Limits of Local Algorithms over Sparse Random Graphs

Size: px
Start display at page:

Download "Limits of Local Algorithms over Sparse Random Graphs"

Transcription

1 Limits of Local Algorithms over Sparse Ranom Graphs [Extene Abstract] ABSTRACT Davi Gamarnik MIT Local algorithms on graphs are algorithms that run in parallel on the noes of a graph to compute some global structural feature of the graph. Such algorithms use only local information available at noes to etermine local aspects of the global structure, while also potentially using some ranomness. Research over the years has shown that such algorithms can be surprisingly powerful in terms of computing structures like large inepenent sets in graphs locally. These algorithms have also been implicitly consiere in the work on graph limits, where a conjecture ue to Hatami, Lovász an Szegey [17] implie that local algorithms may be able to compute near-maximum inepenent sets in sparse) ranom -regular graphs. In this paper we refute this conjecture an show that every inepenent set prouce by local algorithms is smaller that the largest one by a multiplicative factor of at least 1/ + 1/ ).853, asymptotically as. Our result is base on an important clustering phenomena preicte first in the literature on spin glasses, an recently prove rigorously for a variety of constraint satisfaction problems on ranom graphs. Such properties suggest that the geometry of the solution space can be quite intricate. The specific clustering property, that we prove an apply in this paper shows that typically every two large inepenent sets in a ranom graph either have a significant intersection, or have a nearly empty intersection. As a result, large inepenent sets are clustere accoring to the proximity to each other. While the clustering property was postulate earlier as an obstruction for the success of local algorithms, such as for example, the Belief Propagation algorithm, our result is the first one where the clustering property is use to formally prove limits on local algorithms. A full version of this paper is available as Electronic Colloquium on Computational Complexity, Technical Report TR , February 0, 013. Research supporte by the NSF grants CMMI Permission to make igital or har copies of all or part of this work for personal or classroom use is grante without fee provie that copies are not mae or istribute for profit or commercial avantage an that copies bear this notice an the full citation on the first page. Copyrights for components of this work owne by others than ACM must be honore. Abstracting with creit is permitte. To copy otherwise, or republish, to post on servers or to reistribute to lists, requires prior specific permission an/or a fee. Request permissions from permissions@acm.org. ITCS 14, January 1 14, 014, Princeton, New Jersey, USA. Copyright 014 ACM /14/01...$ Mahu Suan Microsoft Research mahu@mit.eu Categories an Subject Descriptors F.1.3 [Complexity Measures an Classes]: Lower Bouns General Terms Theory Keywors Local algorithms, lower bouns, ranom graphs 1. INTRODUCTION Local algorithms are ecentralize algorithms that run in parallel on noes in a network using only information available from local neighborhoos to compute some global function of ata that is sprea over the network. Local algorithms have been stuie in the past in various communities. They arise as natural solution concepts in parallel an istribute computing see, e.g., [1, 19]). They also lea to efficient sub-linear algorithms algorithms that run in time significantly less than the length of the input an [6, 5, 16, 7] illustrate some of the progress in this irection. Finally local algorithms have also been propose as natural heuristics for solving har optimization problems with the popular Belief Propagation algorithm see for instance [9, 3]) being one such example. In this work we stuy the performance of a natural class of local algorithms on ranom regular graphs an show limits on the performance of these algorithms. 1.1 Motivation The motivation for our work comes from the a notion of local algorithms that has appeare in a completely ifferent mathematical context, namely that of the theory of graph limits, evelope in several papers, incluing [8], [7], [0], [6], [5], [1], [17]. In the realms of this theory it was conjecture that every reasonable combinatorial optimization problem on ranom graphs can be solve by means of some local algorithms. To the best of our knowlege this conjecture for the first time was formally state in Hatami, Lovász an Szegey in [17, Conjecture 7.13], an thus, from now on, we will refer to it as Hatami-Lovász-Szegey or HLS) conjecture, though informally it was pose by Szegey earlier, an was reference in several papers, incluing Lyons an Nazarov [], an Csoka an Lippner [11]. In the concrete context of the problem of fining largest inepenent sets in sparse ranom regular graphs, the conjecture is state as follows. Let T,r be a roote -regular tree with epth r.

2 Namely, every noe incluing the root, has egree r, except for the leaves, an the istance from the root to every leaf is r. Consier a function f r : [0, 1] T,r {0, 1} which maps every such tree whose noes are associate with real values from [0, 1] to a ecision encoe by 0 an 1, where the ecision is a function of the values associate with the noes of the tree. In light of the fact that in a ranom -regular graph G r) the typical noe has epth-r neighborhoo isomorphic to T,r, for any constant r, such a function f r can be use to generate ranom) subsets I of G r) as follows: associate with every noe of G r) a uniform ranom values from [0, 1] inepenently for each noe) an apply function f r to each noe. The set of noes for which f r prouces value 1 efines I, an is calle i.i.. factor. It is clear that f r essentially escribes a local algorithm for proucing sets I sweeping issue of computability of f r uner the rug). The HLS conjecture postulates the existence of a sequence of f r, r = 1,,..., such that the set I thus prouce is an inepenent subset of G r) an asymptotically achieves the largest possible value as r. Namely, largest inepenent subsets of ranom regular graphs are i.i.. factors. The precise connection between this conjecture an the theory of graph limits is beyon the scope of this paper. Instea we refer the reaer to the relevant papers [17],[1] 1. The concept of i.i.. factors appears also in one of the open problem by Davi Alous [] in the context of coing invariant processes on infinite trees. It turns out that an analogue for the HLS conjecture is inee vali for another important combinatorial optimization problem - the matching problem. Lyons an Nazarov [] establishe it for the case of bi-partite locally T,r -tree-like graphs, an Csoka an Lippner establishe this result for general locally T,r -tree-like graphs. Further, one can moify the framework of i.i.. factors by encapsulating non-t,r type neighborhoos, for example by making f r epen not only on the realization of ranom uniform in [0, 1] values, but also on the realization of the graph-theoretic neighborhoos aroun the noes. Some probabilistic boun on a egree might be neee to make this efinition rigorous though we will not attempt this formalization in this paper). In this case one can consier, for example, i.i.. factors when neighborhoos are istribute as r generations of a branching process with Poisson istribution, an then ask which combinatorial optimization problems efine now on sparse Erös-Rényi graphs Gn, /n) can be solve as i.i.. factors. Here Gn, /n) is a ranom graph on n noes with each of the n ) eges selecte with probability /n, inepenently for all eges, an > 0 is a fixe constant. In this case it is possible to show that when c e, the maximum inepenent set problem on Gn, /n) can be solve nearly optimally by the well known Belief Propagation BP) algorithm with constantly many rouns. Since the BP is a local algorithm, then the maximum inepenent set on Gn, /n) is an i.i.. factor, in the extene framework efine above. We shoul note that the original proof of Karp an Sipser [18] of the very similar result, relie on a ifferent metho.) Thus, the framework of local algorithms viewe as i.i.. factors is rich enough to solve several interesting combinatorial optimization problems. 1 We also remark that the language use in [17] is quite ifferent from ours. Other works, e.g., [] however o use langauge similar to ours in escribing this conjecture. Nevertheless, in this paper we refute the HLS conjecture in the context of maximum inepenent set problem on ranom regular graphs G n). Specifically, we show that for large enough, with high probability as n, every inepenent set proucible as an i.i.. factor is a multiplicative factor γ < 1 smaller than the largest inepenent subset of G n). We establish that γ is asymptotically at most though we conjecture that the result hols simply for γ = 1/, as we iscuss in the boy of the paper). 1. Techniques Our result is prove in two steps. First we show that large inepenent sets in ranom regular graphs are somewhat clustere. Next we show that solutions of natural algorithms are not clustere. We elaborate on the two parts below. Clustering is a powerful though fairly simple to establish in our case) phenomenon associate with the solutions of some combinatorial optimization problems on ranom graphs. Roughly it states that such solutions are clustere in some natural topology associate with the solution space. We will escribe the precise version we care about shortly.) Such phenomena were first conjecture in the theory of spin glasses an later confirme by rigorous means. Initially, this clustering property was iscusse in terms of so-calle overlap structure of the solutions of the Sherrington-Kirkpatrick moel [8]. Later works highlighte this phenomenon in the context of ranom K-SAT problem. In particular, in inepenent works Achlioptas, Coja-Oghlan an Ricci-Tersenghi [1], an Mezar, Mora an Zecchina [4] prove this phenomenon rigorously. We o not efine the ranom K-SAT problem here an instea refer the reaer to the aforementione papers. What these results state is that in certain regimes, the set of satisfying assignments, with high probability, can be clustere into groups such that two solutions within the same cluster agree on a certain minimum number of variables, while two solutions from ifferent clusters have to isagree on a certain minimum number of variables. In particular, one can ientify a certain non-empty interval [z 1, z ] [0, 1] such that no two solutions of the ranom K-SAT problem agree on precisely z fraction of variables for all z [z 1, z ]. One can further show that the onset of clustering property occurs when the ensity of clauses to variables becomes at least K /K, while at the same time the formula remains satisfiable with high probability, when the ensity is below K log. Interestingly, the known algorithms for fining solutions of ranom instances of K-SAT problem also stop working aroun the K /K threshol. It was wiely conjecture that the onset of the clustering phase is the main obstruction for fining such algorithms. In fact, Coja Oghlan [9] showe that the BP algorithm, which was earlier conjecture to be a goo contener for solving the ranom instances of K-SAT problems, also fails when the ensity of clauses to variables is at least K log K/K, though Coja-Oghlan s approach oes not explicitly rely on the clustering property, an one coul argue that the connection between the clustering property an the failure of the BP algorithm is coinciental. Closer to the topic of this paper, the clustering property was also recently establishe for inepenent sets in Erös- Rényi graphs by Coja-Oghlan an Efthymiou [10]. To escribe their result, we first note a well-known fact that the largest inepenent subset of Gn, /n) has size approximately log /)n, when is large see the next section for

