the cetral path parameter µ. (iii) Vavasis ad Ye [12] provided a O(N 2 ) upper boud for the umber of approximately straight segmets of the cetral path. 3 Proofs of Propositio 2.1 ad Propositio Prelimiary Lemmas Lemma 3.1. With b = 4 δ (1,...,1), ε = ( h = (+1) 2+7 (+1) δ 1 δ,..., 22+8 (+1) +k 1 δ +k 2 we have A h 3b 2. A = ε 2(+1), d = (2+4,...,2 k+5,...,2 5 ), 2+k+6 (+1) 2k 1,..., (+1) 2 1 δ 2k 2 d ε d 2 +1 ε 2 d ε 0 0 k 1 ε k d k +1 d k ε d 1 +1 ε 1 d 1 2ε d +1 δ 2 2 ) Proof. As ε = 2(+1) ad d = (2+4,...,2 k+5,...,2 5 ), h ca be rewritte as ( ) h = 4 δ ( 4 ε 1 ε ),..., d k ( 4 ε k 1 ε 1 d ),..., 3 ε k ε 1 ε ad A h 3b 2 ca be rewritte as 4 δ ( 4 ε 1 ε ) + 4 δ which is equivalet to ( 1 ε 2 1 ε 3 2 ( 1 ε k+1 2 ε k + 1 ε δ + 1 ( 4 ε 1 ε ) 4 δ ( 4 ε 1 ε 2) 2d k d k + 1 ( 4 ε 1 ε k ) 4 δ ( 4 ε 1 ε k+1) ( 2 ε + 1 ε δ ( 4 ε 1 ε ) + 4 δ ) ) 4 ε 1 ε δ 6 δ 2d 3 d + 1 ε 6 δ, d k 8 ε 1 ε k+1 1 ε ) d 4 ε 1 ε , for k = 2,..., 1 for k = 2,..., 1 ad

7 Atoie Deza, Eissa Nematollahi ad Tamás Terlaky 7 As 1 ε 1 2 ε , 1 ε 3 2 0, 1 ε a + 1 ε k+1 ε 3 2 0, the above system is implied by ( ε 1 ε k+1 2 ) ε k d k 8 ε for k = 2,..., 1 2 ε d 4 ε, 1 as 2 = 2 a = 2 k ( ) k ε k+1 ε k ε k ε k 2 k+2, the above system is implied by 2 +5 d k 2 k+4 for k = 2,..., 1 d 2 which is true sice d = (2 +4,...,2 k+5,...,2 5 ). Corollary 3.2. With the same assumptios as i Lemma 3.1 ad h = h, we have Ah b. Proof. Sice 0 h k h k < 1 ad d k = 2 k+5, we have: h 1 h 1 + 2( h k h k )ε k 1 d k +1 h 1 h 1 h 1 h 1 +1 ( h 2 h 2 )ε d 2 2 δ ( h k+1 h k+1 )ε k d k+1 2 δ for k = 2,..., 1 + 2( h h )ε 1 d +1 2 δ thus, A( h h) b 2, which implies, sice A h 3b 2 by Lemma 3.1, that Ah b. Corollary 3.3. With the same assumptios as i Lemma 3.1 ad h = h, we have: h kε k 1 d k +1 h k+1 ε k d k δ for k = 1,..., 1. Proof. For k = 1,..., 1, oe ca easily check that the first k iequalities of Ah b imply h k ε k 1 d k +1 h k+1ε k d k δ. The aalytic ceter χ = (ξ1,...,ξ ) of C δ is the uique solutio to the problem cosistig of maximizig the product of the slack variables: s 1 = x 1 s k = x k εx k 1 for k = 2,..., s 1 = 1 x 1 s k = 1 εx k 1 x k for k = 2,..., s 1 = + s 1 repeated h 1 times. s = d + s. repeated h times.

8 8 How good are iterior poit methods? Equivaletly, χ is the solutio of the followig maximizatio problem: max x i.e., with the covetio x 0 = 0, max x k=1 (l s k + l s k + h k l s k ), k=1 ( ) l(x k εx k 1 ) + l(1 εx k 1 x k ) + h k l(d k + x k εx k 1 ). The optimality coditios (the gradiet is equal to zero at optimality) for this cocave maximizatio problem give: where 1 σ k ε σk+1 1 σ k ε σ k+1 + h k σ k h k+1ε σ k+1 1 σ 1 σ + h = 0 for k = 1,..., 1 σ = 0 (1) σk > 0, σ k > 0, σ k > 0 for k = 1,...,, σ 1 = ξ 1 σ k = ξ k εξ k 1 for k = 2,..., σ 1 = 1 ξ 1 σ k = 1 εξ k 1 ξ k for k = 2,..., σ k = d k + σ k for k = 1,...,. The followig lemma states that, for C δ, the aalytic ceter χ belogs to the eighborhood of the vertex v {} = (0,...,0,1). Lemma 3.4. For C δ, we have: χ N δ (v {} ). Proof. Addig the -th equatio of (1) multiplied by ε 1 to the j-th equatio of (1) multiplied by ε j 1 for j = k..., 1, we have, for k = 1,..., 1, ε k 1 σ k εk 1 σ k 2ε 1 σ 2 2 i=k ε i σ i+1 + h kε k 1 σ k 2h ε 1 σ = 0, implyig: 2h ε 1 σ h kε k 1 σ k = εk 1 σ k ( ε k 1 σ k + 2ε 1 σ i=k ε i σ i+1 ) εk 1 σk, which implies, sice σ d + 1, σ k d k ad h 1 h kε k 1 d k by Corollary 3.3, 2h ε 1 d + 1 h 1 εk 1, σ k

