Computing the maximum using (min,+) formulas

Size: px
Start display at page:

Download "Computing the maximum using (min,+) formulas"

Transcription

1 Computig the maximum usig (mi,+) formulas Meea Mahaja 1, Prajakta Nimbhorkar 2, ad Auj Tawari 1 1 The Istitute of Mathematical Scieces, HBNI, Cheai, Idia. {meea,aujvt}@imsc.res.i 2 Cheai Mathematical Istitute, Idia. prajakta@cmi.ac.i Abstract We study computatio by formulas over (mi, +). We cosider the computatio of max{x 1,..., x } over N as a differece of (mi, +) formulas, ad show that size + log is sufficiet ad ecessary. Our proof also shows that ay (mi, +) formula computig the miimum of all sums of 1 out of variables must have log leaves; this too is tight. Our proofs use a complexity measure for (mi, +) fuctios based o miterm-like behaviour ad o the etropy of a associated graph ACM Subject Classificatio F.1.1 Models of Computatio; F.1.3 Complexity Measures ad Classes; F.2.3 Tradeoffs betwee Complexity Measures Keywords ad phrases formulas, circuits, lower bouds, tropical semirig Digital Object Idetifier /LIPIcs.MFCS Itroductio A (mi, +) formula is a formula (tree) i which the leaves are labeled by variables or costats. The iteral odes are gates labeled by either mi or +. A mi gate computes the miimum value amog its iputs while a + gate simply adds the values computed by its iputs. Such formulas ca compute ay fuctio expressible as the miimum over several liear polyomials with o-egative iteger coefficiets. I this work, we cosider the followig problem: Suppose we are give iput variables x 1, x 2,..., x ad we wat to fid a formula which computes the maximum value take by these variables, max(x 1, x 2,..., x ). If variables are restricted to take o-egative iteger values, t is easy to show that o (mi, +) formula ca compute max. Suppose ow we stregthe this model by allowig mius gates as well. Now we have a very small liear sized formula: max(x 1, x 2,..., x ) = 0 mi(0 x 1, 0 x 2,..., 0 x ). It is clear that mius gates add sigificat power to the model of (mi, +) formulas. But how may miuses do we actually eed? It turs out that oly oe mius gate, at the top, is sufficiet. Here is oe such formula: (Sum of all variables) - mi i (Sum of all variables except x i ). The secod expressio above ca be computed by a (mi, +) formula of size log usig recursio. So, we ca compute max usig mi, + ad oe mius gate at the top, at the cost of a slightly super-liear size. Ca we do ay better? We show that this simple differece formula is ideed the best we ca achieve for this model. The mai motivatio behid studyig this questio is the followig questio asked i [8]: Does there exist a aturally occurig fuctio f for which (mi, +) circuits are superpolyomially weaker tha (max, +) circuits? There are two possibilities: 1. Show that max ca be implemeted usig a small (mi, +) circuit. 2. Come up with a explicit fuctio f which has small (max, +) circuits but requires large (mi, +) circuits. Meea Mahaja ad Prajakta Nimbhorkar ad Auj Tawari; licesed uder Creative Commos Licese CC-BY 42d Iteratioal Symposium o Mathematical Foudatios of Computer Sciece (MFCS 2017). Editors: Kim G. Larse, Has L. Bodlaeder, ad Jea-Fracois Raski; Article No. 74; pp. 74:1 74:11 Leibiz Iteratioal Proceedigs i Iformatics Schloss Dagstuhl Leibiz-Zetrum für Iformatik, Dagstuhl Publishig, Germay

2 74:2 Computig max usig (mi,+) formulas Sice we show that o (mi, +) formula (or circuit) ca compute max, optio 1 is ruled out. I the weaker model of formulas istead of circuits, we show that ay differece of (mi, +) formulas computig max should have size at least log. This yields us a separatio betwee (max, +) formulas ad differece of (mi, +) formulas. Backgroud May dyamic programmig algorithms correspod to (mi, +) circuits over a appropriate semirig. Notable examples iclude the Bellma-Ford-Moore (BFM) algorithm for the sigle-source-shortest-path problem (SSSP) [2, 5, 14], the Floyd-Warshall (FW) algorithm for the All-Pairs-Shortest-Path (APSP) problem [4, 18], ad the Held-Karp (HK) algorithm for the Travellig Salesma Problem (TSP) [6]. All these algorithms are just recursively costructed (mi, +) circuits. For example, both the BFM ad the FW algorithms give O( 3 ) sized (mi, +) circuits while the HK algorithm gives a O( 2 2 ) sized (mi, +) circuit. Matchig lower bouds were proved for TSP i [7], for APSP i [8], ad for SSSP i [10]. So, provig tight lower bouds for circuits over (mi, +) ca help us uderstad the power ad limitatios of dyamic programmig. We refer the reader to [8, 9] for more results o (mi, +) circuit lower bouds. Note that algorithms for problems like computig the diameter of a graph are aturally expressed usig (mi, max, +) circuits. This makes the cost of covertig a max gate to a (mi, +) circuit or formula a iterestig measure. A related questio arises i the settig of coutig classes defied by arithmetic circuits ad formulas. Circuits over N, with specific resource bouds, cout acceptig computatio paths or proof-trees i a related resource-bouded Turig machie model defiig a class C. The coutig fuctio class is deoted #C. The differece of two such fuctios i a class #C is a fuctio i the class DiffC. O the other had, circuits with the same resource bouds, but over Z, or equivaletly, with subtractio gates, describe the fuctio class GapC. For most complexity classes C, a straightforward argumet shows that that DiffC ad GapC coicide. See [1] for further discussio o this. I this framework, we restrict attetio to computatio over N ad see that as a member of a Gap class over (mi, +), max has liear-size formulas, whereas as a member of a Diff class, it requires Ω( log ) size. Our results ad techiques: We ow formally state our results ad briefly commet o the techiques used to prove them. 1. For 2, o (mi, +) formula over N ca compute max(x 1, x 2,..., x ). (Theorem 10) The proof is simple: apply a carefully chose restrictio to the variables ad show that the (mi, +) formula does ot output the correct value of max o this restrictio. 2. max(x 1, x 2,..., x ) ca be computed by a differece of two (mi, +) formulas with total size + log. More geerally, the fuctio computig the sum of the topmost k values amogst the variables ca be computed by a differece of two (mi, +) formulas with total size + ( log ) mi{k, k}. (Theorem 11) Note that the sum of the topmost k values ca be computed by the followig formula: (Sum of all variables) - mi S (Sum of all variables except those i S). Here S rages over all possible subsets of {x 1, x 2,..., x } of cardiality k. Usig recursio, we obtai the claimed size boud. 3. Let F 1, F 2 be (mi, +) formulas over N such that F 1 F 2 = max(x 1, x 2,..., x ). The F 1 must have at least leaves ad F 2 at least log leaves. (Theorem 13)

