Near-Optimal Lower Bounds for ɛ-nets for Half-spaces and Low Complexity Set Systems

Size: px
Start display at page:

Download "Near-Optimal Lower Bounds for ɛ-nets for Half-spaces and Low Complexity Set Systems"

Transcription

1 Near-Optimal Lower Bouds for -ets for Half-spaces ad Low Compleity Set Systems Adrey Kupavskii Nabil H. Mustafa Jáos Pach Abstract Followig groudbreakig work by Haussler ad Welzl 1987, the use of small -ets has become a stadard techique for solvig algorithmic ad etremal problems i geometry ad learig theory. Two sigificat recet developmets are: i a upper boud o the size of the smallest -ets for set systems, as a fuctio of their so-called shallow-cell compleity Cha, Grat, Köema, ad Sharpe; ad ii the costructio of a set system whose members ca be obtaied by itersectig a poit set i R 4 by a family of half-spaces such that the size of ay -et for them is Ω 1 log 1 Pach ad Tardos. The preset paper completes both of these aveues of research. We i give a lower boud, matchig the result of Cha et al., ad ii geeralize the costructio of Pach ad Tardos to half-spaces i R d, for ay d 4, to show that the geeral upper boud, O d log 1, of Haussler ad Welzl for the size of the smallest -ets is tight. 1 Itroductio Let X be a fiite set ad let R be a system of subsets of a uderlyig set cotaiig X. I computatioal geometry, the pair X, R is usually called a rage space. A subset X X is called a -et for X, R if X R for every R R with R X X. The use of small-sized -ets i geometrically defied rage spaces has become a stadard techique i discrete ad computatioal geometry, with may combiatorial ad algorithmic cosequeces. I most applicatios, -ets precisely ad provably capture the most importat quatitative ad qualitative properties that oe would epect from a radom sample. Typical applicatios iclude the eistece of spaig trees ad simplicial partitios with low crossig umber, upper bouds for discrepacy of set systems, LP roudig, rage searchig, streamig algorithms; see [13, 18]. For ay subset Y X, defie the projectio of R o Y to be the set system R Y := { Y R : R R }. The Vapik-Chervoekis dimesio or, i short, the VC-dimesio of the rage space X, R is the miimum iteger d such that R Y < 2 R for ay subset Y X with Y > d. Accordig to the Sauer Shelah lemma [21, 23] discovered earlier by Vapik ad Chervoekis [24], for ay rage space X, R whose VC-dimesio is at most d ad for ay subset Y X, we have R Y d i=0 Y i = O Y d. A straightforward samplig argumet shows that every rage space X, R has a -et of size O 1 log R X. The remarkable result of Haussler ad Welzl [10], based o the previous work of Vapik ad Chervoekis [24], shows that much smaller -ets eist if we assume that our rage space has small VC-dimesio. Haussler ad Welzl [10] showed that if the VC-dimesio of a rage space X, R is at most d, the by pickig a radom sample of size Θ d log d, we obtai a -et with positive probability. Actually, they oly used the weaker assumptio that R Y = O Y d for every Y X. This boud was later improved to 1 + o1 d log 1, as 1 ad d is large [11]. I the sequel, we will refer to this result as the -et theorem. The key feature of the -et theorem is that it guaratees the eistece of a Moscow Istitute of Physics ad Techology, G-SCOP laboratory, Greoble. kupavskii@yade.ru. The work of Adrey Kupavskii has bee supported i part by the Swiss Natioal Sciece Foudatio Grats ad ad by the grat N of the Russia Foudatio for Basic Research. Uiversité Paris-Est, LIGM, Equipe A3SI, ESIEE Paris, Frace. mustafa@esiee.fr. The work of Nabil H. Mustafa has bee supported by the grat ANR SAGA JCJC-14-CE EPFL, Lausae ad Réyi Istitute, Budapest. pach@cims.yu.edu. The work of Jáos Pach has bee partially supported by Swiss Natioal Sciece Foudatio Grats ad

2 -et whose size is idepedet of both X ad R X. Furthermore, if oe oly requires the VC-dimesio of X, R to be bouded by d, the this boud caot be improved. It was show i [11] that give ay > 0 ad iteger d 2, there eist rage spaces with VC-dimesio at most d, ad for which ay -et must have size at least 1 2 d + 1 dd+2 + o1 d log 1. The effectiveess of -et theory i geometry derives from the fact that most geometrically defied rage spaces X, R arisig i applicatios have bouded VC-dimesio ad, hece, satisfy the precoditios of the -et theorem. There are two importat types of geometric set systems, both ivolvig poits ad geometric objects i R d, that are used i such applicatios. Let R be a family of possibly ubouded geometric objects i R d, such as the family of all half-spaces, all balls, all polytopes with a bouded umber of facets, or all semialgebraic sets of bouded compleity, i.e., subsets of R d defied by at most D polyomial equatios or iequalities i the d variables, each of degree at most D. Give a fiite set of poits X R d, we defie the primal rage space X, R as the set system iduced by cotaimet i the objects from R. Formally, it is a set system with the set of elemets X ad sets {X R : R R}. The combiatorial properties of this rage space deped o the projectio R X. Usig this termiology, Rado s theorem [13] implies that the primal rage space o a groud set X, iduced by cotaimet i half-spaces i R d, has VC-dimesio at most d + 1 [18]. Thus, by the -et theorem, this rage space has a -et of size O d log 1. I may applicatios, it is atural to cosider the dual rage space, i which the roles of the poits ad rages are swapped. As above, let R be a family of geometric objects rages i R d. Give a fiite set of objects S R, the dual rage space iduced by them is defied as the set system hypergraph o the groud set S, cosistig of the sets S := {S S : S} for all R d. It ca be show that if for ay X R d the VC-dimesio of the rage space X, R is less tha d, the the VC-dimesio of the dual rage space iduced by ay subset of R is less tha 2 d [13]. Recet progress. I may geometric scearios, however, oe ca fid smaller -ets tha those whose eistece is guarateed by the -et theorem. It has bee kow for a log time that this is the case, e.g., for primal set systems iduced by cotaimet i balls i R 2 ad half-spaces i R 2 ad R 3. Over the past two decades, a umber of specialized techiques have bee developed to show the eistece of small-sized -ets for such set systems [3, 4, 5, 6, 7, 8, 11, 12, 14, 16, 20, 25, 26]. Based o these successes, it was geerally believed that i most geometric scearios oe should be able to substatially stregthe the -et theorem, ad obtai perhaps eve a O 1 upper boud for the size of the smallest -ets. I this directio, there have bee two sigificat recet developmets: oe positive ad oe egative. Upper bouds. Followig the work of Clarkso ad Varadaraja [8], it has bee gradually realized that if oe replaces the coditio that the rage space X, R has bouded VC-dimesio by a more refied combiatorial property, oe ca prove the eistece of -ets of size o 1 log 1. To formulate this property, we eed to itroduce some termiology. Give a fuctio ϕ : N R +, we say that the rage space X, R has shallow-cell compleity ϕ if there eists a costat c = cr > 0 such that, for every Y X ad for every positive iteger l, the umber of at most l-elemet sets i R Y is O Y ϕ Y l c. Note that if the VC-dimesio of X, R is d, the for every Y X, the umber of elemets of the projectio of the set system R to Y satisfies R Y = O Y d. However, the coditio that X, R has shallow-cell compleity ϕ for some fuctio ϕ = O d, 0 < d < d 1 ad some costat c = cr, implies ot oly that R Y = O Y 1+d +c, but it reveals some otrivial fier details about the distributio of the sizes of the smaller members of R Y. Several of the rage spaces metioed earlier tured out to have low shallow-cell compleity. For istace, the primal rage spaces iduced by cotaimet of poits i disks i R 2 or half-spaces i R 3 have shallow-cell compleity ϕ = O1. I geeral, it is kow [13] that the primal rage space iduced by cotaimet of poits by half-spaces i R d has shallow-cell compleity ϕ = O d/2 1. Defie the uio compleity of a family of objects R, as the maimum umber of faces boudary pieces of all dimesios that the uio of ay members of R ca have; see [1]. Applyig a simple probabilistic techique developed by Clarkso ad Shor [9], oe ca fid a iterestig relatioship betwee the uio compleity of a family of objects R ad the shallow-cell compleities of the dual rage spaces iduced by subsets S R. Suppose that the uio compleity of a family R of objects i the plae is O ϕ, for some well-behaved o-decreasig fuctio ϕ. The the umber of at most l-elemet subsets i the dual rage space iduced by ay S R is O l 2 S S l ϕ l = O S ϕ S l [22]; i.e., the dual rage space iduced by S has shallow-cell compleity O ϕ. Accordig to the above 2