3 precise etails). In [10] it is shown that the set of inepenent sets of size at least approximately log /)n namely those within factor 1/ of the optimal), are also clustere. Namely, one can split them into groups such that intersection of two inepenent sets within a group has a large carinality, while intersection of two inepenent sets from ifferent groups has a small carinality. One shoul note that algorithms for proucing large inepenent subsets of ranom graphs also stop short factor 1/ of the optimal, both in the case of sparse an in the ense ranom graph cases, as exhibite by the well-known Karp s open problem regaring inepenent subsets of Gn, 1/) [3]. This is almost the result we nee for our analysis with two exceptions. First, we nee to establish this clustering property for ranom regular as oppose Erös-Rényi graphs. Secon, an more signficanlty, the result in [10] applies to typical inepenent sets an oes not rule out the possibility that there exist two inepenent sets with some intermeiate intersection carinality, though the number of such pairs is insignificant compare to the total number of inepenent sets. For our result we nee to show that, without exception, every pair of large inepenent sets has either large or small intersection. We inee establish this, but at the cost of losing aitional factor 1/ ). In particular, we show see Theorem.6) that for large enough, with high probability as n, every two inepenent subsets of G n) with carinality asymptotically 1 + β)log /)n, where 1 β > either have intersection size at least 1 + z)log /)n or at most 1 z)log /)n, for some z < β. The result is establishe using a straightforwar first moment argument: we compute the expecte number of pairs of inepenent sets with intersection lying in the interval [1 z)log /)n, 1 + z)log /)n], an show that this expectation converges to zero exponentially fast. We remark that even though our conclusion is somewhat stronger than that of previous clustering results, our proof of the clustering phenomenon is extremely simple. With this result at han, the refutation of the HLS conjecture is fairly simple to erive. We prove see Theorem.7) that if local algorithms can construct inepenent sets of size asymptotically 1 + β)log /)n, then, by means of a simple coupling construction, we can construct two inepenent sets with intersection size z for all z in the interval [1 + β) log /) n, 1 + β)log /)n], clearly violating the clustering property. The aitional factor 1/ ) is an artifact of the analysis, an hence we believe that our result hols for all β 0, 1]. Namely, no local algorithm is capable of proucing inepenent sets with size larger than factor 1/ of the optimal, asymptotically in. We note again that this coincies with the barrier for known algorithms. It is noteworthy that our result is the first one where algorithmic harness erivation relies irectly on the the geometry of the solution space, viz a vi the clustering phenomena, an thus the connection between algorithmic harness an clustering property is not coinciental. The remainer of the paper is structure as follows. We introuce some basic material an the HLS conjecture in the next section. In the same section we state our main theorem non-valiity of the conjecture Theorem.5). We also state two seconary theorems, the first escribing the overlap structure of inepenent sets in ranom graphs Theorem.6) - the main tool in the proof of our result, an the secon escribing overlaps that can be foun if local algorithms work well Theorem.7). We prove our main theorem easily from the two seconary theorems in Section 3. We prove Theorem.7 in Section 4. We omit the proof of Theorem.6 from this version.. PRELIMINARIES AND MAIN RESULT For convenience, we repeat here some of the notions an efinitions alreay introuce in the first section. Basic graph terminology. All graphs in this paper are unerstoo to be simple unirecte graphs. Given a graph G with noe set V G) an ege set EG), a subset of noes I V G) is an inepenent set if u, v) / EG) for all u, v I. A path between noes u an v with length r is a sequence of noes u 1,..., u r 1 such that u, u 1), u 1, u ),..., u r 1, v) EG). The istance between noes u an v is the length of the shortest path between them. For every positive integer value r an every noe u V G), B G u, r) enotes the epth-r neighborhoo of u in G. Namely, B G u, r) is the subgraph of G inuce by noes v with istance at most r from u. When G is clear from context we rop the subscript. The egree of a vertex u V G) is the number of vertices v such that u, v) EG). The egree of a graph G is the maximum egree of a vertex of G. A graph G is -regular if the egree of every noe is. Ranom graph preliminaries. Given a positive real, Gn, /n) enotes the Erös-Rényi graph on n noes [n] {1,,..., n}, with ege probability /n. Namely each of the n ) eges of a complete graph on n noes belongs to EGn, /n)) with probability /n, inepenently for all eges. Given a positive integer, G n) enotes a graph chosen uniformly at ranom from the space of all -regular graphs on n noes. This efinition is meaningful only when n is an even number, which we assume from now on. Given a positive integer m, let In,, m) enote the set of all inepenent sets in Gn, /n) with carinality m. I n, m) stans for a similar set for the case of ranom regular graphs. Given integers 0 k m, let On,, m, k) enote the set of pairs I, J In,, m) such that I J = k. The efinition of the set O n, m, k) is similar. The sizes of the sets On,, m, k) an O n, m, k), an in particular whether these sets are empty or not, is one of our focuses. Denote by αn, ) the size of a largest in carinality inepenent subset of Gn, /n), normalize by n. Namely, αn, ) = n 1 max{m : In,, m) }. α n) stans for the similar quantity for ranom regular graphs. It is known that αn, ) an α n) have eterministic limits as n. Theorem.1. For every R + there exists α) such that w.h.p. as n, αn, ) α). 1) Similarly, for every positive integer there exists α such that w.h.p. as n α n) α. )