3 M. Mahaja ad P. Nimbhorkar ad A. Tawari 74:3 A major igrediet i our proof is the defiitio of a measure for fuctios computable by costat-free (mi, +) formulas, ad relatig this measure to formula size. The measure ivolves terms aalogous to miterms of a mootoe Boolea fuctio, ad uses the etropy of a associated graph uder the uiform distributio o its vertices. I the settig of mootoe Boolea fuctios, this techique was used i i [15] to give formula size lower bouds. We adapt that techique to the (mi, +) settig. The same techique also yields the followig lower boud: Also, ay (mi, +) formula computig the miimum over the sums of 1 variables must have at least log leaves. This is tight. (Lemma 12 ad Corollary 18) 2 Prelimiaries 2.1 Notatio Let X deote the set of variables {x 1,..., x }. We use x to deote (x 1, x 2,..., x, 1). We use e i to deote the ( + 1)-dimesioal vector with a 1 i the ith coordiate ad zeroes elsewhere. For i [], we also use e i to deote a assigmet to the variables x 1, x 2,..., x where x i is set to 1 ad all other variables are set to 0. Defiitio 1. For 0 r, the -variate fuctios Sum, MiSum r ad MaxSum r are as defied below. Sum = i=1 x i { } MiSum r = mi x i S, S = r i S { } MaxSum r = max x i S, S = r i S Note that MiSum 0 ad MaxSum 0 are the costat fuctio 0, ad MiSum 1 ad MaxSum 1 are just the mi ad max respectively. Observatio 2. For 1 r <, MiSum = MaxSum = Sum = MiSum r +MaxSum r. 2.2 Formulas A (mi, +) formula is a directed tree. Each leaf of a formula has a label from X N; that is, it is labeled by a variable x i or a costat α N. Each iteral ode has exactly two childre ad is labeled by oe of the two operatios mi or +. The output ode of the formula computes a fuctio of the iput variables i the atural way. The iput odes of a formula are also referred to as gates. If all leaves of a formula are labeled from X, we say that the formula is costat-free. A (mi, +, ) formula is similarly defied; the operatio at a iteral ode may also be, i which case the childre are ordered ad the ode computes the differece of their values. We defie the size of a formula as the umber of leaves i the formula. For a formula F, we deote by L(F ) its size, the umber of leaves i it. For a fuctio f, we deote by L(f) the smallest size of a formula computig f. By L cf (f) we deote the smallest size of a costat-free formula computig f. M FC S

4 74:4 Computig max usig (mi,+) formulas 2.3 Graph Etropy The otio of the etropy of a graph or hypergraph, with respect to a probability distributio o its vertices, was first defied by Körer i [11]. I that ad subsequet works (e.g. [12, 13, 3, 15]), equivalet characterizatios of graph etropy were established ad are ofte used ow as the defiitio itself, see for istace [16, 17]. I this paper, we use graph etropy oly with respect to the uiform distributio, ad simply call it graph etropy. We use the followig defiitio, which is exactly the defiitio from [17] specialised to the uiform distributio. Defiitio 3. Let G be a graph with vertex set V (G) = {1,..., }. The vertex packig polytope V P (G) of the graph G is the covex hull of the characteristic vectors of idepedet sets of G. The etropy of G is defied as H(G) = mi a V P (G) i=1 1 log 1 a i. It ca easily be see that H(G) is a o-egative real umber, ad moreover, H(G) = 0 if ad oly if G has o edges. We list o-trivial properties of graph etropy that we use. Lemma 4 ([12, 13]). Let F = (V, E(F )) ad G = (V, E(G)) be two graphs o the same vertex set. The followig hold: 1. Mootoocity. If E(F ) E(G), the H(F ) H(G) 2. Subadditivity. Let Q be the graph with vertex set V ad edge set E(F ) E(G). The H(Q) H(F ) + H(G). Lemma 5 (see for istace [16, 17]). The followig hold: 1. Let K be the complete graph o vertices. The H(K ) = log. 2. Let G be a graph o vertices, whose edges iduce a bipartite graph o m (out of ) vertices. The H(G) m. 3 Trasformatios ad Easy bouds We cosider the computatio of max{x 1,..., x } over N usig (mi, +) formulas. To start with, we describe some properties of (mi, +) formulas that we use repeatedly. The first property, Propositio 7 below, is expressig the fuctio computed by a formula as a depth-2 polyomial where + plays the role of multiplicatio ad mi plays the role of additio. The ext properties, Propositio 8 ad 9 below, deal with removig redudat sub-expressios created by the costat zero or movig commo parts aside. Defiitio 6. Let F be a (mi, +) formula with leaves labeled from X N. For each gate v F, we costruct a set S v N +1 as described below. We costruct the sets iductively based o the depth of v. 1. Case 1. v is a leaf labeled α for some α N. The S v = {α e +1 }. (Recall, e i is the uit vector with 1 at the ith coordiate ad zero elsewhere). 2. Case 2: v is a leaf labeled x i for some i []. The S v = {e i }. 3. Case 3: v = mi{u, w}. The S v = S u S w. 4. Case 4: v = u + w. The S v = {ã + b ã S u, b S w } (coordiate-wise additio). Let r be the output gate of F. We deote by S(F ) the set S r so costructed.