3 defiitios, this meas that for ay S R ad for ay positive iteger l, the umber of subsets S S l for which there is a poit p R 2 cotaied i all elemets of S, but i oe of the elemets of S \ S, is at most O S ϕ S l. For small values of l, the poits p are ot heavily covered. Thus, the correspodig cells S S S \ T S\S T of the arragemet S are shallow, ad the umber of these shallow cells is bouded from above. This eplais the use of the term shallow-cell compleity. A series of elegat results [20, 3, 6, 26] illustrate that if the shallow-cell compleity of a set system is ϕ = o, the it permits smaller -ets tha what is guarateed by the -et theorem. The followig theorem represets the curret state of the art; see [15] for a simple proof of this statemet. Theorem A. Let X, R be a rage space with shallow-cell compleity ϕ, where ϕ = O d for some costat d. The, for every > 0, it has a -et of size O 1 log ϕ 1, where the costat hidde i the O-otatio depeds o d. Proof. Sketch. The mai result i [6] shows the eistece of -ets of size O 1 log ϕ X for ay o-decreasig fuctio ϕ 1. To get a boud idepedet of X, first compute a small /2-approimatio A X for X, R [13]. It is kow that there is such a A with A = O d 2 log 1 = O 1 3, ad for ay R R, we have R A A R X 2. I particular, ay R R with R X cotais at least a 2 -fractio of the elemets of A. Therefore, a /2-et for A, R A is a -et for X, R. Computig a /2-et for A, R A gives the required set of size O 2 log ϕ A = O 1 log ϕ 1 3 = O 1 log ϕ 1. Note that i the bouds o the sizes of -ets based o VC-dimesio, we eplicitly state the depedece o d. O the other had, i the bouds based o shallow-cell compleity, we will assume that d is a costat. Lower bouds. It was cojectured for a log time [14] that most geometrically defied rage spaces of bouded Vapik-Chervoekis dimesio have liear-sized -ets, i.e., -ets of size O 1. These hopes were shattered by Alo [2], who established a superliear but barely superliear! lower boud o the size of -ets for the primal rage space iduced by straight lies i the plae. Shortly after, Pach ad Tardos [19] maaged to establish a tight lower boud of Ω 1 log 1 for the size of -ets i primal rage spaces iduced by half-spaces i R 4, ad i several other geometric scearios. Theorem B. [19] Let F deote the family of half-spaces i R 4. For ay > 0 ad ay sufficietly large iteger, there eists a set X R 4 of poits such that i the primal rage spaces X, F, the size of every -et is at least 1 9 log 1. Our cotributios. The aim of this paper is to complete both aveues of research opeed by Theorems A ad B. Our first theorem, proved i Sectio 2, geeralizes Theorem B to R d, for d 4. It provides a asymptotically tight boud i terms of both ε ad d, ad hece completely settles the -et problem for half-spaces. Theorem 1. For ay iteger d 4, real > 0 ad ay sufficietly large iteger 0, there eist primal rage spaces X, F iduced by -elemet poit sets X ad collectios of half-spaces F i R d such that the size of every -et for X, F is at least d/4 9 log 1. As was metioed i the first subsectio, for ay d 1, the VC-dimesio of ay rage space iduced by poits ad half-spaces i R d is at most d + 1. Thus, Theorem 1 matches, up to a costat factor idepedet of d ad, the upper boud implied by the -et theorem of Haussler ad Welzl. Noga Alo poited out to us that it is very easy to show that for a fied > 0, the lower boud for -ets i rage spaces iduced by half-spaces i R d has to grow at least liearly i d. To see this, suppose that we wat to obtai a 1 3-et, say, for the rage space iduced by ope half-spaces o a set X of 3d poits i geeral positio i R d. Notice that for this we eed at least d + 1 poits. Ideed, ay d poits of X spa a hyperplae, ad oe of the ope half-spaces determied by this hyperplae cotais at least X 3 poits. The key elemet of the proof of Theorem B [19] was to costruct a set B of k + 32 k 2 ais-parallel rectagles i the plae such that for ay subset of them there is a set Q of at most 2 k 1 poits that hit oe of the rectagles that belog to this subset ad all the rectagles i its complemet the precise statemet is give i Sectio 3. We geeralize this statemet to R d by costructig roughly d 2 times more ais-parallel boes 2 tha i the plaar case, but the size of the set Q remais the same size. I Sectio 3, we prove 1 Their result is i fact for the more geeral problem of small weight -ets. 2 A ais-parallel bo i R d is the Cartesia product of d itervals. For simplicity, i the sequel, they will be called boes. 3

4 Lemma 2. Let k, d 2 be itegers. The there eists a set B of d 2 k + 32k 2 ais-parallel boes i R d such that for ay subset S B, oe ca fid a 2 k 1 -elemet set Q of poits with the property that i Q B for ay B B \ S, ad ii Q B = for ay B S. I the et sectio we show how this lemma implies the boud of Theorem 1, which is d 4 times better tha the boud i Theorem B. The proof of Lemma 2 will be give i Sectio 3. I Sectio 4, we show that the boud i Theorem A caot be improved. Defiitio 1. A fuctio ϕ : R + R + is called submultiplicative if there eists a 0 R + such that for every α, 0 < α < 1, ad > 0, we have ϕ α 1/α ϕ. Some eamples of submultiplicative fuctios are c for ay positive c, 2 log, log, log log, log, ad the iverse Ackerma fuctio. Theorem 3. Let d be a fied positive iteger ad let ϕ : R + R +, ϕ, be a mootoically icreasig submultiplicative fuctio that teds to ifiity such that ϕ = O d. The, for ay > 0, there eist rage spaces X, F that have i shallow-cell compleity ϕ, ad for which ii the size of ay -et is Ω 1 log ϕ 1. Theorem 3 becomes iterestig whe ϕ = o ad the upper boud O 1 log ϕ 1 i Theorem A improves o the geeral upper boud O 1 log 1 guarateed by the -et theorem. Theorem 3 shows that, eve if ϕ = o, this improved boud is asymptotically tight. The best upper ad lower bouds for the size of small -ets i rage spaces with a give shallow-cell compleity ϕ are based o purely combiatorial argumets, ad they imply directly or idirectly all kow results o -ets i geometrically defied rage spaces see [17] for a detailed discussio. This suggests that the itroductio of the otio of shallow-cell compleity provided the right framework for -et theory. 2 Proof of Theorem 1 usig Lemma 2 Let B be a set of d-dimesioal ais-parallel boes i R d. We recall that the dual rage space iduced by B is the set system hypergraph o the groud set B cosistig of the sets B p := {B B : p B} for all p R d. Lemma 4. Let d 1 be a iteger, ad cosider the dual rage space iduced by a set of ais-parallel boes B i R d. The there eists a fuctio f : B R 2d such that for every poit p R d, there is a half-space H i R 2d with {fb : B B p } = H {fb : B B}. Proof. By traslatio, we ca assume that all boes i B lie i the positive orthat of R d. Cosider the fuctio g : B R 2d mappig a bo B = [ l 1, r 1] [ l 2, r 2] [ l d, r d ] to the poit l 1, 1/ r 1, l 2, 1/ r 2,..., l d, 1/r d lyig i the positive orthat of R2d. Furthermore, for ay p = a 1, a 2,..., a d R d i the positive orthat, let C p deote the bo [0, a 1 ] [0, 1/a 1 ] [0, a 2 ] [0, 1/a 2 ] [0, a d ] [0, 1/a d ] i R 2d. Clearly, a poit p lies i a bo B i R d if ad oly if gb C p i R 2d. Thus, g maps the set of boes i B to a set of poits i R 2d, such that for ay poit p i the positive orthat of R d, the set of boes B p B that cotai p are mapped to the set of poits that belog to the bo C p. Note that C p cotais the origi. We complete the proof by applyig the followig simple trasformatio [19, Lemma 2.3] to the set Q = gb: to each poit q Q i the positive orthat of R 2d, we ca assig aother poit q i the positive orthat of R 2d such that for each bo i R 2d that cotais the origi, there is a half-space with the property that q belogs to the bo if ad oly if q belogs to the correspodig half-space. The mappig fb = gb for every B B meets the requiremets of the lemma. Lemma 5. Give ay iteger d 2, a real umber > 0, ad a sufficietly large iteger 0, there eists a set B of ais-parallel boes i R d such that the size of ay -et for the dual set system iduced by B is at least d 2 9 log 1. Proof. Let = α with k N, k 2, ad 1 2 k 1 3 α 2 3. Applyig Lemma 2, we obtai a set B of d 2 k + 32k 2 boes i R d. We claim that the dual rage space iduced by these boes does ot admit a -et of size 1 α B. 4