4 Furthermore as. α) = log 1 o1)), 3) α = log 1 o1)), 4) The convergence 1) an ) were establishe in Bayati, Gamarnik an Tetali [4]. The limits 3) an 4) follow from much oler results by Frieze [13] for the case of Erös-Rényi graphs an by Frieze an Luczak [14] for the case of ranom regular graphs, which establishe these limits in the lim sup n an lim inf n sense. The fallout of these results is that graphs Gn, /n) an G n) have inepenent sets of size up to approximately log /)n, when n an are large, namely in the oubly asymptotic sense when we first take n to infinity an then to infinity. Local graph terminology. A ecision function is a measurable function f = fu, G, x) where G is a graph on vertex set [n] for some positive integer n, u [n] is a vertex an x [0, 1] N is a sequence of real numbers for some N n an returns a Boolean value {0, 1}. A ecision function f is sai to compute an inepenent set if for every graph G an every sequence x an for every pair u, v) EG) it is the case that either fu, G, x) = 0 or fv, G, x) = 0, or both. We refer to such an f as an inepenence function. For an inepenence function f, graph G on vertex set [n] an x [0, 1] N for N n, we let I G f, x) enote the inepenent set of G returne by f, i.e., I G f, x) = {u [n] fu, G, x) = 1}. We will assume later that X is chosen ranomly accoring to some probability istribution. In this case I G f, x) is a ranomly chosen inepenent set in G. We now efine the notion of a local ecision function, i.e., one whose actions epen only on the local structure of a graph an the local ranomness. The efinition is a natural one, but we formalize it below for completeness. Let G 1 an G be graphs on vertex sets [n 1] an [n ] respectively. Let u 1 [n 1] an u [n ]. We say that π : [n 1] [n ] is an r-local isomorphism mapping u 1 to u if π is a graph isomorphism from B G1 u 1, r) to B G u, r) so in particular it is a bijection from B G1 u 1, r) to B G u, r), an further it preserves ajacency within B G1 u 1, r) an B G u, r)). For G 1, G, u 1, u an an r-local isomorphism π, we say sequences x 1) [0, 1] N 1 an x ) [0, 1] N are r-locally equivalent if for every v B G1 u 1, r) we have x 1) v = x ) πv). Finally we say fu, G, x) is an r-local function if for every pair of graphs G 1, G, for every pair of vertices u 1 V G 1) an u V G ), for every r-local isomorphism π mapping u 1 to u an r-locally equivalent sequences x 1) an x ) we have fu 1, G 1, x 1) ) = fu, G, x ) ). We often use the notation f r to enote an r-local function. Let n,r 1 + 1) r 1)/ ) enote the number of vertices in a roote tree of egree an epth r. We let T,r enote a canonical roote tree on vertex set [n,r ] with root being 1. For n n,r, x [0, 1] n an an r-local function f r, we let f rx) enote the quantity f r1, T,r, x). Let X be chosen accoring to a uniform istribution on [0, 1] n. The set subset of noes I G n)f r, X) is calle i.i.. factor prouce by the r-local function f r. As we will see below the αf r) 1 E n X[f rx)] accurately captures to within an aitive o1) factor) the ensity of an inepenent returne by an r-local inepenence function f r on G n). First we recall the following folklore proposition which we will also use often in this paper. Proposition.. As n, with probability tening to 1 almost all local neighborhoos in G n) look like a tree. Formally, for every, r an ɛ, for sufficiently large n, P G n) {u [n] BG n)u, r) = T,r } ɛn ) ɛ. This immeiately implies that the expecte value of the inepenent set I G n)f r, X) prouce by f r is αf r)n+on). In fact the following concentration result hols. Proposition.3. As n, with probability tening to 1 the inepenent set prouce by a r-local function f on G n) is of size αf) n + on). Formally, for every, r, ɛ an every r-local function f, for sufficiently large n, P G n),x [0,1] N IG n)f r, X) αf r)n ɛn ) ɛ. Proof. The proof follows from by the fact that the variance of I G n),x is On) an its expectation is αf r)n+on), an so the concentration follows by Chebychev s inequality. The boun on the variance in turn follows from the fact that for every graph G, there are at most On) pairs of vertices u an v for which the events fu, G, X) an fv, G, X) are not inepenent for ranom X. Details omitte. The Hatami-Lovász-Szegey Conjecture an our result. We now turn to escribing the Hatami-Lovász-Szegey HLS) conjecture an our result. Recall α efine by ). The HLS conjecture can be state as follows. Conjecture.4. There exists a sequence of r-local inepenence functions f r, r 1 such that almost surely If r, n) is an inepenent set in G n) an αf r) α as r. Namely, the conjecture asserts the existence of a local algorithm r-local inepenence function f r) which is capable of proucing inepenent sets in G r) of carinality close to the largest that exist. For such an algorithm to be efficient the function f ru, G, x) shoul also be efficiently computable uniformly. Even setting this issue asie, we show that there is a limit on the power of local algorithms to fin large inepenent sets in G n) an in particular the HLS conjecture oes not hol. Let ˆα = sup r sup fr αf r), where the secon supremum is taken over all r-local inepenence functions f r. Theorem.5. [Main] For every ɛ > 0 an all sufficiently large, ˆα 1 α ɛ. That is, for every ɛ > 0 an for all sufficiently large, a largest inepenent set obtainable by r-local functions is at most ɛ for all r. Thus for all large enough there is a multiplicative gap between ˆα an the inepenence ratio α. That being sai, our result oes not rule out that for small, ˆα in fact equals α, thus leaving the HLS conjecture open in this regime.

5 The two main ingreients in our proof of Theorem.5 both eal with the overlaps between inepenent sets in ranom regular graphs. Informally, our first result on the size of the overlaps shows that in ranom graphs the overlaps are not of intermeiate size this is formalize in Theorem.6. We then show that we can apply any r-local function f r twice, with couple ranomness, to prouce two inepenent sets of intermeiate overlap where the size of the overlap epens on the size of the inepenent sets foun by f r an the level of coupling. This is formalize in Theorem.7 Theorem.5 follows immeiately by combinig the two theorems an appropriate setting of parameters). Overlaps in ranom graphs. We now state our main theorem about the overlap of large inepenent sets. We interpret the statement after we make the formal statement. Theorem.6. For β 1/, 1) an 0 < z < β 1 < β an, let s = 1 + β) 1 log an let Kz) enote the set 1 z)n log 1+z)n log of integers between an. Then, for all large enough, we have ) lim P k Kz) On,, sn, k) = 0, 5) n an ) lim P k Kz) O n, sn, k) = 0. 6) n In other wors, both in the Erös-Rényi an in the ranom regular graph moels, when β > 1/, an is large enough, with probability approaching unity as n, one cannot fin a pair of inepenent sets I an J with size ns, such that their overlap intersection) has carinality n1 z) log n1+z) log. at least an at most Note that for all β > 1/, there exists z satisfying 0 < z < β 1 an so the theorem is not vacuous in this setting. Furthermore as β 1, z can be chosen arbitrarily close to 1 making the forbien overlap region extremely broa. That is, as the size of the inepenent sets in consieration approaches the maximum possible namely as β 1), an as, we can take z 1. In other wors, with probability approaching one, two nearly largest inepenent sets either overlap almost entirely or almost o not have an intersection. This is the key result for establishing our harness bouns for existence of local algorithms. A slightly ifferent version of the first of these results can be foun as Lemma 1 in [10]. The latter paper shows that if an inepenent set I is chosen uniformly at ranom from the set with size nearly 1 + β)n log /, then with high probability with respect to the choice of I), there exists an empty overlap region in the sense escribe above. In fact, this empty overlap region exists for every β 0, 1), as oppose to just 1 > β > 1/ + 1/ ) as in our case. Unfortunately, this result cannot be use for our purposes, since this result oes not rule out the existence of rare sets I for which no empty overlap exists. Overlapping from local algorithms. Next we turn to the formalizing the notion of using a local function f r twice on couple ranomness to prouce overlapping inepenent sets. Fix an r-local inepenence function f r. Given a vector X = X u, 1 u n) of variables X u [0, 1], recall that I G f r, X) enotes the inepenent set of G given by u I G f r, X) if an only if f ru, G, X) = 1. Recall that X is chosen accoring to the uniform istribution on [0, 1] n. Namely, X u are inepenent an uniformly istribute over [0, 1]. In what follows we consier some joint istributions on pairs of vectors X, Y) such that marginal istributions on the vector X an Y are uniform on [0, 1] n, though X an Y are epenent on each other. The intuition behin the proof of Theorem.5 is as follows. Note that if X = Y then I G f r, X) = I G f r, Y). As a result the overlap I G f r, X) I G f r, Y) between I G f r, X) an I G f r, Y) is αf r)n + on) in expectation. On the other han, if X an Y are inepenent, then the overlap between I G f r, X) an I G f r, Y) is α f r)n + on) in expectation, since the ecision to pick a vertex u in I is inepenent for most vertices when X an Y are inepenent. In particular, note that if the local neighborhoo aroun u is a tree, which accoring to Proposition. happens with probability approaching unity, then the two ecisions are inepenent, an u I with probability αf r).) Our main theorem shows that by coupling the variables, the overlap can be arrange to be of any intermeiate size, to within an aitive on) factor. In particular, if αf r) excees we will be able to show that the overlap can be arrange to be between the values 1 z)n log 1+z)n log an, escribe in Theorem.6 which contraicts the statement of this theorem. Theorem.7. Fix a positive integer. For constant r, let f ru, G, x) be an r-local inepenence function an let α = αf r). For every γ [α, α] an ɛ > 0, an for every sufficiently large n, there exists a istribution on variables X, Y) [0, 1] n [0, 1] n such that P G n),x,y) IG n)f r, X) I G n)f r, Y) γ ± ɛ)n ) ɛ. 3. PROOF OF THEOREM.5 We now show how Theorems.6 an.7 immeiately imply Theorem.5. Proof Proof of Theorem.5. Fix an r-local function f r an let α = αf r). Fix 0 < η < 1. We will prove below that for sufficiently large we have α/α 1/+1/ )+ η. The theorem will then follow. Let ɛ = η log. By Proposition.3 we have that almost surely an inepenent set returne by f r on G n) is of size at least α ɛ)n. Furthermore for every γ [α, α] we have, by Theorem.7, that G n) almost surely has two inepenent sets I an J, with I, J α ɛ)n an I J [γ ɛ)n, γ + ɛ)n]. 7) Finally, by Theorem.1, we have that for sufficiently large, I, J 1 log )1 + η)n 4 1 log n an so α 1 log, allowing us to set γ = 1 / log. Now we apply Theorem.6 with z = ɛ/ log an β > 1+z. Note that for this choice we have z < 1 an z < β 1 < β < 1. We will also use later the fact that for this choice we have β 1/ +z = 1/ +ɛ 1 log.) Theorem.6 asserts that almost surely G n) has no inepenent sets of size at least 1 + β) 1 log n with intersection size in [1 z) 1 log n, 1 + z) 1 log n]. Since I J [γ ɛ)n, γ +ɛ)n] = [1 z) 1 log n, 1+z) 1 log n], we conclue that min{ I, J } 1 + β) 1 log n. Combining with Equation 7) we get that α ɛ)n min{ I, J }