5 M. Mahaja ad P. Nimbhorkar ad A. Tawari 74:5 It is straightforward to see that if F has o costats (so Case 1 is ever ivoked), the a +1 remais 0 throughout the costructio of the sets S v. Hece if F is costat-free, the for each ã S(F ), a +1 = 0. By costructio, the set S(F ) describes the fuctio computed by F. Thus we have the followig: Propositio 7. Let F be a formula with mi ad + gates, with leaves labeled by elemets of {x 1,..., x } N. For each gate v F, let f v deote the fuctio computed at v. The f v = mi{ ã x ã S v }. The followig propositio is a easy cosequece of the costructio i Defiitio 6. Propositio 8. Let F be a (mi, +) formula over N. Let G be the formula obtaied from F by replacig all costats by the costat 0. Let H be the costat-free formula obtaied from G by elimiatig 0s from G through repeated replacemets of 0 + A by A, mi{0, A} by 0. The 1. L(H) L(G) = L(F ), 2. S(G) = { b b +1 = 0, ã S(F ), i [], a i = b i }, ad 3. G ad H compute the same fuctio mi{ b x b S(G)}. (Note: It is ot the claim that S(G) = S(H). Ideed, this may ot be the case. eg. let F = x + mi{1, x + y}. The S(F ) = {101, 210}, S(G) = {100, 210}, S(H) = {100}, However, the fuctios computed are the same.) The ext propositio shows how to remove commo cotributors to S(F ) without icreasig the formula size. Propositio 9. Let F be a (mi, +) formula computig a fuctio f. If, for some i [], a i > 0 for every ã S(F ), the f x i ca be computed by a (mi, +) formula F of size at most size(f ). If a +1 > 0 for every ã S(F ), the f 1 ca be computed by a (mi, +) formula F of size at most size(f ). I both cases, S(F ) = { b ã S(F ), b = ã e i }. Proof. First cosider i []. Let X be the subset of odes i F defied as follows: X = {v F ã S v : a i > 0} Clearly, the output gate r of F belogs to X. By the costructio of the sets S v, wheever a mi ode v belogs to X, both its childre belog to X, ad wheever a + ode belogs to X, at least oe of its childre belogs to X. We pick a set T X as follows. Iclude r i T. For each mi ode i T, iclude both its childre i T. For each + ode i T, iclude i T oe child that belogs to X (if both childre are i X, choose ay oe arbitrarily). This gives a sub-formula of F where all leaves are labeled x i. Replace these occurreces of x i i F by 0 to get formula F. It is easy to see that S(F ) = {ã e i ã S}. Hece F computes f x i. For i = a +1, the same process as above yields a subformula where each leaf is labeled by a positive costat. Subtractig 1 from the costat at each leaf i T gives the formula computig f 1. It is ituitively clear that o (mi, +) formula ca compute max. A formal proof usig Propositio 7 appears below. Theorem 10. For 2, o (mi, +) formula over N ca compute max{x 1,..., x }. M FC S

6 74:6 Computig max usig (mi,+) formulas Proof. Suppose, to the cotrary, some formula C computes max. The its restrictio D to x 1 = X, x 2 = Y, x 3 = x 4 =... = x = 0, correctly computes max{x, Y }. Sice all leaves of D are labeled from {x 1, x 2 } N, the set S(D) is a set of triples. Let S N 3 be this set. For all X, Y N, max{x, Y } equals E(X, Y ) = mi{ax + BY + C (A, B, C) S}. Let K deote the maximum value take by C i ay triple i S. If for some B, C N, the triple (0, B, C) belogs to S, the E(K + 1, 0) C K < K + 1 = max{0, K + 1}. So for all (A, B, C) S, A 0, so A 1. Similarly, for all (A, B, C) S, B 1. Hece for all (A, B, C) S, A + B 2. Now E(1, 1) = mi{a + B + C (A, B, C) S} 2 > 1 = max{1, 1}. So E(X, Y ) does ot compute max(x, Y ) correctly. However, if we also allow the subtractio operatio at iteral odes, it is very easy to compute the maximum i liear size; max(x 1,..., x ) = mi{ x 1, x 2,..., x }. Here a is implemeted as 0 a, ad if we allow oly variables, ot costats, at leaves, we ca compute a as (x 1 x 1 ) a. Thus the subtractio operatio adds sigificat power. How much? Ca we compute the maximum with very few subtractio gates? It turs out that the max fuctio ca be computed as the differece of two (mi, +) formulas. Equivaletly, there is a (mi, +, ) formula with a sigle gate at the root, that computes the max fuctio. This formula is ot liear i size, but it is ot too big either; we show that it has size O( log ). A simple geeralisatio allows us to compute the sum of the largest k values. Theorem 11. For each 1, ad each 0 k, the fuctio MaxSum k ca be computed by a differece of two (mi, +) formulas with total size + ( log ) mi{k, k}. I particular, the fuctio max{x 1,..., x } ca be computed by a differece of two (mi, +) formulas with total size + log. Proof. Note that MaxSum k = Sum MiSum k. Lemma 12 below shows that MiSum k ca be computed by a formula of size ( log ) mi{k, k} for 0 k. Sice Sum ca be computed by a formula of size, the claimed upper boud for MaxSum k follows. Lemma 12. For all, k such that 1 ad 0 k <, the fuctios MiSum k, MiSum k ca be computed by a (mi, +) formula of size ( log ) k. Hece the fuctios MiSum k, MiSum k ca be computed by (mi, +) formulas of size ( log ) mi{k, k}. Proof. We prove the upper boud for MiSum k. The boud for MiSum k follows from a essetially idetical argumet. We prove this by iductio o k. Base Case: k = 0. For every 1, MiSum k = Sum ad ca be computed with size. Iductive Hypothesis: For all k < k, ad all > k, MiSum k ca computed i size ( log ) k. Iductive Step: We wat to prove the claim for k, where k 1, ad for all > k. We proceed by iductio o. Base Case: = k + 1. MiSum k = MiSum 1 is the miimum of the variables, ad ca be computed i size. Iductive Hypothesis: For all k < m <, MiSum m k m ca be computed i size m( log m ) k.