5 Assume for cotradictio that there is a -et S B with S 1 α B. Accordig to Lemma 2, there eists a set Q of 2 k 1 poits i R d with the property that o bo i S cotais ay poit of Q, but every member of B \ S does. By the pigeohole priciple, there is a poit p Q cotaied i at least B \ S Q α B Q = α B = B 2k 1 members of B \ S. Thus, oe of the at least B members of B hit by p belog to S, cotradictig the assumptio that S was a -et. Hece, the size of ay -et i the dual rage space iduced by B is at least 1 α B = 1 α d 2 k + 32k 2 = 1 αα 2 d 2 k d 2 1 log 1. The system of boes costructed above has a fied umber of elemets, depedig o the value of 1/. We ca obtai arbitrarily large costructios by replacig each bo of B B with several slightly traslated copies of B we refer the reader to [19] for details. Now we are i a positio to establish Theorem 1. By Lemma 4, ay lower boud for the size of -ets i the dual rage space iduced by the set B of boes i R d gives the same lower boud for the size of a -et i the primal rage space o the set of poits fb R 2d correspodig to these boes, i which the rages are half-spaces i R 2d. For ay iteger d 4 ad ay real > 0, Lemma 5 guaratees the eistece of a set B of ais-parallel boes i R d/2 such that ay -et for the dual set system iduced by B has size at least d/2 2 9 log 1 = d/4 9 boud. 3 Proof of Lemma 2 The proof of Lemma 2 is based o the followig key statemet. log 1. This fact, together with Lemma 4, implies the stated Lemma C [19]. Let k 2 be a iteger. The there eists a set R of k + 32 k 2 ais-parallel rectagles i R 2 such that for ay S R, there eists a 2 k 1 -elemet set Q of poits i R 2 with the property that i Q R for ay R R \ S, ad ii Q R = for ay R S. Deote the - ad y-coordiates of a poit p R 2 by p ad yp respectively, ad set m = d 2. Let R = {R 1,..., R t }, t = k + 32 k 2, be a set of rectagles satisfyig the coditios of Lemma C. By scalig, oe ca assume that R [0, 1] 2 for every R R. Give that a bo i R d is the product of d itervals, the idea of the costructio is to lift the rectagles i Lemma C, i.e., the set R, to boes i R d. So a rectagle R R ca be mapped to a bo i R d which is the product d itervals: the first two beig the itervals defiig R, ad the other d 2 itervals i the product beig the full iterval [0, 1]. Oe ca the agai lift the same set R i a o-iterferig way by mappig R to a bo whose 3-rd ad 4-th itervals are the itervals of R ad the remaiig itervals are [0, 1]. I this way, by packig itervals of each R R ito disjoit coordiates, oe ca lift R m times to get a set of d 2 R boes i Rd. Formally, for i = 1... m, defie the ijective fuctios f i that map a poit i R 2 to a product of d itervals i R d, as follows. f i p = [0, 1] [0, 1] p yp [0, 1] [0, 1], p R 2. }{{}}{{} 2i 2 itervals d 2i itervals This mappig lifts each rectagle R R to the bo f i R = {f i p : p R}, ad each set of rectagles R R to the set of boes f i R = {f i R : R R }. We ow show that B = m i=1 f ir is the desired set of d 2 k + 32k 2 boes i R d. Let S B be a fied set of boes. For ay ide i [1, m], set R i R to be the set of preimage rectagles uder f i of the boes i S f i R, i.e., R i satisfies S f i R = f i R i. Let Q i = {q1, i..., q i 2 } R 2 be the set of k 1 poits hittig all rectagles i R \ R i ad o rectagle i R i ; such a set eists by Lemma C. Now we argue that the set { q 1 j, yqj 1,..., qj m, yqj m : j [1, 2 ]} k 1 if d is eve, Q = { q 1 j, yqj 1,..., qj m, yqj m, 1 : j [1, 2 ]} k 1 if d is odd, 5

6 of 2 k 1 poits i R d is the required set for S; i.e., Q hits all the boes i B \ S, ad oe of the boes i B S. Take ay bo B B \ S; the there eists a ide i ad a rectagle R R \ R i such that R is the preimage rectagle of B uder f i. By Lemma C, R cotais a poit q Q i, ad thus B = f i R cotais the poit q Q with q ad yq i its 2i 1-th ad 2i-th coordiates, as all the remaiig itervals defiig B are [0, 1] ad so each such iterval cotais the correspodig coordiate of q. O the other had, let B S be a bo with the preimage rectagle R R i. By Lemma C, R is ot hit by ay poit of Q i, ad thus for ay poit q Q, the 2i 1-th ad 2i-th coordiates caot both be cotaied i the correspodig two itervals defiig B. Therefore, q does ot hit B. 4 Proof of Theorem 3 The goal of this sectio is to establish lower bouds o the sizes of -ets i rage spaces with give shallow-cell compleity ϕ, where ϕ is a submultiplicative fuctio. We will use the followig property of submultiplicative fuctios. Claim 6. Let ϕ : R + R + be a submultiplicative fuctio. The i for all sufficietly large, y R +, we have ϕy ϕϕy, ad ii if there eists a sufficietly large R + ad a costat c such that ϕ c, the ϕ c for every. Proof. Both of these properties follow immediately from the submultiplicativity of ϕ : i. ϕy = ϕy log y ϕy logy y ϕ y log y ϕ y log y y = ϕ ϕy. ii. ϕ log ϕ c = ϕ c log = c log = c. Theorem 3 is a cosequece of the followig more precise statemet. Theorem 7. Let ϕ : R + R + be a mootoically icreasig submultiplicative fuctio which teds to ifiity ad is bouded from above by a polyomial of costat degree. For ay 0 < δ < 1 10, oe ca fid a 0 > 0 with the followig property: for ay 0 < < 0, there eists a rage space with shallow-cell compleity ϕ o a set of = log ϕ 1 elemets, i which the size of ay -et is at least 1 2 δ log ϕ 1. Proof. The parameters of the rage space are as follows: = log ϕ 1, m = = log ϕ 1 ϕ 1 2δ, p =. m Let d be the smallest iteger such that ϕ = O d. By Claim 6, part ii, for ay large eough, we have d 1 ϕ c 1 d, for a suitable costat c 1 1. I the most iterestig case where ϕ = o, we have d = 1. For a small eough, we have c 1 log ϕ 1, so that m = log ϕ 1 log c1 d d log c 1 d log. 1 Cosider a rage space [], F with a groud set [] = {1, 2,..., } ad with a system of m-elemet subsets F, where each m-elemet subset of [] is added to F idepedetly with probability p. The et claim follows by a routie applicatio of the Cheroff boud. Claim 8. With high probability, F 2ϕ 1 2δ. Theorem 7 follows by combiig the et two lemmas that show that, with high probability, the rage space [], F i does ot admit a -et of size less tha 1 2 δ log ϕ 1, ad ii has shallow-cell compleity ϕ. For the proofs, we eed to assume that = δ, d, ϕ is a sufficietly large costat, or, equivaletly, that 0 = 0 δ, d is sufficietly small. Lemma 9. With high probability, the rage space [], F has shallow-cell compleity ϕ. 6