6 1 + β) 1 log n an so α 1 + β) 1 log + ɛ, which by the given boun on β yiels α 1 + 1/ ) 1 log + ɛ = 1 + 1/ + η) 1 log. On the other han we also have α η) 1 log. It follows that α/α 1/+1/ +η as esire. 4. PROOF OF THEOREM.7 For parameter p [0, 1], we efine the p-correlate istribution on vectors of ranom variables X, Y) to be the following: Let X, Z be inepenent uniform vectors over [0, 1] n. Now let Z u = X u with probability p an Y u with probability 1 p inepenently for every u V G). Let fu, G, x) an α be as in the theorem statement. Recall that fx) = f1, T,r, x) is the ecision of f on the canonical tree of egree an epth r roote at the vertex 1. Let γp) be the probability that fx) = 1 an fy) = 1, for p-correlate variables X, Y). As with Proposition.3 we have the following. Lemma 4.1. For every, r, ɛ > 0 an r-local function f, for sufficiently large n we have: P I G n)f, X) I G n)f, Y) γp) n ɛn ) ɛ, where X, Y) are p-correlate istributions on [0, 1] n. Proof. By Proposition. we have that almost surely almost all local neighborhoos are trees an so for most vertices u the probability that u is chosen to be in the inepenent sets If, X) an If, Y) is γp). By linearity of expectations we get that E[ If, X) If, Y) ] = γp) n + on). Again observing that most local neighborhoos are isjoint we have that the variance of If, X) If, Y) is On). We conclue, by applying the Chebychev boun, that If, X) If, Y) is concentrate aroun the expectation an the lemma follows. We also note that for p = 1 an p = 0 the quantity γp) follow immeiately from their efinition. Proposition 4.. γ1) = α an γ0) = α. Now to prove Theorem.7 it suffices to prove that for every γ [α, α] there exists a p such that γp) = γ. We show this next by showing that γp) is continuous. Lemma 4.3. For every r, γp) is a continuous function of p. Proof. Let W u, u T,r ) be ranom variables associate with noes in T,r, uniformly istribute over [0, 1], which are inepenent for ifferent u an also inepenent from X u an Z u. We use W u as generators for the events Y u = X u vs Y u = Z u. In particular, given p, set Y u = X u if W u p an Y u = Z u otherwise. This process is exactly the process of setting variables Y u to X u an Z u with probabilities p an 1 p respectively, inepenently for all noes u. Now fix any p 1 < p, an let δ < p p 1)/ r+1. We use the notation f rx u, Z u, W u, p) to enote the value of f r when the see variables realization is W u, u T,r ), an the threshol value p is use. Namely, f rx u, Z u, W u, p) = f r X u1{w u p} + Z u1{w u > p}, u T,r ). Here, for ease of notation, the reference to the tree T,r is roppe. Utilizing this notation we have γp) = P f rx u) = f rx u, Z u, W u, p) = 1). Manipulating the expressions somewhat we fin γp ) γp 1) r+1 p p 1). Since r is fixe, the continuity of γp) is establishe. We are now reay to prove Theorem.7. Proof Proof of Theorem.7. Given γ [α, α] by Lemma 4.3 we have that there exists a p such that γ = γp). For this choice of p, let X, Y) be a pair of p-correlate istributions. Applying Lemma 4.1 to this choice of p, we get that with probability at least 1 ɛ we have I G n)f, X) I G n)f, Y) [γ ɛ)n, γ + ɛ)n] as esire. 5. PROOF OF THEOREM.6 We omit the proof of this theorem from this version. Details can be foun in the full version [15]. Acknowlegements The first author wishes to thank Microsoft Research New Englan for proviing exciting an hospitable environment uring his visit in 01 when this work was conucte. The same author also wishes acknowlege the general support of NSF grant CMMI REFERENCES [1] D. Achlioptas, A. Coja-Oghlan, an F. Ricci-Tersenghi. On the solution space geometry of ranom formulas. Ranom Structures an Algorithms, 38:51 68, 011. [] D. Alous. Some open problems. alous/ Research/OP/inex.html. [3] N. Alon an J. Spencer. Probabilistic Metho. Wiley, 199. [4] M. Bayati, D. Gamarnik, an P. Tetali. Combinatorial approach to the interpolation metho an scaling limits in sparse ranom graphs. Annals of Probability, to appear. Conference version in Proc. 4n Ann. Symposium on the Theory of Computing STOC), 010. [5] C. Borgs, J. Chayes, an D. Gamarnik. Convergent sequences of sparse graphs: A large eviations approach. arxiv: [math.pr], 013. [6] C. Borgs, J. Chayes, L. Lovász, an J. Kahn. Left an right convergence of graphs with boune egree [7] C. Borgs, J. Chayes, L. Lovász, V. Sós, an K. Vesztergombi. Convergent graph sequences II: Multiway cuts an statistical physics. Ann. of Math., 176:151 19, 01. [8] C. Borgs, J. Chayes, L. Lovász, V. Sós, an K. Vesztergombi. Convergent graph sequences I: Subgraph frequencies, metric properties, an testing. Avances in Math., 19: , 08. [9] A. Coja-Oghlan. On belief propagation guie ecimation for ranom k-sat. In Proceeings of the Twenty-Secon Annual ACM-SIAM Symposium on Discrete Algorithms, pages SIAM, 011.

7 [10] A. Coja-Oghlan an C. Efthymiou. On inepenent sets in ranom graphs. In Proceeings of the Twenty-Secon Annual ACM-SIAM Symposium on Discrete Algorithms, pages SIAM, 011. [11] E. Csoka an G. Lippner. Invariant ranom matchings in cayley graphs. arxiv preprint arxiv: , 01. [1] G. Elek an G. Lippner. Borel oracles. an analytical approach to constant-time algorithms. In Proc. Amer. Math. Soc, volume 138, pages , 010. [13] A. Frieze. On the inepenence number of ranom graphs. Discrete Mathematics, 81: , [14] A. Frieze an T. Luczak. On the inepenence an chromatic numbers of ranom regular graphs. Journal of Combinatorial Theory, Series B, 541):13 13, 199. [15] D. Gamarnik an M. Suan. Limits of local algorithms over sparse ranom graphs. Electronic Colloquium on Computational Complexity ECCC), 0:55, 013. [16] A. Hassiim, J. A. Kelner, H. N. Nguyen, an K. Onak. Local graph partitions for approximation an testing. In FOCS, pages 31. IEEE Computer Society, 009. [17] H. Hatami, L. Lovász, an B. Szegey. Limits of local-global convergent graph sequences. Preprint at [18] R. Karp an M. Sipser. Maximum matchings in sparse ranom graphs. In n Annual Symposium on Founations of Computer Science, pages , [19] N. Linial. Locality in istribute graph algorithms. SIAM J. Comput., 11):193 01, 199. [0] L. Lovász an B. Szegey. Limits of ense graph sequences. Journal of Combinatorial Theory, Series B, 96: , 006. [1] M. Luby. A simple parallel algorithm for the maximal inepenent set problem. In R. Segewick, eitor, STOC, pages ACM, [] R. Lyons an F. Nazarov. Perfect matchings as ii factors on non-amenable groups. European Journal of Combinatorics, 37): , 011. [3] M. Mezar an A. Montanari. Information, Physics an Computation. Oxfor grauate texts, 009. [4] M. Mézar, T. Mora, an R. Zecchina. Clustering of solutions in the ranom satisfiability problem. Physical Review Letters, 9419):19705, 005. [5] H. N. Nguyen an K. Onak. Constant-time approximation algorithms via local improvements. In FOCS, pages IEEE Computer Society, 008. [6] M. Parnas an D. Ron. Approximating the minimum vertex cover in sublinear time an a connection to istribute algorithms. Theor. Comput. Sci., ): , 007. [7] R. Rubinfel, G. Tamir, S. Vari, an N. Xie. Fast local computation algorithms. In B. Chazelle, eitor, ICS, pages Tsinghua University Press, 011. [8] M. Talagran. Mean Fiel Moels for Spin Glasses: Volume I: Basic Examples. Springer, 010. [9] M. J. Wainwright an M. I. Joran. A variational principle for graphical moels. In New Directions in Statistical Signal Processing: From Systems to Brain. Cambrige, MA: MIT Press, 005.

Limits of local algorithms over sparse random graphs

Limits of local algorithms over sparse random graphs Limits of local algorithms over sparse ranom graphs The MIT Faculty has mae this article openly available. Please share how this access benefits you. Your story matters. Citation As Publishe Publisher

More information

Lower Bounds for the Smoothed Number of Pareto optimal Solutions

Lower Bounds for the Smoothed Number of Pareto optimal Solutions Lower Bouns for the Smoothe Number of Pareto optimal Solutions Tobias Brunsch an Heiko Röglin Department of Computer Science, University of Bonn, Germany brunsch@cs.uni-bonn.e, heiko@roeglin.org Abstract.