7 M. Mahaja ad P. Nimbhorkar ad A. Tawari 74:7 Iductive Step: Let m = /2, m = /2, Let X, X l, X r deote the sets of variables {x 1,..., x }, {x 1,..., x m }, {x m +1,..., x }. Note that X l = m, X r = m, m + m =. Let p deote log. Note that log m = log m = p 1. To compute MiSum k o X, we first compute, for various values of t, MiSum m t m o X r, ad add them up. We the take the miimum of o X l, MiSum m (k t) m these sums. Note that if m = t or m = k t, the that summad is simply 0 ad we oly compute the other summad. Now MiSum k { mi MiSum m t m (X) ca be computed as } (X l) + MiSum m (k t) m (X r) max{0, k m } t mi{m, k} For all the sub-expressios appearig i the above costructio, we ca use iductively costructed formulas. Usig the iductive hypotheses (both for t < k ad for t = k, m < ), we see that the umber of leaves i the resultig formula is give by = mi{m,k} t=max{0,k m } [ m (p 1) t + m (p 1) k t] k [ m (p 1) t + m (p 1) k t] t=0 [ k ] [ k ] m (p 1) t + m (p 1) t t=0 t=0 [ k ] = (m + m ) (p 1) t t=0 [(p 1) + 1] k = p k I the rest of this paper, our goal is to prove a matchig lower boud for the max fuctio. Note that the costructios i Theorem 11 ad Lemma 12 yield formulas that do ot use costats at ay leaves. Ituitively, it is clear that if a formula computes the maximum correctly for all atural umbers, the costats caot help. So the lower boud should hold eve i the presece of costats, ad ideed our lower boud does hold eve if costats are allowed. 4 The mai lower boud I this sectio, we prove the followig theorem: Theorem 13. Let F 1, F 2 be (mi, +) formulas over N such that F 1 F 2 = max(x 1,..., x ). The L(F 1 ), ad L(F 2 ) log. The proof proceeds as follows: we first trasform F 1 ad F 2 over a series of steps to formulas G 1 ad G 2 o larger tha F 1 ad F 2, such that G 1 G 2 equals F 1 F 2 ad hece still computes max, ad G 1 ad G 2 have some ice properties. These properties immediately imply that L(F 1 ) L(G 1 ). We further trasform G 2 to a costat-free formula H o larger tha G 2. We the defie a measure for fuctios computable by costat-free (mi, +) formulas, relate this measure to formula size, ad use the properties of G 2 ad H to show that the fuctio h computed by H has large measure ad large formula size. Trasformatio 1: For b {1, 2}, let S b deote the set S(F b ). For i [ + 1], let A i be the miimum value appearig i the ith coordiate i ay tuple i S 1 S 2. Let à deote M FC S

8 74:8 Computig max usig (mi,+) formulas the tuple (A 1,..., A, A +1 ). By repeatedly ivokig Propositio 9, we obtai formulas G b computig F b à x, with L(G b) L(F b ). For b {1, 2}, let T b deote the set S(G b ). We ow establish the followig properties of G 1 ad G 2. Lemma 14. Let F 1, F 2 be (mi, +) formulas such that F 1 F 2 computes max. Let G 1, G 2 be obtaied as described above. The 1. L(G 1 ) L(F 1 ), L(G 2 ) L(F 2 ), 2. max(x) = F 1 F 2 = G 1 G 2, 3. For every i [], for every ã T 1, a i > 0. Hece L(G 1 ). 4. For every i [], there exists ã T 2, a i = There exist ã T 1, b T 2, a +1 = b +1 = For every i, j [] with i j, for every ã T 2, a i + a j > 0. Proof. 1. This follows from propositio Obvious. 3. Suppose for some ã T 1 ad for some i [], a i = 0. Cosider the iput assigmet d where d i = 1 + a +1 ad d j = 0 for j [] \ {i}. The max{d 1,..., d } = 1 + a +1. However, ã d = a +1. Therefore o iput d, G 1 ( d) a +1. Sice G 2 0 o all assigmets, we get G 1 ( d) G 2 ( d) a +1 < max( d), cotradictig the assumptio that G 1 G 2 computes max. 4. This follows from the previous poit ad the choice of A i for each i. 5. From the choice of A +1, we kow that there is a ã i T 1 T 2 with a +1 = 0. Suppose there is such a tuple i exactly oe of the sets T 1, T 2. The exactly oe of G 1 ( 0), G 2 ( 0) equals 0, ad so G 1 G 2 does ot compute max( 0). 6. Suppose to the cotrary, some ã T 2 has a i = a j = 0. Cosider the iput assigmet d where d i = d j = 1 + a +1 ad d k = 0 for k [] \ {i, j}. The max{d 1,..., d } = 1 + a +1. Sice every x k figures i every tuple of T 1, G 1 ( d) d i + d j = 2a But G 2 ( d) a +1. Hece G 1 ( d) G 2 ( d) does ot compute max( d). We have already show above that L(F 1 ) L(G 1 ). Now the more tricky part: we eed to lower boud L(G 2 ). Trasformatio 2: Let H be the formula obtaied by simply replacig every costat i G 2 by 0. Let H be the costat-free formula obtaied from H by elimiatig the zeroes, repeatedly replacig 0 + A by A, mi{0, A} by 0. Let h be the fuctio computed by H. The, L cf (h) L(H) L(H ) = L(G 2 ) L(F 2 ). It thus suffices to show that L cf (h) log. To this ed, we defie a complexity measure µ, relate it to costat-free formula size, ad show that it is large for the fuctio h. Defiitio 15. For a -variate fuctio f computable by a costat-free (mi, +) formula, we defie (f) 1 = {i f(e i ) 1, f(0) = 0}. (f) 2 = {(i, j) f(e i + e j ) 1, f(e i ) = 0, f(e j ) = 0}. We defie G(f) to be the graph whose vertex set is [] ad edge set is (f) 2. The measure µ for fuctio f is defied as follows: µ = (f) 1 + H(G(f))

9 M. Mahaja ad P. Nimbhorkar ad A. Tawari 74:9 The followig lemma relates µ(f) with L(f). This relatio has bee used before, see for istace [15] for applicatios to mootoe Boolea circuits. Sice we have ot see a applicatio i the settig of (mi, +) formulas, we (re-)prove this i detail here; however, it is really the same proof. Lemma 16. Let f be a -variate fuctio computable by a costat-free (mi, +) formula. The L cf (f) µ(f). Proof. The proof is by iductio o the depth of a witessig formula F that computes f ad has L cf (F ) = L cf (f). Base case: F is a iput variable, say x i. The (f) 1 = {x i }, ad G(f) is the empty graph, so µ(f) = 1. Hece 1 = L cf (f) = µ(f). Iductive step: F is either F +F or mi{f, F } for some formulas F, F computig fuctios f, f respectively. Sice F is a optimal-size formula for f, F ad F are optimalsize formulas for f ad f as well. So L cf (f) = L(F ) = L(F )+L(F ) = L cf (f )+L cf (f ). Case a. F = F + F. The (f) 1 = (f ) 1 (f ) 1 ad G(f) G(f ) G(f ). Hece, µ(f) (f ) 1 (f ) 1 + H(G(f ) G(f )) (Lemma 4) (f ) 1 + (f ) 1 + H(G(f )) + H(G(f )) (Lemma 4) = µ(f ) + µ(f ) 1 L cf (f ) + 1 L cf (f ) (Iductio) = 1 L cf (f) (L cf (f) = L cf (f ) + L cf (f )) Case b. F = mi(f, F ). Let (f ) 1 = A ad (f ) 1 = B. The (f) 1 = A B ad G(f) G(f ) G(f ) G(A \ B, B \ A). Here, G(P, Q) deotes the bipartite graph G with parts P ad Q. Hece, µ(f) 1 ( A B ) + H(G(f ) G(f ) G(A \ B, B \ A)) (Lemma 4) 1 ( A B ) + H(G(f )) + H(G(f )) + H(G(A \ B, B \ A)) (Lemma 4) 1 ( A B ) + H(G(f )) + H(G(f )) + 1 ( A \ B + B \ A ) (Lemma 5) 1 ( A + B ) + H(G(f )) + H(G(f )) = µ(f ) + µ(f ) 1 L cf (f ) + 1 L cf (f ) (Iductio) = 1 L cf (f) (L cf (f) = L cf (f ) + L cf (f )) Hece, µ(f) 1 L cf (f). Usig this measure, we ca ow show the required lower boud. Lemma 17. For the fuctio h obtaied after Trasformatio 2, µ(h) log. Proof. Recall that we replaced costats i G 2 by 0 to get H, the elimiated the 0s to get costat-free H computig h. By Propositio 8, we kow that S(H ) = { b b +1 = 0, ã T 2, a i = b i i []} ad that h = mi{ x b b S(H )}. M FC S