7 The reaso why this lemma holds is that for ay X [] ad ay k, it is very ulikely that the umber of at most k-elemet sets eceeds the umber permitted by the shallow-cell compleity coditio. To boud the probability of the uio of these evets for all X ad k, we simply use the uio boud. Proof. It is eough to show that for all sufficietly large 0, every X [], X =, ad every l m, the umber of sets of size eactly l i F X is Oϕ, as this implies that the umber of sets i F X of size at most l is O ϕl. I the computatios below, we will also assume that l d + 1 2; otherwise if l d, ad assumig 0 2d, we have l d ϕ, d where the last iequality follows by the assumptio o ϕ, provided that is sufficietly large. We distiguish two cases. Case 1: > probability, we have ϕ δ/d. I this case, we trivially upper-boud F X by F. By Claim 8, with high F 2 ϕ 1 2δ 2 ϕ ϕ 1 2δ by Claim 6 2 ϕ ϕ ϕ δ/d 1 2δ as ϕδ/d 1 2δ usig 2 c 1 ϕϕ δ ϕt c1 t d 2c 1ϕ 1 δ 2c 1ϕ 1 δ+δ/d = Oϕ. Case 2: ϕ δ/d. Deote the largest iteger that satisfies this iequality by 1. It is clear that 1 = o recall that ϕ is mootoically icreasig ad teds to ifiity. We also deote the system of all l-elemet subsets of F X by F l X ad the set of all l-elemet subsets of X by X l. Let E be the evet that F does ot have the required ϕ -shallow-cell compleity property. The Pr[E] m l=2 Pr[E l], where E l is the evet that for some X [], X =, there are more tha ϕ elemets i F l X. The, for ay fied l d + 1 2, we have Pr[E l ] 1 = 0 Pr 1 = 0 1 = 0 1 = 0 1 [ ] X [], X =, F l X > ϕ l [ ] Pr For a fied X, X =, {S F X, S = l} = s s= ϕ l s= ϕ l s= ϕ l = 0 s= ϕ 1 l = 0 s= ϕ 1 l = 0 s= ϕ l s Pr [ For a fied X, X =, S ] we have F l X = S X, S = s, l l 1 1 p m l s 1 p m l l s 2 s e e e l s e l e l+1 l 1 l l ϕ p m s s p 3 m l m l s m l 4 e em l 1 e 2 mϕ 1 2δ s 5 ϕ 7

8 I the trasitio to the epressio 3, we used several times i the boud a b ea b b for ay a, b N; ii the iequality 1 p b 1 bp for ay iteger b 1 ad real 0 p 1; ad iii we upper-bouded the last factor of 2 by 1. I the trasitio from 3 to 4 we lower-bouded s by ϕ. We also used the estimate m l m m l, which ca be verified as follows. m l = m l m m l 1 m i m + i + 1 i=0 l m m m m l m l. Fially, to obtai 5, we substituted the formula for p ad used the fact that l l m l = l m l l l l, 2 as 1 = o, m = 4 for < 0 1/4 ad l 2. Deote 2 = 1 δ. We split the epressio 5 ito two sums Σ 1 ad Σ 2. Let Σ 1 := Σ 2 := l = 0 s= ϕ 1 l = 2 s= ϕ e em l 1 e 2 mϕ 1 2δ ϕ e em l 1 e 2 mϕ 1 2δ ϕ s s These two sums will be bouded separately. We have Σ 1 l = 0 s= ϕ l = 0 s= ϕ l = 0 s= ϕ e e e em em s l 1 c 1 2δ 1 e 2 m d 2dδ d 2dδ ϕ 2δ 6 l 1 d+2dδ s Cm d+1 2dδ for some C > 0 δ/2 l 1 d+2dδ s Cm d+1 7 e ϕ l δ 2 2dδ δ2 e 2 l ϕdδ2 2 8 = 0 = 0 2 ϕdδ = o m = 0 = 0 To obtai 6, we used the property that ϕ ϕϕ c 1ϕ d, provided that,, are sufficietly large. To establish 7, we used the fact that 2 = 1 δ ad that em ed log δ/2. I the trasitio to 8, we eeded that l d + 1, d 1 ad that Cm d+1 Cd log d+1 = o δ2 /2. The we lower-bouded s by ϕ. To arrive at 9, we used that l. The last iequality follows from the fact that 0 is large eough, so that ϕ ϕ 0 8/dδ 2 ad that m = o. Net, we tur to boudig Σ 2. First, observe that ϕ 1 2δ ϕ 1 2δ 1 δ 1 δ ϕ 1 2δ 1 δ ϕ 1 δ, where we used the submultiplicativity ad mootoicity of the fuctio ϕ ad the fact that 2 = 1 δ. Secod, ote by that restrictig to be small eough, by the submultiplicativity of ϕ, we have that for ay 2, m log ϕ 1 ϕδ/4 1 ϕδ/3 2 ϕ δ/

9 Substitutig the boud for ϕ 1 2δ i Σ 2, settig C = e 2, ad by 10, we obtai Σ 2 1 l = 2 s= ϕ 1 l = 2 1 = 2 1 = 2 = 1 1 e e em l 1 s Cϕ 2δ/3 em ϕ Cϕ 2δ/3 11 ϕ ϕ e 1+/ϕ m l/ϕ Cϕ 2δ/3 ϕ C ϕ δ/3 ϕ for some costat C > ϕ 2 2ϕ 2 2 2ϕ 2 Cϕ δ/ ϕ 2 2 = o 1 m. I the trasitio to 11, we used that l 2 ad 1 δ 1 /ϕ δ/d 1, which implies that 1 /ϕ δ/2d ad, therefore, em < e log ϕ1/ e log ϕ ϕ δ/2d ϕ δ/2d ϕ δ/3d. To get 12, we used that for some costat c > 1 we have l/ϕ c m/ϕ c log ϕ/ϕ = O1 ad that m ϕ δ/3 for 2 by 10. To obtai 13, observe that 1/2ϕ2 = O1. At the last equatio, we used that 1 = o, e/ 1 as ad that 2 ϕ = Ω 1 δ. We have show that for every l = 2,..., m, we have Pr[E l ] = o1/m, as m teds to ifiity. Thus, we ca coclude that Pr[E] m l=2 Pr[E l] = o1 ad, hece, with high probability the rage space [], F has shallow-cell compleity ϕ.. Now we are i a positio to prove that with high probability the rage space [], F does ot admit a small -et. Lemma 10. With high probability, the size of ay -et of the rage space [], F is at least 1 2 δ log ϕ 1. Proof. Deote by µ the probability that the rage space has a -et of size t = 1 2 δ log ϕ 1 = 1 2 δ. The we have µ Pr [ X is a -et for F ] 1 p t m e p t m 2 e ϕδ/2 = o1. 14 t t X [] X =t To verify the last two iequalities, otice that, sice 1 a > e b for b > a, 0 < < 1 a 1 b, we have t m t p p m m t = ϕ 1 2δ 1 t ϕ 1 2δ 1 m t m δ t t ϕ 1 2δ e mt δ = ϕ 1 2δ e 1 2δ 1+δ log ϕ 1 = ϕ 1 2δ ϕ 1 2δ 1+δ 1 ϕ 1 2δ ϕ 1 2δ 1+δ ϕ δ/2. 13 Here the last but oe iequality follows from 1 holds, because δ 1/10. ad from the mootoicity of ϕ. The last iequality Thus, Lemma 9 ad Lemma 10 imply that with high probability the rage space [], F has shallow-cell compleity ϕ ad it admits o -et of size less tha 1 2 δ log ϕ 1. This completes the proof of the theorem. 9

