How To Solve The Sum Of Square Roots Problem Over Polynomals

Size: px
Start display at page:

Download "How To Solve The Sum Of Square Roots Problem Over Polynomals"

Transcription

1 On the Sum of Square Roots of Polynomals and Related Problems Neeraj Kayal Chandan Saha Abstract The sum of square roots problem over ntegers s the task of decdng the sgn of a non-zero sum, S = n =1 δ a, where δ {+1, 1} and a s are postve ntegers that are upper bounded by N (say). A fundamental open queston n numercal analyss and computatonal (n log N)O(1) geometry s whether S 1/2 when S 0. We study a formulaton of ths problem over polynomals: Gven an expresson S = n =1 c f (x), where c s belong to a feld of characterstc 0 and f s are unvarate polynomals wth degree bounded by d and f (0) 0 for all, s t true that the mnmum exponent of x whch has a nonzero coeffcent n the power seres S s upper bounded by (n d) O(1), unless S = 0? We answer ths queston affrmatvely. Further, we show that ths result over polynomals can be used to settle (postvely) the sum of square roots problem for a specal class of ntegers: Suppose each nteger a s of the form, a = X d + b 1 X d b d, d > 0, where X s a postve real number and b j s are ntegers. Let B = max({ b j },j, 1) and d = max {d }. If X > (B + 1) (n d)o(1) then a non-zero S = n =1 δ a s lower bounded as S 1/X (n d)o(1). The constant n the O(1) notaton, as fxed by our analyss, s roughly 2. We then consder the followng more general problem: gven an arthmetc crcut computng a multvarate polynomal f(x) and nteger d, s the degree of f(x) less than or equal to d? We gve a corp PP -algorthm for ths problem, mprovng prevous results of Allender, Bürgsser, Kjeldgaard-Pedersen and Mltersen (2009), and Koran and Perfel (2007). A prelmnary verson of ths paper appeared n the Proceedngs of the 26th IEEE Conference on Computatonal Complexty (CCC), Mcrosoft Research Inda, neeraka@mcrosoft.com Max Planck Insttute for Informatcs, csaha@mp-nf.mpg.de 1

2 1 Introducton The sum of square roots s the followng well-known problem: gven a set {a 1, a 2,..., a n } of postve ntegers and {δ 1, δ 2,..., δ n } { 1, +1} n, determne the sgn of the sum n δ a (1) =1 It was posed as an open problem by Garey, Graham and Johnson [GGJ76] n connecton wth the Eucldean travellng salesman problem. Eucldean TSP s not known to be n NP but s easly seen to be n NP relatve to the sum of square roots problem. The sum of square roots problem also arses naturally n some other problems nvolvng computatonal geometry (cf. [MR08]) and semdefnte progammng (cf. the survey by Goemans [Goe98]). Although t has been conjectured [Mal96] that the problem les n P, the best known result so far [ABKPM09] s contanment n the countng herarchy CH, whch s a subclass of PSPACE that contans the polynomal herarchy PH. For the related but easer problem of determnng whether the sum (1) s zero or not, a determnstc polynomal tme algorthm s known due to the work of Blömer [Blö91]. (The conference verson of Blömer s paper [Blö91] contans a randomzed polynomal tme algorthm, whch was later derandomzed n [Blö93]). A possble approach towards answerng the sum of square roots queston s to solve the followng number-theoretc problem whose current status s stll a conjecture. Problem 1.1 (Lower boundng a non-zero sgned sum of square roots of ntegers). Gven a sum S = n =1 δ a, where δ {+1, 1} and a s are postve ntegers upper bounded by N, fnd a tght lower bound on S n terms of n and N when S 0. Is t true that for a non-zero S, S 1/2 poly(n,log N) for some fxed polynomal poly( )? If the answer to the above queston s yes then computng the square roots up to poly(n, log N) bts of precson suffces to determne the sgn of the sum of square roots. Now, t s known (cf. [Bre76]) that square root of a number less than N can be computed up to m bts of precson n poly(m, log N) tme, whch together wth an affrmatve answer to the above queston would mply that the sum of square roots problem s n P. In fact, a stronger result s known n ths regard: t was shown by Ref and Tate [RT92] that TC 0 crcuts of polynomal sze can compute both terated sums and square roots up to polynomal bts of precson, whch means that the sum of square roots problem can even be shown to be n the complexty class TC 0 f we manage to solve Problem 1.1 postvely. Reducng the mmense gap between the known upper and lower bounds for Problem 1.1 s a challengng number-theoretc problem. Now, there s a well known analogy between ntegers and polynomals (cf. [EHM05]). We refer the reader to a survey by Immerman and Landau [IL93] for some algorthmc aspects of ths analogy. Whle the polynomal analog of a number-theoretc problem s typcally much easer than the nteger verson, nevertheless the nvestgaton of the polynomal verson can on occasson gve some nsght nto the nteger problem tself. For example, Allender et al. [ABKPM09] proved a hardness result of a closely related problem, whch they call BtSLP, by frst observng that the correspondng problem for polynomals s #P-hard. Ths motvates us to examne the polynomal analogue of the sum of square roots problem, the precse statement of whch s gven below as an nterestng problem n ts own rght. Problem 1.2 (Sum of square roots of polynomals). Gven an expresson S = n =1 c f (x), where c F (a feld of characterstc 0) and f (x) are unvarate polynomals wth degree bounded by d 2