10 74:10 Computig max usig (mi,+) formulas From item 4 i Lemma 14, it follows that (h) 1 =. (For every i, there is a b S(H ) with b i = 0. So h(e i ) e i b = 0.) Sice (h) 1 is empty, (i, j) G(h) exactly whe h(e i +e j ) 1. From item 6 i Lemma 14, it follows that every pair (i, j) is i G(h). Thus G(h) is the complete graph K. From Lemma 5 we coclude that µ(h) = log. Lemmas 16 ad 17 imply that L cf (h) log. Sice L cf (h) L(H) L(H ) = L(G 2 ) L(F 2 ), we coclude that L(F 2 ) log. This completes the proof of Theorem 13. A major igrediet i this proof is usig the measure µ. This yields lower bouds for costat-free formulas. For fuctios computable i a costat-free maer, it is hard to see how costats ca help. However, to trasfer a lower boud o L cf (f) to a lower boud o L(f), this idea of costats caot help eeds to be formalized. The trasformatios described before we defie µ do precisely this. For the MiSum 1 fuctio, applyig the measure techique immediately yields the lower boud L cf (MiSum 1 ) log. Trasferrig this lower boud to formulas with costats is a corollary of our mai result, ad with it we see that the upper boud from Lemma 12 is tight for MiSum 1. Corollary 18. Ay (mi, +) formula computig MiSum 1 must have size at least log. Proof. Let F be ay formula computig MiSum 1. Applyig Theorem 13 to F 1 = x x ad F 2 = F, we obtai L(F ) log. 5 Discussio Our results hold whe variables take values from N. I the stadard (mi, +) semi-rig, the value is also allowed, sice it serves as the idetity for the mi operatio. The proof of our mai result Theorem 13 does ot carry over to this settig. The mai stumblig block is the removal of the commo part of S(F ). However, if we allow as a value that a variable ca take, but ot as a costat appearig at a leaf, the the lower boud proof still seems to work. However, the upper boud o loger works; while takig a differece, what is? Apart from the may atural settigs where the tropical semirig (mi, +, N { }, 0, ) crops up, it is also iterestig because it ca simulate the Boolea semirig for mootoe computatio. The mappig is straightforward: 0, 1,, i the Boolea semirig are replaced by, 0, mi, + respectively i the tropical semirig. Provig lower bouds for (mi, +) formulas could be easier tha for mootoe Boolea formulas because the (mi, +) formula has to compute a fuctio correctly at all values, ot just at 0,. Hece it would be iterestig to exted our lower boud to this settig with as well. Our trasformatios crucially use the fact that there is a miimum elemet, 0. Thus, we do ot see how to exted these results to computatios over itegers. It appears that we will eed to iclude, ad sice we are curretly uable to hadle eve +, there is already a barrier. The lower boud method uses graph etropy which is always bouded above by log. Thus this method caot give a lower boud larger tha log. It would be iterestig to obtai a modified techique that ca show that all the upper bouds i Theorem 11 ad Lemma 12 are tight. It would also be iterestig to fid a direct combiatorial proof of our lower boud result, without usig graph etropy.

11 M. Mahaja ad P. Nimbhorkar ad A. Tawari 74:11 Refereces 1 Eric Alleder. Arithmetic circuits ad coutig complexity classes. I Ja Krajicek, editor, Complexity of Computatios ad Proofs, Quaderi di Matematica Vol. 13, pages Secoda Uiversita di Napoli, A earlier versio appeared i the Complexity Theory Colum, SIGACT News 28, 4 (Dec. 1997) pp Richard Bellma. O a routig problem. Quarterly of Applied Mathematics, 16:87 90, Imre Csiszár, Jáos Körer, László Lovász, Katali Marto, ad Gábor Simoyi. Etropy splittig for atiblockig corers ad perfect graphs. Combiatorica, 10(1):27 40, Robert W Floyd. Algorithm 97: shortest path. Commuicatios of the ACM, 5(6):345, Lester R Ford Jr. Network flow theory. Techical Report P-923, Rad Corporatio, Michael Held ad Richard M Karp. A dyamic programmig approach to sequecig problems. Joural of the Society for Idustrial ad Applied Mathematics, 10(1): , Mark Jerrum ad Marc Sir. Some exact complexity results for straight-lie computatios over semirigs. Joural of the ACM (JACM), 29(3): , Stasys Juka. Lower bouds for tropical circuits ad dyamic programs. Theory of Computig Systems, 57(1): , Stasys Juka. Tropical complexity, Sido sets, ad dyamic programmig. SIAM Joural o Discrete Mathematics, 30(4): , Stasys Juka ad Georg Schitger. O the optimality of Bellma Ford Moore shortest path algorithm. Theoretical Computer Sciece, 628: , Jáos Körer. Codig of a iformatio source havig ambiguous alphabet ad the etropy of graphs. I Trasactios of 6th Prague Coferece o Iformatio Theory, pages Academia, Prague, Jáos Körer. Fredma-Komlós bouds ad iformatio theory. SIAM. J. o Algebraic ad Discrete Methods, 7(4): , Jáos Körer ad Katali Marto. New bouds for perfect hashig via iformatio theory. Europea Joural of Combiatorics, 9(6): , Edward F Moore. The shortest path through a maze. Bell Telephoe System., Ila Newma ad Avi Wigderso. Lower bouds o formula size of boolea fuctios usig hypergraph etropy. SIAM Joural o Discrete Mathematics, 8(4): , Gábor Simoyi. Graph etropy: A survey. Combiatorial Optimizatio, 20: , Gábor Simoyi. Perfect graphs ad graph etropy: A updated survey. I Perfect Graphs, pages Joh Wiley ad Sos, Stephe Warshall. A theorem o Boolea matrices. Joural of the ACM (JACM), 9(1):11 12, M FC S