10 Ackowledgemets. We thak the aoymous referees for carefully readig our mauscript ad for poitig out several problems i the presetatio, icludig a mistake i the proof of Lemma 10. A prelimiary versio of this paper was accepted i SoCG 2016 Symposium o Computatioal Geometry. We also thak the aoymous reviewers of the coferece versio for their feedback ad valuable suggestios. Refereces [1] P. K. Agarwal, J. Pach, ad M. Sharir. State of the uio of geometric objects: A review. I J. Goodma, J. Pach, ad R. Pollack, editors, Computatioal Geometry: Twety Years Later, pages America Mathematical Society, [2] N. Alo. A o-liear lower boud for plaar epsilo-ets. Discrete & Computatioal Geometry, 472: , [3] B. Aroov, E. Ezra, ad M. Sharir. Small-size -ets for ais-parallel rectagles ad boes. SIAM J. Comput., 397: , [4] P. Ashok, U. Azmi, ad S. Govidaraja. Small strog epsilo ets. Comput. Geom., 479: , [5] N. Bus, S. Garg, N. H. Mustafa, ad S. Ray. Tighter estimates for epsilo-ets for disks. Comput. Geom., 53:27 35, [6] T. M. Cha, E. Grat, J. Köema, ad M. Sharpe. Weighted capacitated, priority, ad geometric set cover via improved quasi-uiform samplig. I Proceedigs of Symposium o Discrete Algorithms SODA, pages , [7] B. Chazelle ad J. Friedma. A determiistic view of radom samplig ad its use i geometry. Combiatorica, 103: , [8] K. Clarkso ad K. Varadaraja. Improved approimatio algorithms for geometric set cover. Discrete & Computatioal Geometry, 37:43 58, [9] K. L. Clarkso ad P. W. Shor. Applicatio of radom samplig i computatioal geometry, II. Discrete & Computatioal Geometry, 4: , [10] D. Haussler ad E. Welzl. Epsilo-ets ad simple rage queries. Discrete & Computatioal Geometry, 2: , [11] J. Komlós, J. Pach, ad G. J. Woegiger. Almost tight bouds for epsilo-ets. Discrete & Computatioal Geometry, 7: , [12] J. Matoušek. O costats for cuttigs i the plae. Discrete & Computatioal Geometry, 204: , [13] J. Matoušek. Lectures i Discrete Geometry. Spriger-Verlag, New York, NY, [14] J. Matoušek, R. Seidel, ad E. Welzl. How to et a lot with little: Small epsilo-ets for disks ad halfspaces. I Proceedigs of Symposium o Computatioal Geometry, pages 16 22, [15] N. H. Mustafa, K. Dutta, ad A. Ghosh. A simple proof of optimal epsilo-ets. Combiatorica, to appear. [16] N. H. Mustafa ad S. Ray. Near-optimal geeralisatios of a theorem of Macbeath. I Proceedigs of the Symposium o Theoretical Aspects of Computer Sciece STACS, pages , [17] N. H. Mustafa ad K. Varadaraja. Epsilo-approimatios ad epsilo-ets. I J. E. Goodma, J. O Rourke, ad C. D. Tóth, editors, Hadbook of Discrete ad Computatioal Geometry. CRC Press LLC, 2016, to appear. [18] J. Pach ad P. K. Agarwal. Combiatorial Geometry. Joh Wiley & Sos, New York, NY, [19] J. Pach ad G. Tardos. Tight lower bouds for the size of epsilo-ets. Joural of the AMS, 26: ,

11 [20] E. Pyrga ad S. Ray. New eistece proofs for epsilo-ets. I Proceedigs of the Symposium o Computatioal Geometry SoCG, pages , [21] N. Sauer. O the desity of families of sets. Joural of Combiatorial Theory, Series A, 13:145147, [22] Micha Sharir. O k-sets i arragemet of curves ad surfaces. Discrete & Computatioal Geometry, 6: , [23] S. Shelah. A combiatorial problem; stability ad order for models ad theories i ifiitary laguages. Pacific Joural of Mathematics, 41: , [24] V. N. Vapik ad A. Ya. Chervoekis. O the uiform covergece of relative frequecies of evets to their probabilities. Theory of Probability ad its Applicatios, 162: , [25] K. Varadaraja. Epsilo ets ad uio compleity. I Proceedigs of the Symposium o Computatioal Geometry SoCG, pages 11 16, [26] K. Varadaraja. Weighted geometric set cover via quasi uiform samplig. I Proceedigs of the Symposium o Theory of Computig STOC, pages , New York, USA, ACM. 11

A Simple Proof of the Shallow Packing Lemma

A Simple Proof of the Shallow Packing Lemma A Simple Proof of the Shallow Packig Lemma Nabil Mustafa To cite this versio: Nabil Mustafa. A Simple Proof of the Shallow Packig Lemma. Discrete ad Computatioal Geometry, Spriger Verlag, 06, 55 (3), pp.739-743.

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

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

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

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

if > 6 is sucietly large). Nevertheless, Pichasi has show that the umber of radial poits of a o-colliear set P of poits i the plae that lie i a halfpl

if > 6 is sucietly large). Nevertheless, Pichasi has show that the umber of radial poits of a o-colliear set P of poits i the plae that lie i a halfpl Radial Poits i the Plae Jaos Pach y Micha Sharir z Jauary 6, 00 Abstract A radial poit for a ite set P i the plae is a poit q 6 P with the property that each lie coectig q to a poit of P passes through

More information

An Introduction to Randomized Algorithms

An Introduction to Randomized Algorithms A Itroductio to Radomized Algorithms The focus of this lecture is to study a radomized algorithm for quick sort, aalyze it usig probabilistic recurrece relatios, ad also provide more geeral tools for aalysis

More information

Lecture 2. The Lovász Local Lemma

Lecture 2. The Lovász Local Lemma Staford Uiversity Sprig 208 Math 233A: No-costructive methods i combiatorics Istructor: Ja Vodrák Lecture date: Jauary 0, 208 Origial scribe: Apoorva Khare Lecture 2. The Lovász Local Lemma 2. Itroductio

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

Product measures, Tonelli s and Fubini s theorems For use in MAT3400/4400, autumn 2014 Nadia S. Larsen. Version of 13 October 2014.

Product measures, Tonelli s and Fubini s theorems For use in MAT3400/4400, autumn 2014 Nadia S. Larsen. Version of 13 October 2014. Product measures, Toelli s ad Fubii s theorems For use i MAT3400/4400, autum 2014 Nadia S. Larse Versio of 13 October 2014. 1. Costructio of the product measure The purpose of these otes is to preset the

More information

If a subset E of R contains no open interval, is it of zero measure? For instance, is the set of irrationals in [0, 1] is of measure zero?

If a subset E of R contains no open interval, is it of zero measure? For instance, is the set of irrationals in [0, 1] is of measure zero? 2 Lebesgue Measure I Chapter 1 we defied the cocept of a set of measure zero, ad we have observed that every coutable set is of measure zero. Here are some atural questios: If a subset E of R cotais a

More information

Solution. 1 Solutions of Homework 1. Sangchul Lee. October 27, Problem 1.1

Solution. 1 Solutions of Homework 1. Sangchul Lee. October 27, Problem 1.1 Solutio Sagchul Lee October 7, 017 1 Solutios of Homework 1 Problem 1.1 Let Ω,F,P) be a probability space. Show that if {A : N} F such that A := lim A exists, the PA) = lim PA ). Proof. Usig the cotiuity

More information

Many Touchings Force Many Crossings

Many Touchings Force Many Crossings May Touchigs Force May Crossigs Jáos Pach 1, ad Géza Tóth 1 École Polytechique Fédérale de Lausae, St. 8, Lausae 1015, Switzerlad pach@cims.yu.edu Réyi Istitute, Hugaria Academy of Scieces 1364 Budapest,

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

Infinite Sequences and Series

Infinite Sequences and Series Chapter 6 Ifiite Sequeces ad Series 6.1 Ifiite Sequeces 6.1.1 Elemetary Cocepts Simply speakig, a sequece is a ordered list of umbers writte: {a 1, a 2, a 3,...a, a +1,...} where the elemets a i represet

More information

Chapter 3. Strong convergence. 3.1 Definition of almost sure convergence

Chapter 3. Strong convergence. 3.1 Definition of almost sure convergence Chapter 3 Strog covergece As poited out i the Chapter 2, there are multiple ways to defie the otio of covergece of a sequece of radom variables. That chapter defied covergece i probability, covergece i

More information

The log-behavior of n p(n) and n p(n)/n

The log-behavior of n p(n) and n p(n)/n Ramauja J. 44 017, 81-99 The log-behavior of p ad p/ William Y.C. Che 1 ad Ke Y. Zheg 1 Ceter for Applied Mathematics Tiaji Uiversity Tiaji 0007, P. R. Chia Ceter for Combiatorics, LPMC Nakai Uivercity

More information

Advanced Stochastic Processes.

Advanced Stochastic Processes. Advaced Stochastic Processes. David Gamarik LECTURE 2 Radom variables ad measurable fuctios. Strog Law of Large Numbers (SLLN). Scary stuff cotiued... Outlie of Lecture Radom variables ad measurable fuctios.