More information

Chromatic number for a generalization of Cartesian product graphs

Chromatic number for a generalization of Cartesian product graphs Chromatic number for a generalization of Cartesian prouct graphs Daniel Král Douglas B. West Abstract Let G be a class of graphs. The -fol gri over G, enote G, is the family of graphs obtaine from -imensional

More information

Lecture Introduction. 2 Examples of Measure Concentration. 3 The Johnson-Lindenstrauss Lemma. CS-621 Theory Gems November 28, 2012

Lecture Introduction. 2 Examples of Measure Concentration. 3 The Johnson-Lindenstrauss Lemma. CS-621 Theory Gems November 28, 2012 CS-6 Theory Gems November 8, 0 Lecture Lecturer: Alesaner Mąry Scribes: Alhussein Fawzi, Dorina Thanou Introuction Toay, we will briefly iscuss an important technique in probability theory measure concentration

More information

The chromatic number of graph powers

The chromatic number of graph powers Combinatorics, Probability an Computing (19XX) 00, 000 000. c 19XX Cambrige University Press Printe in the Unite Kingom The chromatic number of graph powers N O G A A L O N 1 an B O J A N M O H A R 1 Department

More information

Survey Sampling. 1 Design-based Inference. Kosuke Imai Department of Politics, Princeton University. February 19, 2013

Survey Sampling. 1 Design-based Inference. Kosuke Imai Department of Politics, Princeton University. February 19, 2013 Survey Sampling Kosuke Imai Department of Politics, Princeton University February 19, 2013 Survey sampling is one of the most commonly use ata collection methos for social scientists. We begin by escribing

More information

Ramsey numbers of some bipartite graphs versus complete graphs

Ramsey numbers of some bipartite graphs versus complete graphs Ramsey numbers of some bipartite graphs versus complete graphs Tao Jiang, Michael Salerno Miami University, Oxfor, OH 45056, USA Abstract. The Ramsey number r(h, K n ) is the smallest positive integer

More information

A new proof of the sharpness of the phase transition for Bernoulli percolation on Z d

A new proof of the sharpness of the phase transition for Bernoulli percolation on Z d A new proof of the sharpness of the phase transition for Bernoulli percolation on Z Hugo Duminil-Copin an Vincent Tassion October 8, 205 Abstract We provie a new proof of the sharpness of the phase transition

More information

A Sketch of Menshikov s Theorem

A Sketch of Menshikov s Theorem A Sketch of Menshikov s Theorem Thomas Bao March 14, 2010 Abstract Let Λ be an infinite, locally finite oriente multi-graph with C Λ finite an strongly connecte, an let p

More information

Perfect Matchings in Õ(n1.5 ) Time in Regular Bipartite Graphs

Perfect Matchings in Õ(n1.5 ) Time in Regular Bipartite Graphs Perfect Matchings in Õ(n1.5 ) Time in Regular Bipartite Graphs Ashish Goel Michael Kapralov Sanjeev Khanna Abstract We consier the well-stuie problem of fining a perfect matching in -regular bipartite

More information

Lower bounds on Locality Sensitive Hashing

Lower bounds on Locality Sensitive Hashing Lower bouns on Locality Sensitive Hashing Rajeev Motwani Assaf Naor Rina Panigrahy Abstract Given a metric space (X, X ), c 1, r > 0, an p, q [0, 1], a istribution over mappings H : X N is calle a (r,

More information

Lecture 5. Symmetric Shearer s Lemma

Lecture 5. Symmetric Shearer s Lemma Stanfor University Spring 208 Math 233: Non-constructive methos in combinatorics Instructor: Jan Vonrák Lecture ate: January 23, 208 Original scribe: Erik Bates Lecture 5 Symmetric Shearer s Lemma Here

More information

Sharp Thresholds. Zachary Hamaker. March 15, 2010

Sharp Thresholds. Zachary Hamaker. March 15, 2010 Sharp Threshols Zachary Hamaker March 15, 2010 Abstract The Kolmogorov Zero-One law states that for tail events on infinite-imensional probability spaces, the probability must be either zero or one. Behavior

More information

u!i = a T u = 0. Then S satisfies

u!i = a T u = 0. Then S satisfies Deterministic Conitions for Subspace Ientifiability from Incomplete Sampling Daniel L Pimentel-Alarcón, Nigel Boston, Robert D Nowak University of Wisconsin-Maison Abstract Consier an r-imensional subspace

More information

Convergence of Random Walks

Convergence of Random Walks Chapter 16 Convergence of Ranom Walks This lecture examines the convergence of ranom walks to the Wiener process. This is very important both physically an statistically, an illustrates the utility of

More information

Time-of-Arrival Estimation in Non-Line-Of-Sight Environments

Time-of-Arrival Estimation in Non-Line-Of-Sight Environments 2 Conference on Information Sciences an Systems, The Johns Hopkins University, March 2, 2 Time-of-Arrival Estimation in Non-Line-Of-Sight Environments Sinan Gezici, Hisashi Kobayashi an H. Vincent Poor

More information

Iterated Point-Line Configurations Grow Doubly-Exponentially

Iterated Point-Line Configurations Grow Doubly-Exponentially Iterate Point-Line Configurations Grow Doubly-Exponentially Joshua Cooper an Mark Walters July 9, 008 Abstract Begin with a set of four points in the real plane in general position. A to this collection

More information

A Random Graph Model for Massive Graphs

A Random Graph Model for Massive Graphs A Ranom Graph Moel for Massive Graphs William Aiello AT&T Labs Florham Park, New Jersey aiello@research.att.com Fan Chung University of California, San Diego fan@ucs.eu Linyuan Lu University of Pennsylvania,

More information

Logarithmic spurious regressions

Logarithmic spurious regressions Logarithmic spurious regressions Robert M. e Jong Michigan State University February 5, 22 Abstract Spurious regressions, i.e. regressions in which an integrate process is regresse on another integrate

More information

Diophantine Approximations: Examining the Farey Process and its Method on Producing Best Approximations

Diophantine Approximations: Examining the Farey Process and its Method on Producing Best Approximations Diophantine Approximations: Examining the Farey Process an its Metho on Proucing Best Approximations Kelly Bowen Introuction When a person hears the phrase irrational number, one oes not think of anything

More information

Math 342 Partial Differential Equations «Viktor Grigoryan

Math 342 Partial Differential Equations «Viktor Grigoryan Math 342 Partial Differential Equations «Viktor Grigoryan 6 Wave equation: solution In this lecture we will solve the wave equation on the entire real line x R. This correspons to a string of infinite

More information

FLUCTUATIONS IN THE NUMBER OF POINTS ON SMOOTH PLANE CURVES OVER FINITE FIELDS. 1. Introduction

FLUCTUATIONS IN THE NUMBER OF POINTS ON SMOOTH PLANE CURVES OVER FINITE FIELDS. 1. Introduction FLUCTUATIONS IN THE NUMBER OF POINTS ON SMOOTH PLANE CURVES OVER FINITE FIELDS ALINA BUCUR, CHANTAL DAVID, BROOKE FEIGON, MATILDE LALÍN 1 Introuction In this note, we stuy the fluctuations in the number

More information

6 General properties of an autonomous system of two first order ODE

6 General properties of an autonomous system of two first order ODE 6 General properties of an autonomous system of two first orer ODE Here we embark on stuying the autonomous system of two first orer ifferential equations of the form ẋ 1 = f 1 (, x 2 ), ẋ 2 = f 2 (, x

More information

New bounds on Simonyi s conjecture

New bounds on Simonyi s conjecture New bouns on Simonyi s conjecture Daniel Soltész soltesz@math.bme.hu Department of Computer Science an Information Theory, Buapest University of Technology an Economics arxiv:1510.07597v1 [math.co] 6 Oct

More information

Discrete Mathematics

Discrete Mathematics Discrete Mathematics 309 (009) 86 869 Contents lists available at ScienceDirect Discrete Mathematics journal homepage: wwwelseviercom/locate/isc Profile vectors in the lattice of subspaces Dániel Gerbner

More information

ensembles When working with density operators, we can use this connection to define a generalized Bloch vector: v x Tr x, v y Tr y

ensembles When working with density operators, we can use this connection to define a generalized Bloch vector: v x Tr x, v y Tr y Ph195a lecture notes, 1/3/01 Density operators for spin- 1 ensembles So far in our iscussion of spin- 1 systems, we have restricte our attention to the case of pure states an Hamiltonian evolution. Toay

More information

MULTIFRACTAL NETWORK GENERATORS

MULTIFRACTAL NETWORK GENERATORS MULTIFRACTAL NETWORK GENERATORS AUSTIN R. BENSON, CARLOS RIQUELME, SVEN P. SCHMIT (0) Abstract. Generating ranom graphs to moel networks has a rich history. In this paper, we explore a recent generative