9 Atoie Deza, Eissa Nematollahi ad Tamás Terlaky 9 implyig, sice 2hε 1 d +1 h 1 -th equatio of (1) implies: h ε 1 σ 1 δ by Corollary 3.2, σ k εk 1 δ for k = 1,..., 1. The ; that is, sice σ < d + 1 ad hε 1 d +1 1 δ by ε 1 σ Corollary 3.2, we have: 1 δ hε 1 d +1 ε 1 σ, implyig: σ ε 1 δ. The cetral path P of C δ ca be defied as the set of aalytic ceters χ (α) = (x 1,...,x 1,α) of the itersectio of the hyperplae H α : x = α with the feasible regio of C δ where 0 < α ξ, see [8]. These itersectios Ω(α) are called the level sets ad χ (α) is the solutio of the followig system: where { 1 s k ε s k+1 1 s k ε s + h k k+1 s h k+1ε k s = 0 for k = 1,..., 1 k+1 s k > 0, s k > 0, s k > 0 for k = 1,..., 1, (2) s 1 = x 1 s k = x k εx k 1 for k = 2,..., 1 s = α εx 1 s 1 = 1 x 1 s k = 1 εx k 1 x k for k = 2,..., 1 s = 1 α εx 1 s k = d k + s k for k = 1,...,. Lemma 3.5. For C δ, Ck δ = {x C : s k ε k 1 δ,s k ε k 1 δ} ad Ĉk δ = {x C : s k 1 ε k 2 δ,s k 2 ε k 3 δ,...,s 1 δ}, we have: Cδ k P Ĉk δ for k = 2,...,. Proof. Let x Cδ k P. Addig the (k 1)-st equatio of (2) multiplied by εk 2 to the i-th equatio of (1) multiplied by ε i 1 for i = j...,k 2, we have, for k = 2,..., 1, 2h k 1ε k 2 s + h jε j 1 k 1 s + h kε k 1 j s + εj 1 k s + εk 1 j s + εk 1 k s 2εk 2 k s + εj 1 k 3 ε i k 1 s + 2 j s = 0, i+1 which implies, sice s k 1 < d k 1 + 1, s j > d j, s k > d k ad s k εk 1 δ ad s k εk 1 δ as x C k δ, 2h k 1 ε k 2 d k h jε j 1 h kε k 1 d j d k implyig, sice h 1 h jε j 1 d j by Corollary 3.3, εj 1 s j + 2 δ, i=j h 1 + 2h k 1ε k 2 d k h kε k 1 d k εj 1 s j + 2 δ, that is, as 3 δ h 1 + 2h k 1ε k 2 d k 1 +1 h kε k 1 d k equatio of (2), we have by Corollary 3.2: s j εj 1 δ. Cosiderig the (k 1)-st h k 1 ε k 2 s k 1 h kε k 1 s k = εk 2 s + εk 1 k 1 s + εk 1 k s k εk 2 s, k 1

10 10 How good are iterior poit methods? which implies, sice s k 1 < d k 1 + 1, s k > d k ad s k εk 1 δ ad s k εk 1 δ as x C k δ, h k 1 ε k 2 d k h kε k 1 d k which implies, sice 3 δ h k 1ε k 2 d k 1 +1 h kε k 1 d k x Ĉk δ. εk 2 s + 2 k 1 δ, by Corollary 3.3, that s k 1 εk 2 δ ad, therefore, 3.2 Proof of Propositio 2.1 Figure 2: The set P δ for the Klee-Mity 3-cube. For k = 2,...,, while Cδ k, defied i Lemma 3.5, ca be see as the cetral part of the cube C, the sets Tδ k = {x C : s k ε k 1 δ} ad Bδ k = {x C : s k ε k 1 δ}, ca be see, respectively, as the top ad bottom part of C. Clearly, we have C = Tδ k Ck δ Bk δ for each k = 2,...,. Usig the set Ĉk δ defied i Lemma 3.5, we cosider the set Ak δ = T δ k Ĉk δ Bk δ for k = 2...,, ad, for 0 < δ 1 4(+1), we show that the set P δ = k=2 Ak δ, see Figure 2, cotais the cetral path P. By Lemma 3.4, the startig poit χ of P belogs to N δ (v {} ). Sice P C ad C = k=2 (T δ k Ck δ Bk δ ), we have: P = C P = (Tδ k Ck δ Bk δ ) P = (Tδ k (Ck δ P) Bk δ ) P, k=2 k=2 that is, by Lemma 3.5, P (Tδ k Ĉk δ Bk δ ) = A k δ = P δ k=2 k=2

11 Atoie Deza, Eissa Nematollahi ad Tamás Terlaky 11 T 3 0 B 2 0 Ĉ 2 0 T 2 0 Ĉ 3 0 B 3 0 A 2 0 A 3 0 Figure 3: The sets A 2 0 ad A3 0 for the Klee-Mity 3-cube. Remark that the sets Cδ k,ĉk δ,t δ k,bk δ ad Ak δ ca be defied for δ = 0, see Figure 3, ad that the correspodig set P 0 = k=2 Ak 0 is precisely the path followed by the simplex method o the origial Klee-Mity problem as it pivots alog the edges of C. The set P δ is a δ-sized (cross sectio) tube alog the path P 0. See Figure 4 illustratig how P 0 starts at v {}, decreases with respect to the last coordiate x ad eds at v. v {3} v {1,3} v {1,2,3} v {2,3} v {2} v v {1} v {1,2} Figure 4: The path P 0 followed by the simplex method for the Klee-Mity 3-cube. 3.3 Proof of Propositio 2.4 We cosider the poit x of the cetral path which lies o the boudary of the δ-eighborhood of the highest vertex v {}. This poit is defied by: s 1 = δ,s k ε k 1 δ for k = 2,..., 1 ad s 2 ε δ. Note that the otatio s k for the cetral path (perturbed complemetarity) coditios, y k s k = µ for k = 1,...,p, is cosistet with the slacks itroduced after Corollary 3.3 with s +k = s k for k = 1,..., ad s pi +k = s k for k = 1,...,h i+1 ad i = 0,..., 1. Let µ deote the cetral path parameter correspodig to x. I the followig, we prove that µ ε 1 δ which implies that ay poit of the cetral path with correspodig parameter µ µ belog to the iterior of the δ-eighborhood of the highest vertex v {}. I particular, it implies

12 12 How good are iterior poit methods? that by settig the cetral path parameter to µ 0 = 1, we obtai a startig poit which belogs to the iterior of the δ-eighborhood of the vertex v {}. Estimatio of the cetral path parameter µ: The formulatio of the dual problem of C δ is: max z = 2 p k y k d k y i k=+1 k=1 i=p k 1 +1 subject to y k εy k+1 y +k εy +k+1 + p k i=p k 1 +1 y i ε y y 2 + where p 0 = 2 ad p k = 2 + h h k for k = 1,...,. p k+1 i=p k +1 p i=p 1 +1 y i = 0 for k = 1,..., 1 y i = 1 y k 0 for k = 1,...,p, For k = 1,...,, multiplyig by ε k 1 the k-th equatio of the above dual costraits ad summig the up, we have: which implies y 1 y +1 2 ( 2+h εy +2 + ε 2 y ε 1 ) 1 y 2 + 2ε 1 y 2 y h 1 i=2+1 y i i=2+1 implyig, sice for i = 2 + 1,...,2 + h 1, s i yields y i µ, that 2ε 1 y 2 y 1 + µh 1 = µ δ + µh 1. y i = ε 1 Sice for i = p 1 + 1,...,p, s i = d + x ε x 1 d + 1 yields y i µ costrait implies p y 2 y i 1 µh d i=p 1 +1 ( ) which, combied with the previously obtaied iequality, gives µ 2hε 1 d +1 h 1 1 δ ad, sice Corollary 3.2 gives 2hε 1 d +1 h 1 1 δ 2 δ, we have µ ε 1 δ. d +1, the last dual 2ε 1, Ackowledgmets. We would like to thak a associate editor ad the referees for poitig out the paper [7] ad for helpful commets ad correctios. May thaks to Yiyu Ye for