More information

Application to Random Graphs

Application to Random Graphs A Applicatio to Radom Graphs Brachig processes have a umber of iterestig ad importat applicatios. We shall cosider oe of the most famous of them, the Erdős-Réyi radom graph theory. 1 Defiitio A.1. Let

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

Lecture 3 The Lebesgue Integral

Lecture 3 The Lebesgue Integral Lecture 3: The Lebesgue Itegral 1 of 14 Course: Theory of Probability I Term: Fall 2013 Istructor: Gorda Zitkovic Lecture 3 The Lebesgue Itegral The costructio of the itegral Uless expressly specified

More information

It is always the case that unions, intersections, complements, and set differences are preserved by the inverse image of a function.

It is always the case that unions, intersections, complements, and set differences are preserved by the inverse image of a function. MATH 532 Measurable Fuctios Dr. Neal, WKU Throughout, let ( X, F, µ) be a measure space ad let (!, F, P ) deote the special case of a probability space. We shall ow begi to study real-valued fuctios defied

More information

Seunghee Ye Ma 8: Week 5 Oct 28

Seunghee Ye Ma 8: Week 5 Oct 28 Week 5 Summary I Sectio, we go over the Mea Value Theorem ad its applicatios. I Sectio 2, we will recap what we have covered so far this term. Topics Page Mea Value Theorem. Applicatios of the Mea Value

More information

Lecture 2 Long paths in random graphs

Lecture 2 Long paths in random graphs Lecture Log paths i radom graphs 1 Itroductio I this lecture we treat the appearace of log paths ad cycles i sparse radom graphs. will wor with the probability space G(, p) of biomial radom graphs, aalogous

More information

A Note on Matrix Rigidity

A Note on Matrix Rigidity A Note o Matrix Rigidity Joel Friedma Departmet of Computer Sciece Priceto Uiversity Priceto, NJ 08544 Jue 25, 1990 Revised October 25, 1991 Abstract I this paper we give a explicit costructio of matrices

More information

A Simple Proof of Optimal Epsilon Nets

A Simple Proof of Optimal Epsilon Nets A Simple Proof of Optimal Epsilon Nets Nabil H. Mustafa Kunal Dutta Arijit Ghosh Abstract Showing the existence of ɛ-nets of small size has been the subject of investigation for almost 30 years, starting

More information

Math 61CM - Solutions to homework 3

Math 61CM - Solutions to homework 3 Math 6CM - Solutios to homework 3 Cédric De Groote October 2 th, 208 Problem : Let F be a field, m 0 a fixed oegative iteger ad let V = {a 0 + a x + + a m x m a 0,, a m F} be the vector space cosistig

More information

Machine Learning Theory Tübingen University, WS 2016/2017 Lecture 3

Machine Learning Theory Tübingen University, WS 2016/2017 Lecture 3 Machie Learig Theory Tübige Uiversity, WS 06/07 Lecture 3 Tolstikhi Ilya Abstract I this lecture we will prove the VC-boud, which provides a high-probability excess risk boud for the ERM algorithm whe

More information

3. Z Transform. Recall that the Fourier transform (FT) of a DT signal xn [ ] is ( ) [ ] = In order for the FT to exist in the finite magnitude sense,

3. Z Transform. Recall that the Fourier transform (FT) of a DT signal xn [ ] is ( ) [ ] = In order for the FT to exist in the finite magnitude sense, 3. Z Trasform Referece: Etire Chapter 3 of text. Recall that the Fourier trasform (FT) of a DT sigal x [ ] is ω ( ) [ ] X e = j jω k = xe I order for the FT to exist i the fiite magitude sese, S = x [

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

4 The Sperner property.

4 The Sperner property. 4 The Sperer property. I this sectio we cosider a surprisig applicatio of certai adjacecy matrices to some problems i extremal set theory. A importat role will also be played by fiite groups. I geeral,

More information

On the Linear Complexity of Feedback Registers

On the Linear Complexity of Feedback Registers O the Liear Complexity of Feedback Registers A. H. Cha M. Goresky A. Klapper Northeaster Uiversity Abstract I this paper, we study sequeces geerated by arbitrary feedback registers (ot ecessarily feedback

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

Integrable Functions. { f n } is called a determining sequence for f. If f is integrable with respect to, then f d does exist as a finite real number

Integrable Functions. { f n } is called a determining sequence for f. If f is integrable with respect to, then f d does exist as a finite real number MATH 532 Itegrable Fuctios Dr. Neal, WKU We ow shall defie what it meas for a measurable fuctio to be itegrable, show that all itegral properties of simple fuctios still hold, ad the give some coditios

More information

6.3 Testing Series With Positive Terms

6.3 Testing Series With Positive Terms 6.3. TESTING SERIES WITH POSITIVE TERMS 307 6.3 Testig Series With Positive Terms 6.3. Review of what is kow up to ow I theory, testig a series a i for covergece amouts to fidig the i= sequece of partial

More information

Riesz-Fischer Sequences and Lower Frame Bounds

Riesz-Fischer Sequences and Lower Frame Bounds Zeitschrift für Aalysis ud ihre Aweduge Joural for Aalysis ad its Applicatios Volume 1 (00), No., 305 314 Riesz-Fischer Sequeces ad Lower Frame Bouds P. Casazza, O. Christese, S. Li ad A. Lider Abstract.

More information

The Borel hierarchy classifies subsets of the reals by their topological complexity. Another approach is to classify them by size.

The Borel hierarchy classifies subsets of the reals by their topological complexity. Another approach is to classify them by size. Lecture 7: Measure ad Category The Borel hierarchy classifies subsets of the reals by their topological complexity. Aother approach is to classify them by size. Filters ad Ideals The most commo measure

More information

b i u x i U a i j u x i u x j

b i u x i U a i j u x i u x j M ath 5 2 7 Fall 2 0 0 9 L ecture 1 9 N ov. 1 6, 2 0 0 9 ) S ecod- Order Elliptic Equatios: Weak S olutios 1. Defiitios. I this ad the followig two lectures we will study the boudary value problem Here

More information

Commutativity in Permutation Groups

Commutativity in Permutation Groups Commutativity i Permutatio Groups Richard Wito, PhD Abstract I the group Sym(S) of permutatios o a oempty set S, fixed poits ad trasiet poits are defied Prelimiary results o fixed ad trasiet poits are

More information

6 Integers Modulo n. integer k can be written as k = qn + r, with q,r, 0 r b. So any integer.

6 Integers Modulo n. integer k can be written as k = qn + r, with q,r, 0 r b. So any integer. 6 Itegers Modulo I Example 2.3(e), we have defied the cogruece of two itegers a,b with respect to a modulus. Let us recall that a b (mod ) meas a b. We have proved that cogruece is a equivalece relatio

More information

MA131 - Analysis 1. Workbook 3 Sequences II

MA131 - Analysis 1. Workbook 3 Sequences II MA3 - Aalysis Workbook 3 Sequeces II Autum 2004 Cotets 2.8 Coverget Sequeces........................ 2.9 Algebra of Limits......................... 2 2.0 Further Useful Results........................

More information

Axioms of Measure Theory

Axioms of Measure Theory MATH 532 Axioms of Measure Theory Dr. Neal, WKU I. The Space Throughout the course, we shall let X deote a geeric o-empty set. I geeral, we shall ot assume that ay algebraic structure exists o X so that

More information

Chapter 6 Infinite Series

Chapter 6 Infinite Series Chapter 6 Ifiite Series I the previous chapter we cosidered itegrals which were improper i the sese that the iterval of itegratio was ubouded. I this chapter we are goig to discuss a topic which is somewhat

More information

Definition 4.2. (a) A sequence {x n } in a Banach space X is a basis for X if. unique scalars a n (x) such that x = n. a n (x) x n. (4.