Computing the Maximum Using (min,+) Formulas

Computing the Maximum Using (min,+) Formulas Computing the Maximum Using (min,+) Formulas Meena Mahajan 1, Prajakta Nimbhorkar 2, and Anuj Tawari 3 1 The Institute of Mathematical Sciences, HBNI, Chennai, India meena@imsc.res.in 2 Chennai Mathematical

More information

Lecture 16: Monotone Formula Lower Bounds via Graph Entropy. 2 Monotone Formula Lower Bounds via Graph Entropy

Lecture 16: Monotone Formula Lower Bounds via Graph Entropy. 2 Monotone Formula Lower Bounds via Graph Entropy 15-859: Iformatio Theory ad Applicatios i TCS CMU: Sprig 2013 Lecture 16: Mootoe Formula Lower Bouds via Graph Etropy March 26, 2013 Lecturer: Mahdi Cheraghchi Scribe: Shashak Sigh 1 Recap Graph Etropy:

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

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

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

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

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

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

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

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

LONG SNAKES IN POWERS OF THE COMPLETE GRAPH WITH AN ODD NUMBER OF VERTICES

LONG SNAKES IN POWERS OF THE COMPLETE GRAPH WITH AN ODD NUMBER OF VERTICES J Lodo Math Soc (2 50, (1994, 465 476 LONG SNAKES IN POWERS OF THE COMPLETE GRAPH WITH AN ODD NUMBER OF VERTICES Jerzy Wojciechowski Abstract I [5] Abbott ad Katchalski ask if there exists a costat c >

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

CHAPTER I: Vector Spaces

CHAPTER I: Vector Spaces CHAPTER I: Vector Spaces Sectio 1: Itroductio ad Examples This first chapter is largely a review of topics you probably saw i your liear algebra course. So why cover it? (1) Not everyoe remembers everythig

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

Sequences, Mathematical Induction, and Recursion. CSE 2353 Discrete Computational Structures Spring 2018

Sequences, Mathematical Induction, and Recursion. CSE 2353 Discrete Computational Structures Spring 2018 CSE 353 Discrete Computatioal Structures Sprig 08 Sequeces, Mathematical Iductio, ad Recursio (Chapter 5, Epp) Note: some course slides adopted from publisher-provided material Overview May mathematical

More information

4.3 Growth Rates of Solutions to Recurrences

4.3 Growth Rates of Solutions to Recurrences 4.3. GROWTH RATES OF SOLUTIONS TO RECURRENCES 81 4.3 Growth Rates of Solutios to Recurreces 4.3.1 Divide ad Coquer Algorithms Oe of the most basic ad powerful algorithmic techiques is divide ad coquer.

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

Some special clique problems

Some special clique problems Some special clique problems Reate Witer Istitut für Iformatik Marti-Luther-Uiversität Halle-Witteberg Vo-Seckedorff-Platz, D 0620 Halle Saale Germay Abstract: We cosider graphs with cliques of size k

More information

On Some Properties of Digital Roots

On Some Properties of Digital Roots Advaces i Pure Mathematics, 04, 4, 95-30 Published Olie Jue 04 i SciRes. http://www.scirp.org/joural/apm http://dx.doi.org/0.436/apm.04.46039 O Some Properties of Digital Roots Ilha M. Izmirli Departmet

More information

Lecture 27. Capacity of additive Gaussian noise channel and the sphere packing bound

Lecture 27. Capacity of additive Gaussian noise channel and the sphere packing bound Lecture 7 Ageda for the lecture Gaussia chael with average power costraits Capacity of additive Gaussia oise chael ad the sphere packig boud 7. Additive Gaussia oise chael Up to this poit, we have bee

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

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

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

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

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

Lecture Overview. 2 Permutations and Combinations. n(n 1) (n (k 1)) = n(n 1) (n k + 1) =

Lecture Overview. 2 Permutations and Combinations. n(n 1) (n (k 1)) = n(n 1) (n k + 1) = COMPSCI 230: Discrete Mathematics for Computer Sciece April 8, 2019 Lecturer: Debmalya Paigrahi Lecture 22 Scribe: Kevi Su 1 Overview I this lecture, we begi studyig the fudametals of coutig discrete objects.

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

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

THE ASYMPTOTIC COMPLEXITY OF MATRIX REDUCTION OVER FINITE FIELDS

THE ASYMPTOTIC COMPLEXITY OF MATRIX REDUCTION OVER FINITE FIELDS THE ASYMPTOTIC COMPLEXITY OF MATRIX REDUCTION OVER FINITE FIELDS DEMETRES CHRISTOFIDES Abstract. Cosider a ivertible matrix over some field. The Gauss-Jorda elimiatio reduces this matrix to the idetity

More information

Notes on the Combinatorial Nullstellensatz

Notes on the Combinatorial Nullstellensatz Notes o the Combiatorial Nullstellesatz Costructive ad Nocostructive Methods i Combiatorics ad TCS U. Chicago, Sprig 2018 Istructor: Adrew Drucker Scribe: Roberto Ferádez For the followig theorems ad examples

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

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

(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

Sequences I. Chapter Introduction

Sequences I. Chapter Introduction Chapter 2 Sequeces I 2. Itroductio A sequece is a list of umbers i a defiite order so that we kow which umber is i the first place, which umber is i the secod place ad, for ay atural umber, we kow which

More information

On matchings in hypergraphs

On matchings in hypergraphs O matchigs i hypergraphs Peter Frakl Tokyo, Japa peter.frakl@gmail.com Tomasz Luczak Adam Mickiewicz Uiversity Faculty of Mathematics ad CS Pozań, Polad ad Emory Uiversity Departmet of Mathematics ad CS

More information

The Random Walk For Dummies

The Random Walk For Dummies The Radom Walk For Dummies Richard A Mote Abstract We look at the priciples goverig the oe-dimesioal discrete radom walk First we review five basic cocepts of probability theory The we cosider the Beroulli

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

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

# fixed points of g. Tree to string. Repeatedly select the leaf with the smallest label, write down the label of its neighbour and remove the leaf.