13 Atoie Deza, Eissa Nematollahi ad Tamás Terlaky 13 precious suggestios ad hits which triggered this work ad for iformig us about the papers [11, 12]. Research supported by a NSERC Discovery grat ad a MITACS grat for the last two authors, by the Caada Research Chair program for the first ad last authors ad by a NSERC Discovery grat for the first author. Refereces [1] G. B. Datzig: Maximizatio of a liear fuctio of variables subject to liear iequalities. I: T. C. Koopmas (ed.) Activity Aalysis of Productio ad Allocatio. Joh Wiley (1951) [2] A. Deza, E. Nematollahi, R. Peyghami ad T. Terlaky: The cetral path visits all the vertices of the Klee-Mity cube. Optimizatio Methods ad Software (to appear). [3] N. K. Karmarkar: A ew polyomial-time algorithm for liear programmig. Combiatorica 4 (1984) [4] L. G. Khachiya: A polyomial algorithm i liear programmig. Soviet Mathematics Doklady 20 (1979) [5] V. Klee ad G. J. Mity: How good is the simplex algorithm? I: O. Shisha (ed.) Iequalities III, Academic Press (1972) [6] N. Megiddo: Pathways to the optimal set i liear programmig. I: N. Megiddo (ed.) Progress i Mathematical Programmig: Iterior-Poit ad Related Methods, Spriger- Verlag (1988) ; also i: Proceedigs of the 7th Mathematical Programmig Symposium of Japa, Nagoya, Japa (1986) [7] N. Megiddo ad M. Shub: Boudary behavior of iterior poit algorithms i liear programmig. Mathematics of Operatios Research 14-1 (1989) [8] C. Roos, T. Terlaky ad J-Ph. Vial: Theory ad Algorithms for Liear Optimizatio: A Iterior Poit Approach. Wiley-Itersciece Series i Discrete Mathematics ad Optimizatio. Joh Wiley (1997). [9] G. Soeved: A aalytical cetre for polyhedros ad ew classes of global algorithms for liear (smooth, covex) programmig. I: A. Prékopa, J. Szelezsá, ad B. Strazicky (eds.) System Modellig ad Optimizatio: Proceedigs of the 12th IFIP-Coferece, Budapest Lecture Notes i Cotrol ad Iformatio Scieces 84 Spriger Verlag (1986) [10] T. Terlaky ad S. Zhag: Pivot rules for liear programmig - a survey. Aals of Operatios Research 46 (1993) [11] M. Todd ad Y. Ye: A lower boud o the umber of iteratios of log-step ad polyomial iterior-poit liear programmig algorithms. Aals of Operatios Research 62 (1996) [12] S. Vavasis ad Y. Ye: A primal-dual iterior-poit method whose ruig time depeds oly o the costrait matrix. Mathematical Programmig 74 (1996)

14 14 How good are iterior poit methods? [13] S. J. Wright: Primal-Dual Iterior-Poit Methods. SIAM Publicatios (1997). [14] Y. Ye: Iterior-Poit Algorithms: Theory ad Aalysis. Wiley-Itersciece Series i Discrete Mathematics ad Optimizatio. Joh Wiley (1997). Atoie Deza, Eissa Nematollahi, Tamás Terlaky Advaced Optimizatio Laboratory, Departmet of Computig ad Software, McMaster Uiversity, Hamilto, Otario, Caada. deza, ematoe, terlaky@mcmaster.ca

How good are interior point methods? Klee Minty cubes tighten iteration-complexity bounds

How good are interior point methods? Klee Minty cubes tighten iteration-complexity bounds Math Program, Ser A (2008 113:1 14 DOI 101007/s10107-006-0044-x FULL LENGTH PAPER How good are iterior poit methods? Klee Mity cubes tighte iteratio-complexity bouds Atoie Deza Eissa Nematollahi Tamás

More information

A Simpler and Tighter Redundant Klee-Minty Construction

A Simpler and Tighter Redundant Klee-Minty Construction A Simpler and Tighter Redundant Klee-Minty Construction Eissa Nematollahi Tamás Terlaky October 19, 2006 Abstract By introducing redundant Klee-Minty examples, we have previously shown that the central

More information

Definitions and Theorems. where x are the decision variables. c, b, and a are constant coefficients.

Definitions and Theorems. where x are the decision variables. c, b, and a are constant coefficients. Defiitios ad Theorems Remember the scalar form of the liear programmig problem, Miimize, Subject to, f(x) = c i x i a 1i x i = b 1 a mi x i = b m x i 0 i = 1,2,, where x are the decisio variables. c, b,

More information

A Redundant Klee-Minty Construction with All the Redundant Constraints Touching the Feasible Region

A Redundant Klee-Minty Construction with All the Redundant Constraints Touching the Feasible Region A Redundant Klee-Minty Construction with All the Redundant Constraints Touching the Feasible Region Eissa Nematollahi Tamás Terlaky January 5, 2008 Abstract By introducing some redundant Klee-Minty constructions,

More information

Optimization Methods MIT 2.098/6.255/ Final exam

Optimization Methods MIT 2.098/6.255/ Final exam Optimizatio Methods MIT 2.098/6.255/15.093 Fial exam Date Give: December 19th, 2006 P1. [30 pts] Classify the followig statemets as true or false. All aswers must be well-justified, either through a short

More information

w (1) ˆx w (1) x (1) /ρ and w (2) ˆx w (2) x (2) /ρ.

w (1) ˆx w (1) x (1) /ρ and w (2) ˆx w (2) x (2) /ρ. 2 5. Weighted umber of late jobs 5.1. Release dates ad due dates: maximimizig the weight of o-time jobs Oce we add release dates, miimizig the umber of late jobs becomes a sigificatly harder problem. For

More information

15.081J/6.251J Introduction to Mathematical Programming. Lecture 21: Primal Barrier Interior Point Algorithm

15.081J/6.251J Introduction to Mathematical Programming. Lecture 21: Primal Barrier Interior Point Algorithm 508J/65J Itroductio to Mathematical Programmig Lecture : Primal Barrier Iterior Poit Algorithm Outlie Barrier Methods Slide The Cetral Path 3 Approximatig the Cetral Path 4 The Primal Barrier Algorithm

More information

Linear Programming! References! Introduction to Algorithms.! Dasgupta, Papadimitriou, Vazirani. Algorithms.! Cormen, Leiserson, Rivest, and Stein.

Linear Programming! References! Introduction to Algorithms.! Dasgupta, Papadimitriou, Vazirani. Algorithms.! Cormen, Leiserson, Rivest, and Stein. Liear Programmig! Refereces! Dasgupta, Papadimitriou, Vazirai. Algorithms.! Corme, Leiserso, Rivest, ad Stei. Itroductio to Algorithms.! Slack form! For each costrait i, defie a oegative slack variable

More information

Simple Polygons of Maximum Perimeter Contained in a Unit Disk

Simple Polygons of Maximum Perimeter Contained in a Unit Disk Discrete Comput Geom (009) 1: 08 15 DOI 10.1007/s005-008-9093-7 Simple Polygos of Maximum Perimeter Cotaied i a Uit Disk Charles Audet Pierre Hase Frédéric Messie Received: 18 September 007 / Revised:

More information

1 Duality revisited. AM 221: Advanced Optimization Spring 2016

1 Duality revisited. AM 221: Advanced Optimization Spring 2016 AM 22: Advaced Optimizatio Sprig 206 Prof. Yaro Siger Sectio 7 Wedesday, Mar. 9th Duality revisited I this sectio, we will give a slightly differet perspective o duality. optimizatio program: f(x) x R

More information

arxiv: v1 [math.mg] 29 Nov 2018

arxiv: v1 [math.mg] 29 Nov 2018 AN EXTREMAL PROBLEM OF REGULAR SIMPLICES THE HIGHER-DIMENSIONAL CASE ÁKOS GHORVÁTH arxiv:99v [mathmg] 9 Nov Abstract The ew result of this paper coected with the followig problem: Cosider a supportig hyperplae

More information

McMaster University. Advanced Optimization Laboratory. Title: The Central Path Visits all the Vertices of the Klee-Minty Cube.