Definition 4.2. (a) A sequence {x n } in a Banach space X is a basis for X if. unique scalars a n (x) such that x = n. a n (x) x n. (4. 4. BASES I BAACH SPACES 39 4. BASES I BAACH SPACES Sice a Baach space X is a vector space, it must possess a Hamel, or vector space, basis, i.e., a subset {x γ } γ Γ whose fiite liear spa is all of X ad

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

Math 216A Notes, Week 5

Math 216A Notes, Week 5 Math 6A Notes, Week 5 Scribe: Ayastassia Sebolt Disclaimer: These otes are ot early as polished (ad quite possibly ot early as correct) as a published paper. Please use them at your ow risk.. Thresholds

More information

PRACTICE FINAL/STUDY GUIDE SOLUTIONS

PRACTICE FINAL/STUDY GUIDE SOLUTIONS Last edited December 9, 03 at 4:33pm) Feel free to sed me ay feedback, icludig commets, typos, ad mathematical errors Problem Give the precise meaig of the followig statemets i) a f) L ii) a + f) L iii)

More information

1 Review and Overview

1 Review and Overview DRAFT a fial versio will be posted shortly CS229T/STATS231: Statistical Learig Theory Lecturer: Tegyu Ma Lecture #3 Scribe: Migda Qiao October 1, 2013 1 Review ad Overview I the first half of this course,

More information

Lecture 4: April 10, 2013

Lecture 4: April 10, 2013 TTIC/CMSC 1150 Mathematical Toolkit Sprig 01 Madhur Tulsiai Lecture 4: April 10, 01 Scribe: Haris Agelidakis 1 Chebyshev s Iequality recap I the previous lecture, we used Chebyshev s iequality to get a

More information

(A sequence also can be thought of as the list of function values attained for a function f :ℵ X, where f (n) = x n for n 1.) x 1 x N +k x N +4 x 3

(A sequence also can be thought of as the list of function values attained for a function f :ℵ X, where f (n) = x n for n 1.) x 1 x N +k x N +4 x 3 MATH 337 Sequeces Dr. Neal, WKU Let X be a metric space with distace fuctio d. We shall defie the geeral cocept of sequece ad limit i a metric space, the apply the results i particular to some special

More information

8. Applications To Linear Differential Equations

8. Applications To Linear Differential Equations 8. Applicatios To Liear Differetial Equatios 8.. Itroductio 8.. Review Of Results Cocerig Liear Differetial Equatios Of First Ad Secod Orders 8.3. Eercises 8.4. Liear Differetial Equatios Of Order N 8.5.

More information

SOME GENERALIZATIONS OF OLIVIER S THEOREM

SOME GENERALIZATIONS OF OLIVIER S THEOREM SOME GENERALIZATIONS OF OLIVIER S THEOREM Alai Faisat, Sait-Étiee, Georges Grekos, Sait-Étiee, Ladislav Mišík Ostrava (Received Jauary 27, 2006) Abstract. Let a be a coverget series of positive real umbers.

More information

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.436J/15.085J Fall 2008 Lecture 19 11/17/2008 LAWS OF LARGE NUMBERS II THE STRONG LAW OF LARGE NUMBERS

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.436J/15.085J Fall 2008 Lecture 19 11/17/2008 LAWS OF LARGE NUMBERS II THE STRONG LAW OF LARGE NUMBERS MASSACHUSTTS INSTITUT OF TCHNOLOGY 6.436J/5.085J Fall 2008 Lecture 9 /7/2008 LAWS OF LARG NUMBRS II Cotets. The strog law of large umbers 2. The Cheroff boud TH STRONG LAW OF LARG NUMBRS While the weak

More information

Polynomial identity testing and global minimum cut

Polynomial identity testing and global minimum cut CHAPTER 6 Polyomial idetity testig ad global miimum cut I this lecture we will cosider two further problems that ca be solved usig probabilistic algorithms. I the first half, we will cosider the problem

More information

Chapter 7 Isoperimetric problem

Chapter 7 Isoperimetric problem Chapter 7 Isoperimetric problem Recall that the isoperimetric problem (see the itroductio its coectio with ido s proble) is oe of the most classical problem of a shape optimizatio. It ca be formulated

More information

Lecture 14: Graph Entropy

Lecture 14: Graph Entropy 15-859: Iformatio Theory ad Applicatios i TCS Sprig 2013 Lecture 14: Graph Etropy March 19, 2013 Lecturer: Mahdi Cheraghchi Scribe: Euiwoog Lee 1 Recap Bergma s boud o the permaet Shearer s Lemma Number

More information

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 21 11/27/2013

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 21 11/27/2013 MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 21 11/27/2013 Fuctioal Law of Large Numbers. Costructio of the Wieer Measure Cotet. 1. Additioal techical results o weak covergece

More information

Rademacher Complexity

Rademacher Complexity EECS 598: Statistical Learig Theory, Witer 204 Topic 0 Rademacher Complexity Lecturer: Clayto Scott Scribe: Ya Deg, Kevi Moo Disclaimer: These otes have ot bee subjected to the usual scrutiy reserved for

More information

Chapter 4. Fourier Series

Chapter 4. Fourier Series Chapter 4. Fourier Series At this poit we are ready to ow cosider the caoical equatios. Cosider, for eample the heat equatio u t = u, < (4.) subject to u(, ) = si, u(, t) = u(, t) =. (4.) Here,

More information

Notes on iteration and Newton s method. Iteration

Notes on iteration and Newton s method. Iteration Notes o iteratio ad Newto s method Iteratio Iteratio meas doig somethig over ad over. I our cotet, a iteratio is a sequece of umbers, vectors, fuctios, etc. geerated by a iteratio rule of the type 1 f

More information

Discrete Mathematics for CS Spring 2008 David Wagner Note 22

Discrete Mathematics for CS Spring 2008 David Wagner Note 22 CS 70 Discrete Mathematics for CS Sprig 2008 David Wager Note 22 I.I.D. Radom Variables Estimatig the bias of a coi Questio: We wat to estimate the proportio p of Democrats i the US populatio, by takig

More information

Chapter 0. Review of set theory. 0.1 Sets

Chapter 0. Review of set theory. 0.1 Sets Chapter 0 Review of set theory Set theory plays a cetral role i the theory of probability. Thus, we will ope this course with a quick review of those otios of set theory which will be used repeatedly.

More information

Largest families without an r-fork

Largest families without an r-fork Largest families without a r-for Aalisa De Bois Uiversity of Salero Salero, Italy debois@math.it Gyula O.H. Katoa Réyi Istitute Budapest, Hugary ohatoa@reyi.hu Itroductio Let [] = {,,..., } be a fiite

More information

62. Power series Definition 16. (Power series) Given a sequence {c n }, the series. c n x n = c 0 + c 1 x + c 2 x 2 + c 3 x 3 +

62. Power series Definition 16. (Power series) Given a sequence {c n }, the series. c n x n = c 0 + c 1 x + c 2 x 2 + c 3 x 3 + 62. Power series Defiitio 16. (Power series) Give a sequece {c }, the series c x = c 0 + c 1 x + c 2 x 2 + c 3 x 3 + is called a power series i the variable x. The umbers c are called the coefficiets of

More information

1 Review and Overview

1 Review and Overview CS9T/STATS3: Statistical Learig Theory Lecturer: Tegyu Ma Lecture #6 Scribe: Jay Whag ad Patrick Cho October 0, 08 Review ad Overview Recall i the last lecture that for ay family of scalar fuctios F, we

More information

Basics of Probability Theory (for Theory of Computation courses)

Basics of Probability Theory (for Theory of Computation courses) Basics of Probability Theory (for Theory of Computatio courses) Oded Goldreich Departmet of Computer Sciece Weizma Istitute of Sciece Rehovot, Israel. oded.goldreich@weizma.ac.il November 24, 2008 Preface.