More information

arxiv: v1 [math.co] 13 Dec 2017

arxiv: v1 [math.co] 13 Dec 2017 The List Linear Arboricity of Graphs arxiv:7.05006v [math.co] 3 Dec 07 Ringi Kim Department of Mathematical Sciences KAIST Daejeon South Korea 344 an Luke Postle Department of Combinatorics an Optimization

More information

Topic 7: Convergence of Random Variables

Topic 7: Convergence of Random Variables Topic 7: Convergence of Ranom Variables Course 003, 2016 Page 0 The Inference Problem So far, our starting point has been a given probability space (S, F, P). We now look at how to generate information

More information

Leaving Randomness to Nature: d-dimensional Product Codes through the lens of Generalized-LDPC codes

Leaving Randomness to Nature: d-dimensional Product Codes through the lens of Generalized-LDPC codes Leaving Ranomness to Nature: -Dimensional Prouct Coes through the lens of Generalize-LDPC coes Tavor Baharav, Kannan Ramchanran Dept. of Electrical Engineering an Computer Sciences, U.C. Berkeley {tavorb,

More information

On colour-blind distinguishing colour pallets in regular graphs

On colour-blind distinguishing colour pallets in regular graphs J Comb Optim (2014 28:348 357 DOI 10.1007/s10878-012-9556-x On colour-blin istinguishing colour pallets in regular graphs Jakub Przybyło Publishe online: 25 October 2012 The Author(s 2012. This article

More information

Math Notes on differentials, the Chain Rule, gradients, directional derivative, and normal vectors

Math Notes on differentials, the Chain Rule, gradients, directional derivative, and normal vectors Math 18.02 Notes on ifferentials, the Chain Rule, graients, irectional erivative, an normal vectors Tangent plane an linear approximation We efine the partial erivatives of f( xy, ) as follows: f f( x+

More information

Math 1B, lecture 8: Integration by parts

Math 1B, lecture 8: Integration by parts Math B, lecture 8: Integration by parts Nathan Pflueger 23 September 2 Introuction Integration by parts, similarly to integration by substitution, reverses a well-known technique of ifferentiation an explores

More information

Computing Exact Confidence Coefficients of Simultaneous Confidence Intervals for Multinomial Proportions and their Functions

Computing Exact Confidence Coefficients of Simultaneous Confidence Intervals for Multinomial Proportions and their Functions Working Paper 2013:5 Department of Statistics Computing Exact Confience Coefficients of Simultaneous Confience Intervals for Multinomial Proportions an their Functions Shaobo Jin Working Paper 2013:5

More information

The Exact Form and General Integrating Factors

The Exact Form and General Integrating Factors 7 The Exact Form an General Integrating Factors In the previous chapters, we ve seen how separable an linear ifferential equations can be solve using methos for converting them to forms that can be easily

More information

arxiv: v1 [math.co] 29 May 2009

arxiv: v1 [math.co] 29 May 2009 arxiv:0905.4913v1 [math.co] 29 May 2009 simple Havel-Hakimi type algorithm to realize graphical egree sequences of irecte graphs Péter L. Erős an István Miklós. Rényi Institute of Mathematics, Hungarian

More information

Least-Squares Regression on Sparse Spaces

Least-Squares Regression on Sparse Spaces Least-Squares Regression on Sparse Spaces Yuri Grinberg, Mahi Milani Far, Joelle Pineau School of Computer Science McGill University Montreal, Canaa {ygrinb,mmilan1,jpineau}@cs.mcgill.ca 1 Introuction

More information

A Weak First Digit Law for a Class of Sequences

A Weak First Digit Law for a Class of Sequences International Mathematical Forum, Vol. 11, 2016, no. 15, 67-702 HIKARI Lt, www.m-hikari.com http://x.oi.org/10.1288/imf.2016.6562 A Weak First Digit Law for a Class of Sequences M. A. Nyblom School of

More information

A PAC-Bayesian Approach to Spectrally-Normalized Margin Bounds for Neural Networks

A PAC-Bayesian Approach to Spectrally-Normalized Margin Bounds for Neural Networks A PAC-Bayesian Approach to Spectrally-Normalize Margin Bouns for Neural Networks Behnam Neyshabur, Srinah Bhojanapalli, Davi McAllester, Nathan Srebro Toyota Technological Institute at Chicago {bneyshabur,

More information

Equilibrium in Queues Under Unknown Service Times and Service Value

Equilibrium in Queues Under Unknown Service Times and Service Value University of Pennsylvania ScholarlyCommons Finance Papers Wharton Faculty Research 1-2014 Equilibrium in Queues Uner Unknown Service Times an Service Value Laurens Debo Senthil K. Veeraraghavan University

More information

Upper and Lower Bounds on ε-approximate Degree of AND n and OR n Using Chebyshev Polynomials

Upper and Lower Bounds on ε-approximate Degree of AND n and OR n Using Chebyshev Polynomials Upper an Lower Bouns on ε-approximate Degree of AND n an OR n Using Chebyshev Polynomials Mrinalkanti Ghosh, Rachit Nimavat December 11, 016 1 Introuction The notion of approximate egree was first introuce

More information

REAL ANALYSIS I HOMEWORK 5

REAL ANALYSIS I HOMEWORK 5 REAL ANALYSIS I HOMEWORK 5 CİHAN BAHRAN The questions are from Stein an Shakarchi s text, Chapter 3. 1. Suppose ϕ is an integrable function on R with R ϕ(x)x = 1. Let K δ(x) = δ ϕ(x/δ), δ > 0. (a) Prove

More information

Stopping-Set Enumerator Approximations for Finite-Length Protograph LDPC Codes

Stopping-Set Enumerator Approximations for Finite-Length Protograph LDPC Codes Stopping-Set Enumerator Approximations for Finite-Length Protograph LDPC Coes Kaiann Fu an Achilleas Anastasopoulos Electrical Engineering an Computer Science Dept. University of Michigan, Ann Arbor, MI

More information

Lecture 6: Calculus. In Song Kim. September 7, 2011

Lecture 6: Calculus. In Song Kim. September 7, 2011 Lecture 6: Calculus In Song Kim September 7, 20 Introuction to Differential Calculus In our previous lecture we came up with several ways to analyze functions. We saw previously that the slope of a linear

More information

arxiv: v4 [cs.ds] 7 Mar 2014

arxiv: v4 [cs.ds] 7 Mar 2014 Analysis of Agglomerative Clustering Marcel R. Ackermann Johannes Blömer Daniel Kuntze Christian Sohler arxiv:101.697v [cs.ds] 7 Mar 01 Abstract The iameter k-clustering problem is the problem of partitioning

More information

Linear First-Order Equations

Linear First-Order Equations 5 Linear First-Orer Equations Linear first-orer ifferential equations make up another important class of ifferential equations that commonly arise in applications an are relatively easy to solve (in theory)

More information

Anagram-free colorings of graphs

Anagram-free colorings of graphs Anagram-free colorings of graphs Nina Kamčev Tomasz Luczak Benny Suakov Abstract A sequence S is calle anagram-free if it contains no consecutive symbols r 1 r... r k r k+1... r k such that r k+1... r

More information

arxiv: v1 [math.co] 15 Sep 2015

arxiv: v1 [math.co] 15 Sep 2015 Circular coloring of signe graphs Yingli Kang, Eckhar Steffen arxiv:1509.04488v1 [math.co] 15 Sep 015 Abstract Let k, ( k) be two positive integers. We generalize the well stuie notions of (k, )-colorings

More information

Separation of Variables

Separation of Variables Physics 342 Lecture 1 Separation of Variables Lecture 1 Physics 342 Quantum Mechanics I Monay, January 25th, 2010 There are three basic mathematical tools we nee, an then we can begin working on the physical

More information

An Optimal Algorithm for Bandit and Zero-Order Convex Optimization with Two-Point Feedback

An Optimal Algorithm for Bandit and Zero-Order Convex Optimization with Two-Point Feedback Journal of Machine Learning Research 8 07) - Submitte /6; Publishe 5/7 An Optimal Algorithm for Banit an Zero-Orer Convex Optimization with wo-point Feeback Oha Shamir Department of Computer Science an

More information

The number of K s,t -free graphs

The number of K s,t -free graphs The number of K s,t -free graphs József Balogh an Wojciech Samotij August 17, 2010 Abstract Denote by f n (H the number of (labele H-free graphs on a fixe vertex set of size n. Erős conjecture that whenever

More information

TAYLOR S POLYNOMIAL APPROXIMATION FOR FUNCTIONS

TAYLOR S POLYNOMIAL APPROXIMATION FOR FUNCTIONS MISN-0-4 TAYLOR S POLYNOMIAL APPROXIMATION FOR FUNCTIONS f(x ± ) = f(x) ± f ' (x) + f '' (x) 2 ±... 1! 2! = 1.000 ± 0.100 + 0.005 ±... TAYLOR S POLYNOMIAL APPROXIMATION FOR FUNCTIONS by Peter Signell 1.