McMaster University. Advanced Optimization Laboratory. Title: The Central Path Visits all the Vertices of the Klee-Minty Cube. McMaster University Avance Optimization Laboratory Title: The Central Path Visits all the Vertices of the Klee-Minty Cube Authors: Antoine Deza, Eissa Nematollahi, Reza Peyghami an Tamás Terlaky AvOl-Report

More information

Lecture 20. Brief Review of Gram-Schmidt and Gauss s Algorithm

Lecture 20. Brief Review of Gram-Schmidt and Gauss s Algorithm 8.409 A Algorithmist s Toolkit Nov. 9, 2009 Lecturer: Joatha Keler Lecture 20 Brief Review of Gram-Schmidt ad Gauss s Algorithm Our mai task of this lecture is to show a polyomial time algorithm which

More information

A New Solution Method for the Finite-Horizon Discrete-Time EOQ Problem

A New Solution Method for the Finite-Horizon Discrete-Time EOQ Problem This is the Pre-Published Versio. A New Solutio Method for the Fiite-Horizo Discrete-Time EOQ Problem Chug-Lu Li Departmet of Logistics The Hog Kog Polytechic Uiversity Hug Hom, Kowloo, Hog Kog Phoe: +852-2766-7410

More information

Computation of Error Bounds for P-matrix Linear Complementarity Problems

Computation of Error Bounds for P-matrix Linear Complementarity Problems Mathematical Programmig mauscript No. (will be iserted by the editor) Xiaoju Che Shuhuag Xiag Computatio of Error Bouds for P-matrix Liear Complemetarity Problems Received: date / Accepted: date Abstract

More information

Solutions for the Exam 9 January 2012

Solutions for the Exam 9 January 2012 Mastermath ad LNMB Course: Discrete Optimizatio Solutios for the Exam 9 Jauary 2012 Utrecht Uiversity, Educatorium, 15:15 18:15 The examiatio lasts 3 hours. Gradig will be doe before Jauary 23, 2012. Studets

More information

15.083J/6.859J Integer Optimization. Lecture 3: Methods to enhance formulations

15.083J/6.859J Integer Optimization. Lecture 3: Methods to enhance formulations 15.083J/6.859J Iteger Optimizatio Lecture 3: Methods to ehace formulatios 1 Outlie Polyhedral review Slide 1 Methods to geerate valid iequalities Methods to geerate facet defiig iequalities Polyhedral

More information

Math 451: Euclidean and Non-Euclidean Geometry MWF 3pm, Gasson 204 Homework 3 Solutions

Math 451: Euclidean and Non-Euclidean Geometry MWF 3pm, Gasson 204 Homework 3 Solutions Math 451: Euclidea ad No-Euclidea Geometry MWF 3pm, Gasso 204 Homework 3 Solutios Exercises from 1.4 ad 1.5 of the otes: 4.3, 4.10, 4.12, 4.14, 4.15, 5.3, 5.4, 5.5 Exercise 4.3. Explai why Hp, q) = {x

More information

Support vector machine revisited

Support vector machine revisited 6.867 Machie learig, lecture 8 (Jaakkola) 1 Lecture topics: Support vector machie ad kerels Kerel optimizatio, selectio Support vector machie revisited Our task here is to first tur the support vector

More information

2 The LCP problem and preliminary discussion

2 The LCP problem and preliminary discussion Iteratioal Mathematical Forum, Vol. 6, 011, o. 6, 191-196 Polyomial Covergece of a Predictor-Corrector Iterior-Poit Algorithm for LCP Feixiag Che College of Mathematics ad Computer Sciece, Chogqig Three

More information

Lecture Notes for Analysis Class

Lecture Notes for Analysis Class Lecture Notes for Aalysis Class Topological Spaces A topology for a set X is a collectio T of subsets of X such that: (a) X ad the empty set are i T (b) Uios of elemets of T are i T (c) Fiite itersectios

More information

Large holes in quasi-random graphs

Large holes in quasi-random graphs Large holes i quasi-radom graphs Joaa Polcy Departmet of Discrete Mathematics Adam Mickiewicz Uiversity Pozań, Polad joaska@amuedupl Submitted: Nov 23, 2006; Accepted: Apr 10, 2008; Published: Apr 18,

More information

The Method of Least Squares. To understand least squares fitting of data.

The Method of Least Squares. To understand least squares fitting of data. The Method of Least Squares KEY WORDS Curve fittig, least square GOAL To uderstad least squares fittig of data To uderstad the least squares solutio of icosistet systems of liear equatios 1 Motivatio Curve

More information

REGULARIZATION OF CERTAIN DIVERGENT SERIES OF POLYNOMIALS

REGULARIZATION OF CERTAIN DIVERGENT SERIES OF POLYNOMIALS REGULARIZATION OF CERTAIN DIVERGENT SERIES OF POLYNOMIALS LIVIU I. NICOLAESCU ABSTRACT. We ivestigate the geeralized covergece ad sums of series of the form P at P (x, where P R[x], a R,, ad T : R[x] R[x]

More information

Efficient GMM LECTURE 12 GMM II

Efficient GMM LECTURE 12 GMM II DECEMBER 1 010 LECTURE 1 II Efficiet The estimator depeds o the choice of the weight matrix A. The efficiet estimator is the oe that has the smallest asymptotic variace amog all estimators defied by differet

More information

Chapter 2 The Solution of Numerical Algebraic and Transcendental Equations

Chapter 2 The Solution of Numerical Algebraic and Transcendental Equations Chapter The Solutio of Numerical Algebraic ad Trascedetal Equatios Itroductio I this chapter we shall discuss some umerical methods for solvig algebraic ad trascedetal equatios. The equatio f( is said

More information

A constructive analysis of convex-valued demand correspondence for weakly uniformly rotund and monotonic preference

A constructive analysis of convex-valued demand correspondence for weakly uniformly rotund and monotonic preference MPRA Muich Persoal RePEc Archive A costructive aalysis of covex-valued demad correspodece for weakly uiformly rotud ad mootoic preferece Yasuhito Taaka ad Atsuhiro Satoh. May 04 Olie at http://mpra.ub.ui-mueche.de/55889/