More information

Bertrand s Postulate

Bertrand s Postulate Bertrad s Postulate Lola Thompso Ross Program July 3, 2009 Lola Thompso (Ross Program Bertrad s Postulate July 3, 2009 1 / 33 Bertrad s Postulate I ve said it oce ad I ll say it agai: There s always a

More information

arxiv: v3 [math.co] 6 Aug 2014

arxiv: v3 [math.co] 6 Aug 2014 NEAR PERFECT MATCHINGS IN -UNIFORM HYPERGRAPHS arxiv:1404.1136v3 [math.co] 6 Aug 2014 JIE HAN Abstract. Let H be a -uiform hypergraph o vertices where is a sufficietly large iteger ot divisible by. We

More information

Lecture 15: Learning Theory: Concentration Inequalities

Lecture 15: Learning Theory: Concentration Inequalities STAT 425: Itroductio to Noparametric Statistics Witer 208 Lecture 5: Learig Theory: Cocetratio Iequalities Istructor: Ye-Chi Che 5. Itroductio Recall that i the lecture o classificatio, we have see that

More information

Distribution of Random Samples & Limit theorems

Distribution of Random Samples & Limit theorems STAT/MATH 395 A - PROBABILITY II UW Witer Quarter 2017 Néhémy Lim Distributio of Radom Samples & Limit theorems 1 Distributio of i.i.d. Samples Motivatig example. Assume that the goal of a study is to

More information

Subject: Differential Equations & Mathematical Modeling-III

Subject: Differential Equations & Mathematical Modeling-III Power Series Solutios of Differetial Equatios about Sigular poits Subject: Differetial Equatios & Mathematical Modelig-III Lesso: Power series solutios of differetial equatios about Sigular poits Lesso

More information

Singular Continuous Measures by Michael Pejic 5/14/10

Singular Continuous Measures by Michael Pejic 5/14/10 Sigular Cotiuous Measures by Michael Peic 5/4/0 Prelimiaries Give a set X, a σ-algebra o X is a collectio of subsets of X that cotais X ad ad is closed uder complemetatio ad coutable uios hece, coutable

More information

Notes 27 : Brownian motion: path properties

Notes 27 : Brownian motion: path properties Notes 27 : Browia motio: path properties Math 733-734: Theory of Probability Lecturer: Sebastie Roch Refereces:[Dur10, Sectio 8.1], [MP10, Sectio 1.1, 1.2, 1.3]. Recall: DEF 27.1 (Covariace) Let X = (X

More information

Lecture 2: April 3, 2013

Lecture 2: April 3, 2013 TTIC/CMSC 350 Mathematical Toolkit Sprig 203 Madhur Tulsiai Lecture 2: April 3, 203 Scribe: Shubhedu Trivedi Coi tosses cotiued We retur to the coi tossig example from the last lecture agai: Example. Give,

More information

Almost intersecting families of sets

Almost intersecting families of sets Almost itersectig families of sets Dáiel Gerber a Natha Lemos b Cory Palmer a Balázs Patkós a, Vajk Szécsi b a Hugaria Academy of Scieces, Alfréd Réyi Istitute of Mathematics, P.O.B. 17, Budapest H-1364,

More information

A NOTE ON INVARIANT SETS OF ITERATED FUNCTION SYSTEMS

A NOTE ON INVARIANT SETS OF ITERATED FUNCTION SYSTEMS Acta Math. Hugar., 2007 DOI: 10.1007/s10474-007-7013-6 A NOTE ON INVARIANT SETS OF ITERATED FUNCTION SYSTEMS L. L. STACHÓ ad L. I. SZABÓ Bolyai Istitute, Uiversity of Szeged, Aradi vértaúk tere 1, H-6720

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

CHAPTER 5. Theory and Solution Using Matrix Techniques

CHAPTER 5. Theory and Solution Using Matrix Techniques A SERIES OF CLASS NOTES FOR 2005-2006 TO INTRODUCE LINEAR AND NONLINEAR PROBLEMS TO ENGINEERS, SCIENTISTS, AND APPLIED MATHEMATICIANS DE CLASS NOTES 3 A COLLECTION OF HANDOUTS ON SYSTEMS OF ORDINARY DIFFERENTIAL

More information

7.1 Convergence of sequences of random variables

7.1 Convergence of sequences of random variables Chapter 7 Limit theorems Throughout this sectio we will assume a probability space (Ω, F, P), i which is defied a ifiite sequece of radom variables (X ) ad a radom variable X. The fact that for every ifiite

More information

Notes for Lecture 11

Notes for Lecture 11 U.C. Berkeley CS78: Computatioal Complexity Hadout N Professor Luca Trevisa 3/4/008 Notes for Lecture Eigevalues, Expasio, ad Radom Walks As usual by ow, let G = (V, E) be a udirected d-regular graph with

More information

Law of the sum of Bernoulli random variables

Law of the sum of Bernoulli random variables Law of the sum of Beroulli radom variables Nicolas Chevallier Uiversité de Haute Alsace, 4, rue des frères Lumière 68093 Mulhouse icolas.chevallier@uha.fr December 006 Abstract Let be the set of all possible

More information

Math Solutions to homework 6

Math Solutions to homework 6 Math 175 - Solutios to homework 6 Cédric De Groote November 16, 2017 Problem 1 (8.11 i the book): Let K be a compact Hermitia operator o a Hilbert space H ad let the kerel of K be {0}. Show that there

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

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

Topic 9: Sampling Distributions of Estimators

Topic 9: Sampling Distributions of Estimators Topic 9: Samplig Distributios of Estimators Course 003, 2018 Page 0 Samplig distributios of estimators Sice our estimators are statistics (particular fuctios of radom variables), their distributio ca be

More information

page Suppose that S 0, 1 1, 2.

page Suppose that S 0, 1 1, 2. page 10 1. Suppose that S 0, 1 1,. a. What is the set of iterior poits of S? The set of iterior poits of S is 0, 1 1,. b. Give that U is the set of iterior poits of S, evaluate U. 0, 1 1, 0, 1 1, S. The

More information

Linear Differential Equations of Higher Order Basic Theory: Initial-Value Problems d y d y dy

Linear Differential Equations of Higher Order Basic Theory: Initial-Value Problems d y d y dy Liear Differetial Equatios of Higher Order Basic Theory: Iitial-Value Problems d y d y dy Solve: a( ) + a ( )... a ( ) a0( ) y g( ) + + + = d d d ( ) Subject to: y( 0) = y0, y ( 0) = y,..., y ( 0) = y

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

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

Lecture 3: August 31

Lecture 3: August 31 36-705: Itermediate Statistics Fall 018 Lecturer: Siva Balakrisha Lecture 3: August 31 This lecture will be mostly a summary of other useful expoetial tail bouds We will ot prove ay of these i lecture,

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

SRC Technical Note June 17, Tight Thresholds for The Pure Literal Rule. Michael Mitzenmacher. d i g i t a l

SRC Technical Note June 17, Tight Thresholds for The Pure Literal Rule. Michael Mitzenmacher. d i g i t a l SRC Techical Note 1997-011 Jue 17, 1997 Tight Thresholds for The Pure Literal Rule Michael Mitzemacher d i g i t a l Systems Research Ceter 130 Lytto Aveue Palo Alto, Califoria 94301 http://www.research.digital.com/src/

More information

ON POINTWISE BINOMIAL APPROXIMATION

ON POINTWISE BINOMIAL APPROXIMATION Iteratioal Joural of Pure ad Applied Mathematics Volume 71 No. 1 2011, 57-66 ON POINTWISE BINOMIAL APPROXIMATION BY w-functions K. Teerapabolar 1, P. Wogkasem 2 Departmet of Mathematics Faculty of Sciece

More information

Section 1.1. Calculus: Areas And Tangents. Difference Equations to Differential Equations

Section 1.1. Calculus: Areas And Tangents. Difference Equations to Differential Equations Differece Equatios to Differetial Equatios Sectio. Calculus: Areas Ad Tagets The study of calculus begis with questios about chage. What happes to the velocity of a swigig pedulum as its positio chages?

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

Topic 9: Sampling Distributions of Estimators

Topic 9: Sampling Distributions of Estimators Topic 9: Samplig Distributios of Estimators Course 003, 2018 Page 0 Samplig distributios of estimators Sice our estimators are statistics (particular fuctios of radom variables), their distributio ca be

More information

MA131 - Analysis 1. Workbook 2 Sequences I

MA131 - Analysis 1. Workbook 2 Sequences I MA3 - Aalysis Workbook 2 Sequeces I Autum 203 Cotets 2 Sequeces I 2. Itroductio.............................. 2.2 Icreasig ad Decreasig Sequeces................ 2 2.3 Bouded Sequeces..........................

More information

Assignment 5: Solutions

Assignment 5: Solutions McGill Uiversity Departmet of Mathematics ad Statistics MATH 54 Aalysis, Fall 05 Assigmet 5: Solutios. Let y be a ubouded sequece of positive umbers satisfyig y + > y for all N. Let x be aother sequece

More information

Lecture XVI - Lifting of paths and homotopies

Lecture XVI - Lifting of paths and homotopies Lecture XVI - Liftig of paths ad homotopies I the last lecture we discussed the liftig problem ad proved that the lift if it exists is uiquely determied by its value at oe poit. I this lecture we shall

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