# fixed points of g. Tree to string. Repeatedly select the leaf with the smallest label, write down the label of its neighbour and remove the leaf. Combiatorics Graph Theory Coutig labelled ad ulabelled graphs There are 2 ( 2) labelled graphs of order. The ulabelled graphs of order correspod to orbits of the actio of S o the set of labelled graphs.

More information

Some remarks for codes and lattices over imaginary quadratic

Some remarks for codes and lattices over imaginary quadratic Some remarks for codes ad lattices over imagiary quadratic fields Toy Shaska Oaklad Uiversity, Rochester, MI, USA. Caleb Shor Wester New Eglad Uiversity, Sprigfield, MA, USA. shaska@oaklad.edu Abstract

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

An analog of the arithmetic triangle obtained by replacing the products by the least common multiples

An analog of the arithmetic triangle obtained by replacing the products by the least common multiples arxiv:10021383v2 [mathnt] 9 Feb 2010 A aalog of the arithmetic triagle obtaied by replacig the products by the least commo multiples Bair FARHI bairfarhi@gmailcom MSC: 11A05 Keywords: Al-Karaji s triagle;

More information

CSE 1400 Applied Discrete Mathematics Number Theory and Proofs

CSE 1400 Applied Discrete Mathematics Number Theory and Proofs CSE 1400 Applied Discrete Mathematics Number Theory ad Proofs Departmet of Computer Scieces College of Egieerig Florida Tech Sprig 01 Problems for Number Theory Backgroud Number theory is the brach of

More information

FUNDAMENTALS OF REAL ANALYSIS by. V.1. Product measures

FUNDAMENTALS OF REAL ANALYSIS by. V.1. Product measures FUNDAMENTALS OF REAL ANALSIS by Doğa Çömez V. PRODUCT MEASURE SPACES V.1. Product measures Let (, A, µ) ad (, B, ν) be two measure spaces. I this sectio we will costruct a product measure µ ν o that coicides

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

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

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

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

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

FLOOR AND ROOF FUNCTION ANALOGS OF THE BELL NUMBERS. H. W. Gould Department of Mathematics, West Virginia University, Morgantown, WV 26506, USA

FLOOR AND ROOF FUNCTION ANALOGS OF THE BELL NUMBERS. H. W. Gould Department of Mathematics, West Virginia University, Morgantown, WV 26506, USA INTEGERS: ELECTRONIC JOURNAL OF COMBINATORIAL NUMBER THEORY 7 (2007), #A58 FLOOR AND ROOF FUNCTION ANALOGS OF THE BELL NUMBERS H. W. Gould Departmet of Mathematics, West Virgiia Uiversity, Morgatow, WV

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

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

Recurrence Relations

Recurrence Relations Recurrece Relatios Aalysis of recursive algorithms, such as: it factorial (it ) { if (==0) retur ; else retur ( * factorial(-)); } Let t be the umber of multiplicatios eeded to calculate factorial(). The

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

Section 4.3. Boolean functions

Section 4.3. Boolean functions Sectio 4.3. Boolea fuctios Let us take aother look at the simplest o-trivial Boolea algebra, ({0}), the power-set algebra based o a oe-elemet set, chose here as {0}. This has two elemets, the empty set,

More information

Mathematics review for CSCI 303 Spring Department of Computer Science College of William & Mary Robert Michael Lewis

Mathematics review for CSCI 303 Spring Department of Computer Science College of William & Mary Robert Michael Lewis Mathematics review for CSCI 303 Sprig 019 Departmet of Computer Sciece College of William & Mary Robert Michael Lewis Copyright 018 019 Robert Michael Lewis Versio geerated: 13 : 00 Jauary 17, 019 Cotets

More information

Square-Congruence Modulo n

Square-Congruence Modulo n Square-Cogruece Modulo Abstract This paper is a ivestigatio of a equivalece relatio o the itegers that was itroduced as a exercise i our Discrete Math class. Part I - Itro Defiitio Two itegers are Square-Cogruet

More information

Determinants of order 2 and 3 were defined in Chapter 2 by the formulae (5.1)

Determinants of order 2 and 3 were defined in Chapter 2 by the formulae (5.1) 5. Determiats 5.. Itroductio 5.2. Motivatio for the Choice of Axioms for a Determiat Fuctios 5.3. A Set of Axioms for a Determiat Fuctio 5.4. The Determiat of a Diagoal Matrix 5.5. The Determiat of a Upper

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

Resolution Proofs of Generalized Pigeonhole Principles

Resolution Proofs of Generalized Pigeonhole Principles Resolutio Proofs of Geeralized Pigeohole Priciples Samuel R. Buss Departmet of Mathematics Uiversity of Califoria, Berkeley Győrgy Turá Departmet of Mathematics, Statistics, ad Computer Sciece Uiversity

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

γ-max Labelings of Graphs

γ-max Labelings of Graphs γ-max Labeligs of Graphs Supapor Saduakdee 1 & Varaoot Khemmai 1 Departmet of Mathematics, Sriakhariwirot Uiversity, Bagkok, Thailad Joural of Mathematics Research; Vol. 9, No. 1; February 017 ISSN 1916-9795

More information

1. By using truth tables prove that, for all statements P and Q, the statement

1. By using truth tables prove that, for all statements P and Q, the statement Author: Satiago Salazar Problems I: Mathematical Statemets ad Proofs. By usig truth tables prove that, for all statemets P ad Q, the statemet P Q ad its cotrapositive ot Q (ot P) are equivalet. I example.2.3

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

Math 2784 (or 2794W) University of Connecticut

Math 2784 (or 2794W) University of Connecticut ORDERS OF GROWTH PAT SMITH Math 2784 (or 2794W) Uiversity of Coecticut Date: Mar. 2, 22. ORDERS OF GROWTH. Itroductio Gaiig a ituitive feel for the relative growth of fuctios is importat if you really

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 (, 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

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

DIVISIBILITY PROPERTIES OF GENERALIZED FIBONACCI POLYNOMIALS

DIVISIBILITY PROPERTIES OF GENERALIZED FIBONACCI POLYNOMIALS DIVISIBILITY PROPERTIES OF GENERALIZED FIBONACCI POLYNOMIALS VERNER E. HOGGATT, JR. Sa Jose State Uiversity, Sa Jose, Califoria 95192 ad CALVIN T. LONG Washigto State Uiversity, Pullma, Washigto 99163

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

Chapter IV Integration Theory

Chapter IV Integration Theory Chapter IV Itegratio Theory Lectures 32-33 1. Costructio of the itegral I this sectio we costruct the abstract itegral. As a matter of termiology, we defie a measure space as beig a triple (, A, µ), where