More information

Problem Cosider the curve give parametrically as x = si t ad y = + cos t for» t» ß: (a) Describe the path this traverses: Where does it start (whe t =

Problem Cosider the curve give parametrically as x = si t ad y = + cos t for» t» ß: (a) Describe the path this traverses: Where does it start (whe t = Mathematics Summer Wilso Fial Exam August 8, ANSWERS Problem 1 (a) Fid the solutio to y +x y = e x x that satisfies y() = 5 : This is already i the form we used for a first order liear differetial equatio,

More information

CS537. Numerical Analysis and Computing

CS537. Numerical Analysis and Computing CS57 Numerical Aalysis ad Computig Lecture Locatig Roots o Equatios Proessor Ju Zhag Departmet o Computer Sciece Uiversity o Ketucky Leigto KY 456-6 Jauary 9 9 What is the Root May physical system ca be

More information

Several properties of new ellipsoids

Several properties of new ellipsoids Appl. Math. Mech. -Egl. Ed. 008 9(7):967 973 DOI 10.1007/s10483-008-0716-y c Shaghai Uiversity ad Spriger-Verlag 008 Applied Mathematics ad Mechaics (Eglish Editio) Several properties of ew ellipsoids

More information

Minimal surface area position of a convex body is not always an M-position

Minimal surface area position of a convex body is not always an M-position Miimal surface area positio of a covex body is ot always a M-positio Christos Saroglou Abstract Milma proved that there exists a absolute costat C > 0 such that, for every covex body i R there exists a

More information

Algorithm Analysis. Chapter 3

Algorithm Analysis. Chapter 3 Data Structures Dr Ahmed Rafat Abas Computer Sciece Dept, Faculty of Computer ad Iformatio, Zagazig Uiversity arabas@zu.edu.eg http://www.arsaliem.faculty.zu.edu.eg/ Algorithm Aalysis Chapter 3 3. Itroductio

More information

Mathematical Foundations -1- Sets and Sequences. Sets and Sequences

Mathematical Foundations -1- Sets and Sequences. Sets and Sequences Mathematical Foudatios -1- Sets ad Sequeces Sets ad Sequeces Methods of proof 2 Sets ad vectors 13 Plaes ad hyperplaes 18 Liearly idepedet vectors, vector spaces 2 Covex combiatios of vectors 21 eighborhoods,

More information

A 2nTH ORDER LINEAR DIFFERENCE EQUATION

A 2nTH ORDER LINEAR DIFFERENCE EQUATION A 2TH ORDER LINEAR DIFFERENCE EQUATION Doug Aderso Departmet of Mathematics ad Computer Sciece, Cocordia College Moorhead, MN 56562, USA ABSTRACT: We give a formulatio of geeralized zeros ad (, )-discojugacy

More information

A Hadamard-type lower bound for symmetric diagonally dominant positive matrices

A Hadamard-type lower bound for symmetric diagonally dominant positive matrices A Hadamard-type lower boud for symmetric diagoally domiat positive matrices Christopher J. Hillar, Adre Wibisoo Uiversity of Califoria, Berkeley Jauary 7, 205 Abstract We prove a ew lower-boud form of

More information

ACO Comprehensive Exam 9 October 2007 Student code A. 1. Graph Theory

ACO Comprehensive Exam 9 October 2007 Student code A. 1. Graph Theory 1. Graph Theory Prove that there exist o simple plaar triagulatio T ad two distict adjacet vertices x, y V (T ) such that x ad y are the oly vertices of T of odd degree. Do ot use the Four-Color Theorem.

More information

On Algorithm for the Minimum Spanning Trees Problem with Diameter Bounded Below

On Algorithm for the Minimum Spanning Trees Problem with Diameter Bounded Below O Algorithm for the Miimum Spaig Trees Problem with Diameter Bouded Below Edward Kh. Gimadi 1,2, Alexey M. Istomi 1, ad Ekateria Yu. Shi 2 1 Sobolev Istitute of Mathematics, 4 Acad. Koptyug aveue, 630090

More information

Linear Programming and the Simplex Method

Linear Programming and the Simplex Method Liear Programmig ad the Simplex ethod Abstract This article is a itroductio to Liear Programmig ad usig Simplex method for solvig LP problems i primal form. What is Liear Programmig? Liear Programmig is

More information

A class of spectral bounds for Max k-cut

A class of spectral bounds for Max k-cut A class of spectral bouds for Max k-cut Miguel F. Ajos, José Neto December 07 Abstract Let G be a udirected ad edge-weighted simple graph. I this paper we itroduce a class of bouds for the maximum k-cut

More information

TEACHER CERTIFICATION STUDY GUIDE

TEACHER CERTIFICATION STUDY GUIDE COMPETENCY 1. ALGEBRA SKILL 1.1 1.1a. ALGEBRAIC STRUCTURES Kow why the real ad complex umbers are each a field, ad that particular rigs are ot fields (e.g., itegers, polyomial rigs, matrix rigs) Algebra

More information

6.867 Machine learning, lecture 7 (Jaakkola) 1

6.867 Machine learning, lecture 7 (Jaakkola) 1 6.867 Machie learig, lecture 7 (Jaakkola) 1 Lecture topics: Kerel form of liear regressio Kerels, examples, costructio, properties Liear regressio ad kerels Cosider a slightly simpler model where we omit

More information

Supplemental Material: Proofs

Supplemental Material: Proofs Proof to Theorem Supplemetal Material: Proofs Proof. Let be the miimal umber of traiig items to esure a uique solutio θ. First cosider the case. It happes if ad oly if θ ad Rak(A) d, which is a special

More information

Disjoint Systems. Abstract

Disjoint Systems. Abstract Disjoit Systems Noga Alo ad Bey Sudaov Departmet of Mathematics Raymod ad Beverly Sacler Faculty of Exact Scieces Tel Aviv Uiversity, Tel Aviv, Israel Abstract A disjoit system of type (,,, ) is a collectio

More information

A Note on the Symmetric Powers of the Standard Representation of S n

A Note on the Symmetric Powers of the Standard Representation of S n A Note o the Symmetric Powers of the Stadard Represetatio of S David Savitt 1 Departmet of Mathematics, Harvard Uiversity Cambridge, MA 0138, USA dsavitt@mathharvardedu Richard P Staley Departmet of Mathematics,

More information

The random version of Dvoretzky s theorem in l n

The random version of Dvoretzky s theorem in l n The radom versio of Dvoretzky s theorem i l Gideo Schechtma Abstract We show that with high probability a sectio of the l ball of dimesio k cε log c > 0 a uiversal costat) is ε close to a multiple of the

More information

Integer Linear Programming

Integer Linear Programming Iteger Liear Programmig Itroductio Iteger L P problem (P) Mi = s. t. a = b i =,, m = i i 0, iteger =,, c Eemple Mi z = 5 s. t. + 0 0, 0, iteger F(P) = feasible domai of P Itroductio Iteger L P problem