More information

arxiv: v1 [math.co] 27 Nov 2018

arxiv: v1 [math.co] 27 Nov 2018 FURTHER RESULTS ON THE INDUCIBILITY OF -ARY TREES AUDACE A. V. DOSSOU-OLORY AND STEPHAN WAGNER arxiv:1811.11235v1 [math.co] 27 Nov 2018 Abstract. For a -ary tree every vertex has outegree between 2 an

More information

Permanent vs. Determinant

Permanent vs. Determinant Permanent vs. Determinant Frank Ban Introuction A major problem in theoretical computer science is the Permanent vs. Determinant problem. It asks: given an n by n matrix of ineterminates A = (a i,j ) an

More information

Schrödinger s equation.

Schrödinger s equation. Physics 342 Lecture 5 Schröinger s Equation Lecture 5 Physics 342 Quantum Mechanics I Wenesay, February 3r, 2010 Toay we iscuss Schröinger s equation an show that it supports the basic interpretation of

More information

Thermal conductivity of graded composites: Numerical simulations and an effective medium approximation

Thermal conductivity of graded composites: Numerical simulations and an effective medium approximation JOURNAL OF MATERIALS SCIENCE 34 (999)5497 5503 Thermal conuctivity of grae composites: Numerical simulations an an effective meium approximation P. M. HUI Department of Physics, The Chinese University

More information

Some properties of random staircase tableaux

Some properties of random staircase tableaux Some properties of ranom staircase tableaux Sanrine Dasse Hartaut Pawe l Hitczenko Downloae /4/7 to 744940 Reistribution subject to SIAM license or copyright; see http://wwwsiamorg/journals/ojsaphp Abstract

More information

Node Density and Delay in Large-Scale Wireless Networks with Unreliable Links

Node Density and Delay in Large-Scale Wireless Networks with Unreliable Links Noe Density an Delay in Large-Scale Wireless Networks with Unreliable Links Shizhen Zhao, Xinbing Wang Department of Electronic Engineering Shanghai Jiao Tong University, China Email: {shizhenzhao,xwang}@sjtu.eu.cn

More information

arxiv: v2 [math.st] 29 Oct 2015

arxiv: v2 [math.st] 29 Oct 2015 EXPONENTIAL RANDOM SIMPLICIAL COMPLEXES KONSTANTIN ZUEV, OR EISENBERG, AND DMITRI KRIOUKOV arxiv:1502.05032v2 [math.st] 29 Oct 2015 Abstract. Exponential ranom graph moels have attracte significant research

More information

On the Enumeration of Double-Base Chains with Applications to Elliptic Curve Cryptography

On the Enumeration of Double-Base Chains with Applications to Elliptic Curve Cryptography On the Enumeration of Double-Base Chains with Applications to Elliptic Curve Cryptography Christophe Doche Department of Computing Macquarie University, Australia christophe.oche@mq.eu.au. Abstract. The

More information

Chapter 6: Energy-Momentum Tensors

Chapter 6: Energy-Momentum Tensors 49 Chapter 6: Energy-Momentum Tensors This chapter outlines the general theory of energy an momentum conservation in terms of energy-momentum tensors, then applies these ieas to the case of Bohm's moel.

More information

On combinatorial approaches to compressed sensing

On combinatorial approaches to compressed sensing On combinatorial approaches to compresse sensing Abolreza Abolhosseini Moghaam an Hayer Raha Department of Electrical an Computer Engineering, Michigan State University, East Lansing, MI, U.S. Emails:{abolhos,raha}@msu.eu

More information

Introduction to the Vlasov-Poisson system

Introduction to the Vlasov-Poisson system Introuction to the Vlasov-Poisson system Simone Calogero 1 The Vlasov equation Consier a particle with mass m > 0. Let x(t) R 3 enote the position of the particle at time t R an v(t) = ẋ(t) = x(t)/t its

More information

On the number of isolated eigenvalues of a pair of particles in a quantum wire

On the number of isolated eigenvalues of a pair of particles in a quantum wire On the number of isolate eigenvalues of a pair of particles in a quantum wire arxiv:1812.11804v1 [math-ph] 31 Dec 2018 Joachim Kerner 1 Department of Mathematics an Computer Science FernUniversität in

More information

Lecture 2 Lagrangian formulation of classical mechanics Mechanics

Lecture 2 Lagrangian formulation of classical mechanics Mechanics Lecture Lagrangian formulation of classical mechanics 70.00 Mechanics Principle of stationary action MATH-GA To specify a motion uniquely in classical mechanics, it suffices to give, at some time t 0,

More information

Resistant Polynomials and Stronger Lower Bounds for Depth-Three Arithmetical Formulas

Resistant Polynomials and Stronger Lower Bounds for Depth-Three Arithmetical Formulas Resistant Polynomials an Stronger Lower Bouns for Depth-Three Arithmetical Formulas Maurice J. Jansen University at Buffalo Kenneth W.Regan University at Buffalo Abstract We erive quaratic lower bouns

More information

CONTAINMENT GAME PLAYED ON RANDOM GRAPHS: ANOTHER ZIG-ZAG THEOREM

CONTAINMENT GAME PLAYED ON RANDOM GRAPHS: ANOTHER ZIG-ZAG THEOREM CONTAINMENT GAME PLAYED ON RANDOM GRAPHS: ANOTHER ZIG-ZAG THEOREM PAWE L PRA LAT Abstract. We consier a variant of the game of Cops an Robbers, calle Containment, in which cops move from ege to ajacent

More information

On the minimum distance of elliptic curve codes

On the minimum distance of elliptic curve codes On the minimum istance of elliptic curve coes Jiyou Li Department of Mathematics Shanghai Jiao Tong University Shanghai PRChina Email: lijiyou@sjtueucn Daqing Wan Department of Mathematics University of

More information

Acute sets in Euclidean spaces

Acute sets in Euclidean spaces Acute sets in Eucliean spaces Viktor Harangi April, 011 Abstract A finite set H in R is calle an acute set if any angle etermine by three points of H is acute. We examine the maximal carinality α() of

More information

Level Construction of Decision Trees in a Partition-based Framework for Classification

Level Construction of Decision Trees in a Partition-based Framework for Classification Level Construction of Decision Trees in a Partition-base Framework for Classification Y.Y. Yao, Y. Zhao an J.T. Yao Department of Computer Science, University of Regina Regina, Saskatchewan, Canaa S4S

More information

Variable Independence and Resolution Paths for Quantified Boolean Formulas

Variable Independence and Resolution Paths for Quantified Boolean Formulas Variable Inepenence an Resolution Paths for Quantifie Boolean Formulas Allen Van Geler http://www.cse.ucsc.eu/ avg University of California, Santa Cruz Abstract. Variable inepenence in quantifie boolean

More information

Proof of SPNs as Mixture of Trees

Proof of SPNs as Mixture of Trees A Proof of SPNs as Mixture of Trees Theorem 1. If T is an inuce SPN from a complete an ecomposable SPN S, then T is a tree that is complete an ecomposable. Proof. Argue by contraiction that T is not a

More information

1 dx. where is a large constant, i.e., 1, (7.6) and Px is of the order of unity. Indeed, if px is given by (7.5), the inequality (7.

1 dx. where is a large constant, i.e., 1, (7.6) and Px is of the order of unity. Indeed, if px is given by (7.5), the inequality (7. Lectures Nine an Ten The WKB Approximation The WKB metho is a powerful tool to obtain solutions for many physical problems It is generally applicable to problems of wave propagation in which the frequency

More information

Technion - Computer Science Department - M.Sc. Thesis MSC Constrained Codes for Two-Dimensional Channels.