More information

Test One (Answer Key)

Test One (Answer Key) CS395/Ma395 (Sprig 2005) Test Oe Name: Page 1 Test Oe (Aswer Key) CS395/Ma395: Aalysis of Algorithms This is a closed book, closed otes, 70 miute examiatio. It is worth 100 poits. There are twelve (12)

More information

ON THE LEHMER CONSTANT OF FINITE CYCLIC GROUPS

ON THE LEHMER CONSTANT OF FINITE CYCLIC GROUPS ON THE LEHMER CONSTANT OF FINITE CYCLIC GROUPS NORBERT KAIBLINGER Abstract. Results of Lid o Lehmer s problem iclude the value of the Lehmer costat of the fiite cyclic group Z/Z, for 5 ad all odd. By complemetary

More information

SOME TRIBONACCI IDENTITIES

SOME TRIBONACCI IDENTITIES Mathematics Today Vol.7(Dec-011) 1-9 ISSN 0976-38 Abstract: SOME TRIBONACCI IDENTITIES Shah Devbhadra V. Sir P.T.Sarvajaik College of Sciece, Athwalies, Surat 395001. e-mail : drdvshah@yahoo.com The sequece

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

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

Math 155 (Lecture 3)

Math 155 (Lecture 3) Math 55 (Lecture 3) September 8, I this lecture, we ll cosider the aswer to oe of the most basic coutig problems i combiatorics Questio How may ways are there to choose a -elemet subset of the set {,,,

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

Lecture 19. sup y 1,..., yn B d n

Lecture 19. sup y 1,..., yn B d n STAT 06A: Polyomials of adom Variables Lecture date: Nov Lecture 19 Grothedieck s Iequality Scribe: Be Hough The scribes are based o a guest lecture by ya O Doell. I this lecture we prove Grothedieck s

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

On the Optimality of Bellman Ford Shortest Path Algorithm

On the Optimality of Bellman Ford Shortest Path Algorithm Electroic Colloquium o Computatioal Complexity, Revisio 1 of Report No. 127 (2015) O the Optimality of Bellma Ford Shortest Path Algorithm Stasys Juka a,1, Georg Schitger a a Istitute of Computer Sciece,

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

(VII.A) Review of Orthogonality

(VII.A) Review of Orthogonality VII.A Review of Orthogoality At the begiig of our study of liear trasformatios i we briefly discussed projectios, rotatios ad projectios. I III.A, projectios were treated i the abstract ad without regard

More information

(b) What is the probability that a particle reaches the upper boundary n before the lower boundary m?

(b) What is the probability that a particle reaches the upper boundary n before the lower boundary m? MATH 529 The Boudary Problem The drukard s walk (or boudary problem) is oe of the most famous problems i the theory of radom walks. Oe versio of the problem is described as follows: Suppose a particle

More information

2 High-level Complexity vs. Concrete Complexity

2 High-level Complexity vs. Concrete Complexity COMS 6998: Advaced Complexity Sprig 2017 Lecture 1: Course Itroductio ad Boolea Formulas Lecturer: Rocco Servedio Scribes: Jiahui Liu, Kailash Karthik Meiyappa 1 Overview of Topics 1. Boolea formulas (examples,

More information

Random Walks on Discrete and Continuous Circles. by Jeffrey S. Rosenthal School of Mathematics, University of Minnesota, Minneapolis, MN, U.S.A.

Random Walks on Discrete and Continuous Circles. by Jeffrey S. Rosenthal School of Mathematics, University of Minnesota, Minneapolis, MN, U.S.A. Radom Walks o Discrete ad Cotiuous Circles by Jeffrey S. Rosethal School of Mathematics, Uiversity of Miesota, Mieapolis, MN, U.S.A. 55455 (Appeared i Joural of Applied Probability 30 (1993), 780 789.)

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

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

LECTURE NOTES, 11/10/04

LECTURE NOTES, 11/10/04 18.700 LECTURE NOTES, 11/10/04 Cotets 1. Direct sum decompositios 1 2. Geeralized eigespaces 3 3. The Chiese remaider theorem 5 4. Liear idepedece of geeralized eigespaces 8 1. Direct sum decompositios

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

Statistical Machine Learning II Spring 2017, Learning Theory, Lecture 7

Statistical Machine Learning II Spring 2017, Learning Theory, Lecture 7 Statistical Machie Learig II Sprig 2017, Learig Theory, Lecture 7 1 Itroductio Jea Hoorio jhoorio@purdue.edu So far we have see some techiques for provig geeralizatio for coutably fiite hypothesis classes

More information

Lecture 12: November 13, 2018

Lecture 12: November 13, 2018 Mathematical Toolkit Autum 2018 Lecturer: Madhur Tulsiai Lecture 12: November 13, 2018 1 Radomized polyomial idetity testig We will use our kowledge of coditioal probability to prove the followig lemma,

More information

Problem Set 4 Due Oct, 12

Problem Set 4 Due Oct, 12 EE226: Radom Processes i Systems Lecturer: Jea C. Walrad Problem Set 4 Due Oct, 12 Fall 06 GSI: Assae Gueye This problem set essetially reviews detectio theory ad hypothesis testig ad some basic otios

More information

MAT1026 Calculus II Basic Convergence Tests for Series

MAT1026 Calculus II Basic Convergence Tests for Series MAT026 Calculus II Basic Covergece Tests for Series Egi MERMUT 202.03.08 Dokuz Eylül Uiversity Faculty of Sciece Departmet of Mathematics İzmir/TURKEY Cotets Mootoe Covergece Theorem 2 2 Series of Real

More information

Chapter 2. Periodic points of toral. automorphisms. 2.1 General introduction

Chapter 2. Periodic points of toral. automorphisms. 2.1 General introduction Chapter 2 Periodic poits of toral automorphisms 2.1 Geeral itroductio The automorphisms of the two-dimesioal torus are rich mathematical objects possessig iterestig geometric, algebraic, topological ad

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

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

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

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

Dominating Sets and Domination Polynomials of Square Of Cycles

Dominating Sets and Domination Polynomials of Square Of Cycles IOSR Joural of Mathematics IOSR-JM) ISSN: 78-78. Volume 3, Issue 4 Sep-Oct. 01), PP 04-14 www.iosrjourals.org Domiatig Sets ad Domiatio Polyomials of Square Of Cycles A. Vijaya 1, K. Lal Gipso 1 Assistat

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