More information

On Random Line Segments in the Unit Square

On Random Line Segments in the Unit Square O Radom Lie Segmets i the Uit Square Thomas A. Courtade Departmet of Electrical Egieerig Uiversity of Califoria Los Ageles, Califoria 90095 Email: tacourta@ee.ucla.edu I. INTRODUCTION Let Q = [0, 1] [0,

More information

Properties of Fuzzy Length on Fuzzy Set

Properties of Fuzzy Length on Fuzzy Set Ope Access Library Joural 206, Volume 3, e3068 ISSN Olie: 2333-972 ISSN Prit: 2333-9705 Properties of Fuzzy Legth o Fuzzy Set Jehad R Kider, Jaafar Imra Mousa Departmet of Mathematics ad Computer Applicatios,

More information

Homework Set #3 - Solutions

Homework Set #3 - Solutions EE 15 - Applicatios of Covex Optimizatio i Sigal Processig ad Commuicatios Dr. Adre Tkaceko JPL Third Term 11-1 Homework Set #3 - Solutios 1. a) Note that x is closer to x tha to x l i the Euclidea orm

More information

The Pivot Adaptive Method for Solving Linear Programming Problems

The Pivot Adaptive Method for Solving Linear Programming Problems America Joural of Operatios Research 28 8 92- http://wwwscirporg/oural/aor ISS Olie: 26-8849 ISS Prit: 26-883 he Pivot Adaptive Method for Solvig Liear Programmig Problems Saliha elahcee Philippe Martho

More information

Period Function of a Lienard Equation

Period Function of a Lienard Equation Joural of Mathematical Scieces (4) -5 Betty Joes & Sisters Publishig Period Fuctio of a Lieard Equatio Khalil T Al-Dosary Departmet of Mathematics, Uiversity of Sharjah, Sharjah 77, Uited Arab Emirates

More information

IP Reference guide for integer programming formulations.

IP Reference guide for integer programming formulations. IP Referece guide for iteger programmig formulatios. by James B. Orli for 15.053 ad 15.058 This documet is iteded as a compact (or relatively compact) guide to the formulatio of iteger programs. For more

More information

Fundamental Theorem of Algebra. Yvonne Lai March 2010

Fundamental Theorem of Algebra. Yvonne Lai March 2010 Fudametal Theorem of Algebra Yvoe Lai March 010 We prove the Fudametal Theorem of Algebra: Fudametal Theorem of Algebra. Let f be a o-costat polyomial with real coefficiets. The f has at least oe complex

More information

THE LONELY RUNNER PROBLEM FOR MANY RUNNERS. Artūras Dubickas Vilnius University, Lithuania

THE LONELY RUNNER PROBLEM FOR MANY RUNNERS. Artūras Dubickas Vilnius University, Lithuania GLASNIK MATEMATIČKI Vol. 46(66)(2011), 25 30 THE LONELY RUNNER PROBLEM FOR MANY RUNNERS Artūras Dubickas Vilius Uiversity, Lithuaia Abstract. The loely ruer cojecture asserts that for ay positive iteger

More information

Analytic Continuation

Analytic Continuation Aalytic Cotiuatio The stadard example of this is give by Example Let h (z) = 1 + z + z 2 + z 3 +... kow to coverge oly for z < 1. I fact h (z) = 1/ (1 z) for such z. Yet H (z) = 1/ (1 z) is defied for

More information

Recursive Algorithms. Recurrences. Recursive Algorithms Analysis

Recursive Algorithms. Recurrences. Recursive Algorithms Analysis Recursive Algorithms Recurreces Computer Sciece & Egieerig 35: Discrete Mathematics Christopher M Bourke cbourke@cseuledu A recursive algorithm is oe i which objects are defied i terms of other objects

More information

Feedback in Iterative Algorithms

Feedback in Iterative Algorithms Feedback i Iterative Algorithms Charles Byre (Charles Byre@uml.edu), Departmet of Mathematical Scieces, Uiversity of Massachusetts Lowell, Lowell, MA 01854 October 17, 2005 Abstract Whe the oegative system

More information

CS321. Numerical Analysis and Computing

CS321. Numerical Analysis and Computing CS Numerical Aalysis ad Computig Lecture Locatig Roots o Equatios Proessor Ju Zhag Departmet o Computer Sciece Uiversity o Ketucky Leigto KY 456-6 September 8 5 What is the Root May physical system ca

More information

The multiplicative structure of finite field and a construction of LRC

The multiplicative structure of finite field and a construction of LRC IERG6120 Codig for Distributed Storage Systems Lecture 8-06/10/2016 The multiplicative structure of fiite field ad a costructio of LRC Lecturer: Keeth Shum Scribe: Zhouyi Hu Notatios: We use the otatio

More information

The Boolean Ring of Intervals

The Boolean Ring of Intervals MATH 532 Lebesgue Measure Dr. Neal, WKU We ow shall apply the results obtaied about outer measure to the legth measure o the real lie. Throughout, our space X will be the set of real umbers R. Whe ecessary,

More information

APPLICATION OF YOUNG S INEQUALITY TO VOLUMES OF CONVEX SETS

APPLICATION OF YOUNG S INEQUALITY TO VOLUMES OF CONVEX SETS APPLICATION OF YOUNG S INEQUALITY TO VOLUMES OF CONVEX SETS 1. Itroductio Let C be a bouded, covex subset of. Thus, by defiitio, with every two poits i the set, the lie segmet coectig these two poits is

More information

Linear regression. Daniel Hsu (COMS 4771) (y i x T i β)2 2πσ. 2 2σ 2. 1 n. (x T i β y i ) 2. 1 ˆβ arg min. β R n d

Linear regression. Daniel Hsu (COMS 4771) (y i x T i β)2 2πσ. 2 2σ 2. 1 n. (x T i β y i ) 2. 1 ˆβ arg min. β R n d Liear regressio Daiel Hsu (COMS 477) Maximum likelihood estimatio Oe of the simplest liear regressio models is the followig: (X, Y ),..., (X, Y ), (X, Y ) are iid radom pairs takig values i R d R, ad Y

More information

Technical Proofs for Homogeneity Pursuit

Technical Proofs for Homogeneity Pursuit Techical Proofs for Homogeeity Pursuit bstract This is the supplemetal material for the article Homogeeity Pursuit, submitted for publicatio i Joural of the merica Statistical ssociatio. B Proofs B. Proof

More information

Machine Learning Theory Tübingen University, WS 2016/2017 Lecture 11

Machine Learning Theory Tübingen University, WS 2016/2017 Lecture 11 Machie Learig Theory Tübige Uiversity, WS 06/07 Lecture Tolstikhi Ilya Abstract We will itroduce the otio of reproducig kerels ad associated Reproducig Kerel Hilbert Spaces (RKHS). We will cosider couple

More information

Rotationally invariant integrals of arbitrary dimensions

Rotationally invariant integrals of arbitrary dimensions September 1, 14 Rotatioally ivariat itegrals of arbitrary dimesios James D. Wells Physics Departmet, Uiversity of Michiga, A Arbor Abstract: I this ote itegrals over spherical volumes with rotatioally