Technion - Computer Science Department - M.Sc. Thesis MSC Constrained Codes for Two-Dimensional Channels. Technion - Computer Science Department - M.Sc. Thesis MSC-2006- - 2006 Constraine Coes for Two-Dimensional Channels Keren Censor Technion - Computer Science Department - M.Sc. Thesis MSC-2006- - 2006 Technion

More information

LATTICE-BASED D-OPTIMUM DESIGN FOR FOURIER REGRESSION

LATTICE-BASED D-OPTIMUM DESIGN FOR FOURIER REGRESSION The Annals of Statistics 1997, Vol. 25, No. 6, 2313 2327 LATTICE-BASED D-OPTIMUM DESIGN FOR FOURIER REGRESSION By Eva Riccomagno, 1 Rainer Schwabe 2 an Henry P. Wynn 1 University of Warwick, Technische

More information

A Modification of the Jarque-Bera Test. for Normality

A Modification of the Jarque-Bera Test. for Normality Int. J. Contemp. Math. Sciences, Vol. 8, 01, no. 17, 84-85 HIKARI Lt, www.m-hikari.com http://x.oi.org/10.1988/ijcms.01.9106 A Moification of the Jarque-Bera Test for Normality Moawa El-Fallah Ab El-Salam

More information

Bisecting Sparse Random Graphs

Bisecting Sparse Random Graphs Bisecting Sparse Ranom Graphs Malwina J. Luczak,, Colin McDiarmi Mathematical Institute, University of Oxfor, Oxfor OX 3LB, Unite Kingom; e-mail: luczak@maths.ox.ac.uk Department of Statistics, University

More information

The Principle of Least Action

The Principle of Least Action Chapter 7. The Principle of Least Action 7.1 Force Methos vs. Energy Methos We have so far stuie two istinct ways of analyzing physics problems: force methos, basically consisting of the application of

More information

Improved Rate-Based Pull and Push Strategies in Large Distributed Networks

Improved Rate-Based Pull and Push Strategies in Large Distributed Networks Improve Rate-Base Pull an Push Strategies in Large Distribute Networks Wouter Minnebo an Benny Van Hout Department of Mathematics an Computer Science University of Antwerp - imins Mielheimlaan, B-00 Antwerp,

More information

inflow outflow Part I. Regular tasks for MAE598/494 Task 1

inflow outflow Part I. Regular tasks for MAE598/494 Task 1 MAE 494/598, Fall 2016 Project #1 (Regular tasks = 20 points) Har copy of report is ue at the start of class on the ue ate. The rules on collaboration will be release separately. Please always follow the

More information

Entanglement is not very useful for estimating multiple phases

Entanglement is not very useful for estimating multiple phases PHYSICAL REVIEW A 70, 032310 (2004) Entanglement is not very useful for estimating multiple phases Manuel A. Ballester* Department of Mathematics, University of Utrecht, Box 80010, 3508 TA Utrecht, The

More information

Tractability results for weighted Banach spaces of smooth functions

Tractability results for weighted Banach spaces of smooth functions Tractability results for weighte Banach spaces of smooth functions Markus Weimar Mathematisches Institut, Universität Jena Ernst-Abbe-Platz 2, 07740 Jena, Germany email: markus.weimar@uni-jena.e March

More information

arxiv: v4 [math.pr] 27 Jul 2016

arxiv: v4 [math.pr] 27 Jul 2016 The Asymptotic Distribution of the Determinant of a Ranom Correlation Matrix arxiv:309768v4 mathpr] 7 Jul 06 AM Hanea a, & GF Nane b a Centre of xcellence for Biosecurity Risk Analysis, University of Melbourne,

More information

Construction of the Electronic Radial Wave Functions and Probability Distributions of Hydrogen-like Systems

Construction of the Electronic Radial Wave Functions and Probability Distributions of Hydrogen-like Systems Construction of the Electronic Raial Wave Functions an Probability Distributions of Hyrogen-like Systems Thomas S. Kuntzleman, Department of Chemistry Spring Arbor University, Spring Arbor MI 498 tkuntzle@arbor.eu

More information

DEGREE DISTRIBUTION OF SHORTEST PATH TREES AND BIAS OF NETWORK SAMPLING ALGORITHMS

DEGREE DISTRIBUTION OF SHORTEST PATH TREES AND BIAS OF NETWORK SAMPLING ALGORITHMS DEGREE DISTRIBUTION OF SHORTEST PATH TREES AND BIAS OF NETWORK SAMPLING ALGORITHMS SHANKAR BHAMIDI 1, JESSE GOODMAN 2, REMCO VAN DER HOFSTAD 3, AND JÚLIA KOMJÁTHY3 Abstract. In this article, we explicitly

More information

This module is part of the. Memobust Handbook. on Methodology of Modern Business Statistics

This module is part of the. Memobust Handbook. on Methodology of Modern Business Statistics This moule is part of the Memobust Hanbook on Methoology of Moern Business Statistics 26 March 2014 Metho: Balance Sampling for Multi-Way Stratification Contents General section... 3 1. Summary... 3 2.

More information

LECTURE NOTES ON DVORETZKY S THEOREM

LECTURE NOTES ON DVORETZKY S THEOREM LECTURE NOTES ON DVORETZKY S THEOREM STEVEN HEILMAN Abstract. We present the first half of the paper [S]. In particular, the results below, unless otherwise state, shoul be attribute to G. Schechtman.

More information

Research Article When Inflation Causes No Increase in Claim Amounts

Research Article When Inflation Causes No Increase in Claim Amounts Probability an Statistics Volume 2009, Article ID 943926, 10 pages oi:10.1155/2009/943926 Research Article When Inflation Causes No Increase in Claim Amounts Vytaras Brazauskas, 1 Bruce L. Jones, 2 an

More information

Number of wireless sensors needed to detect a wildfire

Number of wireless sensors needed to detect a wildfire Number of wireless sensors neee to etect a wilfire Pablo I. Fierens Instituto Tecnológico e Buenos Aires (ITBA) Physics an Mathematics Department Av. Maero 399, Buenos Aires, (C1106ACD) Argentina pfierens@itba.eu.ar

More information

Multi-View Clustering via Canonical Correlation Analysis

Multi-View Clustering via Canonical Correlation Analysis Technical Report TTI-TR-2008-5 Multi-View Clustering via Canonical Correlation Analysis Kamalika Chauhuri UC San Diego Sham M. Kakae Toyota Technological Institute at Chicago ABSTRACT Clustering ata in

More information

Robustness and Perturbations of Minimal Bases

Robustness and Perturbations of Minimal Bases Robustness an Perturbations of Minimal Bases Paul Van Dooren an Froilán M Dopico December 9, 2016 Abstract Polynomial minimal bases of rational vector subspaces are a classical concept that plays an important

More information

Largest eigenvalues of sparse inhomogeneous Erdős-Rényi graphs

Largest eigenvalues of sparse inhomogeneous Erdős-Rényi graphs Largest eigenvalues of sparse inhomogeneous Erős-Rényi graphs Florent Benaych-Georges Charles Borenave Antti Knowles April 26, 2017 We consier inhomogeneous Erős-Rényi graphs. We suppose that the maximal

More information

Optimization of Geometries by Energy Minimization

Optimization of Geometries by Energy Minimization Optimization of Geometries by Energy Minimization by Tracy P. Hamilton Department of Chemistry University of Alabama at Birmingham Birmingham, AL 3594-140 hamilton@uab.eu Copyright Tracy P. Hamilton, 1997.

More information

Selection principles and the Minimal Tower problem

Selection principles and the Minimal Tower problem Note i Matematica 22, n. 2, 2003, 53 81. Selection principles an the Minimal Tower problem Boaz Tsaban i Department of Mathematics an Computer Science, Bar-Ilan University, Ramat-Gan 52900, Israel tsaban@macs.biu.ac.il,

More information

On the enumeration of partitions with summands in arithmetic progression

On the enumeration of partitions with summands in arithmetic progression AUSTRALASIAN JOURNAL OF COMBINATORICS Volume 8 (003), Pages 149 159 On the enumeration of partitions with summans in arithmetic progression M. A. Nyblom C. Evans Department of Mathematics an Statistics

More information

A note on asymptotic formulae for one-dimensional network flow problems Carlos F. Daganzo and Karen R. Smilowitz

A note on asymptotic formulae for one-dimensional network flow problems Carlos F. Daganzo and Karen R. Smilowitz A note on asymptotic formulae for one-imensional network flow problems Carlos F. Daganzo an Karen R. Smilowitz (to appear in Annals of Operations Research) Abstract This note evelops asymptotic formulae

More information

NOTES ON EULER-BOOLE SUMMATION (1) f (l 1) (n) f (l 1) (m) + ( 1)k 1 k! B k (y) f (k) (y) dy,

NOTES ON EULER-BOOLE SUMMATION (1) f (l 1) (n) f (l 1) (m) + ( 1)k 1 k! B k (y) f (k) (y) dy, NOTES ON EULER-BOOLE SUMMATION JONATHAN M BORWEIN, NEIL J CALKIN, AND DANTE MANNA Abstract We stuy a connection between Euler-MacLaurin Summation an Boole Summation suggeste in an AMM note from 196, which

More information

Implicit Differentiation

Implicit Differentiation Implicit Differentiation Thus far, the functions we have been concerne with have been efine explicitly. A function is efine explicitly if the output is given irectly in terms of the input. For instance,

More information

. Using a multinomial model gives us the following equation for P d. , with respect to same length term sequences.

. Using a multinomial model gives us the following equation for P d. , with respect to same length term sequences. S 63 Lecture 8 2/2/26 Lecturer Lillian Lee Scribes Peter Babinski, Davi Lin Basic Language Moeling Approach I. Special ase of LM-base Approach a. Recap of Formulas an Terms b. Fixing θ? c. About that Multinomial

More information

How to Minimize Maximum Regret in Repeated Decision-Making

How to Minimize Maximum Regret in Repeated Decision-Making How to Minimize Maximum Regret in Repeate Decision-Making Karl H. Schlag July 3 2003 Economics Department, European University Institute, Via ella Piazzuola 43, 033 Florence, Italy, Tel: 0039-0-4689, email:

More information