More information

A Note On The Exponential Of A Matrix Whose Elements Are All 1

A Note On The Exponential Of A Matrix Whose Elements Are All 1 Applied Mathematics E-Notes, 8(208), 92-99 c ISSN 607-250 Available free at mirror sites of http://wwwmaththuedutw/ ame/ A Note O The Expoetial Of A Matrix Whose Elemets Are All Reza Farhadia Received

More information

A HYPERPLANE INEQUALITY FOR MEASURES OF CONVEX BODIES IN R n, n 4

A HYPERPLANE INEQUALITY FOR MEASURES OF CONVEX BODIES IN R n, n 4 A HYPERPANE INEQUAITY FOR MEASURES OF CONVEX BODIES IN R, 4 AEXANDER ODOBSY Abstract. et 4. We show that for a arbitrary measure µ with eve cotiuous desity i R ad ay origi-symmetric covex body i R, µ()

More information

NEW FAST CONVERGENT SEQUENCES OF EULER-MASCHERONI TYPE

NEW FAST CONVERGENT SEQUENCES OF EULER-MASCHERONI TYPE UPB Sci Bull, Series A, Vol 79, Iss, 207 ISSN 22-7027 NEW FAST CONVERGENT SEQUENCES OF EULER-MASCHERONI TYPE Gabriel Bercu We itroduce two ew sequeces of Euler-Mascheroi type which have fast covergece

More information

A Quantitative Lusin Theorem for Functions in BV

A Quantitative Lusin Theorem for Functions in BV A Quatitative Lusi Theorem for Fuctios i BV Adrás Telcs, Vicezo Vespri November 19, 013 Abstract We exted to the BV case a measure theoretic lemma previously proved by DiBeedetto, Giaazza ad Vespri ([1])

More information

( ) (( ) ) ANSWERS TO EXERCISES IN APPENDIX B. Section B.1 VECTORS AND SETS. Exercise B.1-1: Convex sets. are convex, , hence. and. (a) Let.

( ) (( ) ) ANSWERS TO EXERCISES IN APPENDIX B. Section B.1 VECTORS AND SETS. Exercise B.1-1: Convex sets. are convex, , hence. and. (a) Let. Joh Riley 8 Jue 03 ANSWERS TO EXERCISES IN APPENDIX B Sectio B VECTORS AND SETS Exercise B-: Covex sets (a) Let 0 x, x X, X, hece 0 x, x X ad 0 x, x X Sice X ad X are covex, x X ad x X The x X X, which

More information

Sieve Estimators: Consistency and Rates of Convergence

Sieve Estimators: Consistency and Rates of Convergence EECS 598: Statistical Learig Theory, Witer 2014 Topic 6 Sieve Estimators: Cosistecy ad Rates of Covergece Lecturer: Clayto Scott Scribe: Julia Katz-Samuels, Brado Oselio, Pi-Yu Che Disclaimer: These otes

More information

Scheduling under Uncertainty using MILP Sensitivity Analysis

Scheduling under Uncertainty using MILP Sensitivity Analysis Schedulig uder Ucertaity usig MILP Sesitivity Aalysis M. Ierapetritou ad Zheya Jia Departmet of Chemical & Biochemical Egieerig Rutgers, the State Uiversity of New Jersey Piscataway, NJ Abstract The aim

More information

Machine Learning Theory Tübingen University, WS 2016/2017 Lecture 12

Machine Learning Theory Tübingen University, WS 2016/2017 Lecture 12 Machie Learig Theory Tübige Uiversity, WS 06/07 Lecture Tolstikhi Ilya Abstract I this lecture we derive risk bouds for kerel methods. We will start by showig that Soft Margi kerel SVM correspods to miimizig

More information

Markov Decision Processes

Markov Decision Processes Markov Decisio Processes Defiitios; Statioary policies; Value improvemet algorithm, Policy improvemet algorithm, ad liear programmig for discouted cost ad average cost criteria. Markov Decisio Processes

More information

Entropy and Ergodic Theory Lecture 5: Joint typicality and conditional AEP

Entropy and Ergodic Theory Lecture 5: Joint typicality and conditional AEP Etropy ad Ergodic Theory Lecture 5: Joit typicality ad coditioal AEP 1 Notatio: from RVs back to distributios Let (Ω, F, P) be a probability space, ad let X ad Y be A- ad B-valued discrete RVs, respectively.

More information

Introduction to Optimization Techniques. How to Solve Equations

Introduction to Optimization Techniques. How to Solve Equations Itroductio to Optimizatio Techiques How to Solve Equatios Iterative Methods of Optimizatio Iterative methods of optimizatio Solutio of the oliear equatios resultig form a optimizatio problem is usually

More information

1 Introduction. 1.1 Notation and Terminology

1 Introduction. 1.1 Notation and Terminology 1 Itroductio You have already leared some cocepts of calculus such as limit of a sequece, limit, cotiuity, derivative, ad itegral of a fuctio etc. Real Aalysis studies them more rigorously usig a laguage

More information

Supplementary Material for Fast Stochastic AUC Maximization with O(1/n)-Convergence Rate

Supplementary Material for Fast Stochastic AUC Maximization with O(1/n)-Convergence Rate Supplemetary Material for Fast Stochastic AUC Maximizatio with O/-Covergece Rate Migrui Liu Xiaoxua Zhag Zaiyi Che Xiaoyu Wag 3 iabao Yag echical Lemmas ized versio of Hoeffdig s iequality, ote that We

More information

Concavity Solutions of Second-Order Differential Equations

Concavity Solutions of Second-Order Differential Equations Proceedigs of the Paista Academy of Scieces 5 (3): 4 45 (4) Copyright Paista Academy of Scieces ISSN: 377-969 (prit), 36-448 (olie) Paista Academy of Scieces Research Article Cocavity Solutios of Secod-Order

More information

Existence of viscosity solutions with asymptotic behavior of exterior problems for Hessian equations

Existence of viscosity solutions with asymptotic behavior of exterior problems for Hessian equations Available olie at www.tjsa.com J. Noliear Sci. Appl. 9 (2016, 342 349 Research Article Existece of viscosity solutios with asymptotic behavior of exterior problems for Hessia equatios Xiayu Meg, Yogqiag

More information

Self-normalized deviation inequalities with application to t-statistic

Self-normalized deviation inequalities with application to t-statistic Self-ormalized deviatio iequalities with applicatio to t-statistic Xiequa Fa Ceter for Applied Mathematics, Tiaji Uiversity, 30007 Tiaji, Chia Abstract Let ξ i i 1 be a sequece of idepedet ad symmetric

More information

Convergence of random variables. (telegram style notes) P.J.C. Spreij

Convergence of random variables. (telegram style notes) P.J.C. Spreij Covergece of radom variables (telegram style otes).j.c. Spreij this versio: September 6, 2005 Itroductio As we kow, radom variables are by defiitio measurable fuctios o some uderlyig measurable space

More information

Improving the Localization of Eigenvalues for Complex Matrices

Improving the Localization of Eigenvalues for Complex Matrices Applied Mathematical Scieces, Vol. 5, 011, o. 8, 1857-1864 Improvig the Localizatio of Eigevalues for Complex Matrices P. Sargolzaei 1, R. Rakhshaipur Departmet of Mathematics, Uiversity of Sista ad Baluchesta

More information

Problem Set 2 Solutions

Problem Set 2 Solutions CS271 Radomess & Computatio, Sprig 2018 Problem Set 2 Solutios Poit totals are i the margi; the maximum total umber of poits was 52. 1. Probabilistic method for domiatig sets 6pts Pick a radom subset S

More information

Mathematical Methods for Physics and Engineering

Mathematical Methods for Physics and Engineering Mathematical Methods for Physics ad Egieerig Lecture otes Sergei V. Shabaov Departmet of Mathematics, Uiversity of Florida, Gaiesville, FL 326 USA CHAPTER The theory of covergece. Numerical sequeces..

More information

Introduction to Optimization Techniques

Introduction to Optimization Techniques Itroductio to Optimizatio Techiques Basic Cocepts of Aalysis - Real Aalysis, Fuctioal Aalysis 1 Basic Cocepts of Aalysis Liear Vector Spaces Defiitio: A vector space X is a set of elemets called vectors

More information

Continuous Trajectories for Primal-Dual Potential-Reduction Methods

Continuous Trajectories for Primal-Dual Potential-Reduction Methods Cotiuous Trajectories for Primal-Dual Potetial-Reductio Methods Reha H. Tütücü September 7, 2001 Abstract This article cosiders cotiuous trajectories of the vector fields iduced by two differet primal-dual

More information

Algebra of Least Squares

Algebra of Least Squares October 19, 2018 Algebra of Least Squares Geometry of Least Squares Recall that out data is like a table [Y X] where Y collects observatios o the depedet variable Y ad X collects observatios o the k-dimesioal

More information

On forward improvement iteration for stopping problems

On forward improvement iteration for stopping problems O forward improvemet iteratio for stoppig problems Mathematical Istitute, Uiversity of Kiel, Ludewig-Mey-Str. 4, D-24098 Kiel, Germay irle@math.ui-iel.de Albrecht Irle Abstract. We cosider the optimal

More information

Supplementary Material for Fast Stochastic AUC Maximization with O(1/n)-Convergence Rate

Supplementary Material for Fast Stochastic AUC Maximization with O(1/n)-Convergence Rate Supplemetary Material for Fast Stochastic AUC Maximizatio with O/-Covergece Rate Migrui Liu Xiaoxua Zhag Zaiyi Che Xiaoyu Wag 3 iabao Yag echical Lemmas ized versio of Hoeffdig s iequality, ote that We

More information

Optimally Sparse SVMs

Optimally Sparse SVMs A. Proof of Lemma 3. We here prove a lower boud o the umber of support vectors to achieve geeralizatio bouds of the form which we cosider. Importatly, this result holds ot oly for liear classifiers, but

More information

Expected Norms of Zero-One Polynomials

Expected Norms of Zero-One Polynomials DRAFT: Caad. Math. Bull. July 4, 08 :5 File: borwei80 pp. Page Sheet of Caad. Math. Bull. Vol. XX (Y, ZZZZ pp. 0 0 Expected Norms of Zero-Oe Polyomials Peter Borwei, Kwok-Kwog Stephe Choi, ad Idris Mercer

More information

lim za n n = z lim a n n.

lim za n n = z lim a n n. Lecture 6 Sequeces ad Series Defiitio 1 By a sequece i a set A, we mea a mappig f : N A. It is customary to deote a sequece f by {s } where, s := f(). A sequece {z } of (complex) umbers is said to be coverget

More information

CSI 2101 Discrete Structures Winter Homework Assignment #4 (100 points, weight 5%) Due: Thursday, April 5, at 1:00pm (in lecture)

CSI 2101 Discrete Structures Winter Homework Assignment #4 (100 points, weight 5%) Due: Thursday, April 5, at 1:00pm (in lecture) CSI 101 Discrete Structures Witer 01 Prof. Lucia Moura Uiversity of Ottawa Homework Assigmet #4 (100 poits, weight %) Due: Thursday, April, at 1:00pm (i lecture) Program verificatio, Recurrece Relatios

More information

Confidence interval for the two-parameter exponentiated Gumbel distribution based on record values

Confidence interval for the two-parameter exponentiated Gumbel distribution based on record values Iteratioal Joural of Applied Operatioal Research Vol. 4 No. 1 pp. 61-68 Witer 2014 Joural homepage: www.ijorlu.ir Cofidece iterval for the two-parameter expoetiated Gumbel distributio based o record values

More information

subject to A 1 x + A 2 y b x j 0, j = 1,,n 1 y j = 0 or 1, j = 1,,n 2

subject to A 1 x + A 2 y b x j 0, j = 1,,n 1 y j = 0 or 1, j = 1,,n 2 Additioal Brach ad Boud Algorithms 0-1 Mixed-Iteger Liear Programmig The brach ad boud algorithm described i the previous sectios ca be used to solve virtually all optimizatio problems cotaiig iteger variables,

More information

10-701/ Machine Learning Mid-term Exam Solution

10-701/ Machine Learning Mid-term Exam Solution 0-70/5-78 Machie Learig Mid-term Exam Solutio Your Name: Your Adrew ID: True or False (Give oe setece explaatio) (20%). (F) For a cotiuous radom variable x ad its probability distributio fuctio p(x), it

More information

BETWEEN QUASICONVEX AND CONVEX SET-VALUED MAPPINGS. 1. Introduction. Throughout the paper we denote by X a linear space and by Y a topological linear

BETWEEN QUASICONVEX AND CONVEX SET-VALUED MAPPINGS. 1. Introduction. Throughout the paper we denote by X a linear space and by Y a topological linear BETWEEN QUASICONVEX AND CONVEX SET-VALUED MAPPINGS Abstract. The aim of this paper is to give sufficiet coditios for a quasicovex setvalued mappig to be covex. I particular, we recover several kow characterizatios

More information

Linearly Independent Sets, Bases. Review. Remarks. A set of vectors,,, in a vector space is said to be linearly independent if the vector equation

Linearly Independent Sets, Bases. Review. Remarks. A set of vectors,,, in a vector space is said to be linearly independent if the vector equation Liearly Idepedet Sets Bases p p c c p Review { v v vp} A set of vectors i a vector space is said to be liearly idepedet if the vector equatio cv + c v + + c has oly the trivial solutio = = { v v vp} The

More information

Equivalence Between An Approximate Version Of Brouwer s Fixed Point Theorem And Sperner s Lemma: A Constructive Analysis

Equivalence Between An Approximate Version Of Brouwer s Fixed Point Theorem And Sperner s Lemma: A Constructive Analysis Applied Mathematics E-Notes, 11(2011), 238 243 c ISSN 1607-2510 Available free at mirror sites of http://www.math.thu.edu.tw/ame/ Equivalece Betwee A Approximate Versio Of Brouwer s Fixed Poit Theorem

More information