3 and f (0) 0 (for all ), can we show that unless S = 0, the mnmum exponent of x whch has a non-zero coeffcent n the power seres S s bounded by a fxed polynomal n n and d? The condton f (0) 0 n the above problem statement s smply to ensure that f(x) has a well defned power seres expanson around x = 0. Ths problem beng a close cousn of Problem 1.1, t seems reasonable to hope that solvng t mght shed some lght on the latter problem. At the least, one mght expect to solve Problem 1.1 for a nontrval class of ntegers startng from a soluton to Problem 1.2. We are not aware of any pror research work along ths lne. In ths work, we answer the queston posed n Problem 1.2 n the affrmatve. Usng ths result, we show that t s ndeed suffcent to keep polynomal amount of precson n computng the sgn of S n Problem 1.1 f the nput ntegers belong to a specal class that we call (by abusng termnology) the polynomal ntegers. We have mentoned earler that the sum of square roots problem les n the countng herarchy: CH = P P PP P PPPP..., where PP s the class of languages accepted by probablstc polynomal tme machnes. Ths result s due to Allender, Bürgsser, Kjeldgaard-Pedersen and Mltersen [ABKPM09]. In fact, they showed that the more general problem PosSLP, whch s the task of checkng f the nteger produced by a gven dvson-free straght-lne program s greater than zero, belongs to the complexty class P PPPPPP, the thrd level of CH. (The class P s taken to be the 0-th level of CH). The fact that PosSLP s a more general problem than the sum of square roots of ntegers follows from the work of Twar [Tw92] who used Newton Iteraton to construct a small (polynomal-szed) arthmetc crcut that computes a gven sum of square roots to exponentally many dgts of precson. The polynomal analog of the PosSLP problem s the task of comparng the degree of the polynomal computed by a gven arthmetc crcut wth a gven nteger. More precsely: Problem 1.3 (Degree Computaton). Let F be a feld (say the ratonal numbers Q). Gven an arthmetc crcut computng a multvarate polynomal f(x) over F and an nteger d, s the degree of f(x) at most d? Ths problem, whch Allender et al. [ABKPM09] refer to as DegSLP, was also studed by Koran and Perfel [KP07] and they put t n the second level of the countng herarchy. Here we gve a (slght) mprovement to the complexty theoretc upper bound for DegSLP. We show ts contanment n the class corp PP, whch s essentally the frst level of CH modulo the use of randomzaton. 1.1 Prevous work The work of Burnkel, Flescher, Mehlhorn and Schrra [BFMS00] consdered the problem of fndng the sgn of an arthmetc expresson E nvolvng the operatons addtons, subtractons, multplcatons and square root (n fact, dvson as well), and wth nteger operands. They showed that f u s the bound on the value of E when all the subtracton operatons n E are replaced by addtons and k s the number of dstnct square root operatons n E then E 1/u 2k 1 unless E = 0. Ths result mmedately gves an exponental bound on the bt sze of S n Problem 1.1: f S 0 then S 1/2 2n log(nn). (In ths regard, the work of Mehlhorn and Schrra [MS00] s also relevant). It s also noted n [BFMS00] that the bound obtaned for E s nearly optmal n general. For nstance, f E = (2 2k + 1) 1/2k 2 then u = (2 2k + 1) 1/2k and hence by the result n [BFMS00], E 1/5 2k 1. On the other hand, t was also shown that E 1/2 2k. However, Problem 1.1 s just a specal case of the problem studed n [BFMS00] where there s no occurrence of nested square 3

4 roots. So, t remans a concevable possblty that there s a better lower bound for S. Indeed, for a certan choce of parameters a better result s known due to the work of Cheng, Meng, Sun and Chen [CMSC10] (see also [Che06]). By connectng Problem 1.1 to the shortest vector problem over a certan nteger lattce, they showed that S 1/N 2O(N/ log N), whch s an mprovement over earler results when N c n log n for some constant c. However, a less desrable aspect of ths result s the doubly exponental dependency of the bt complexty of S on log N. On the other sde, one seeks to construct/prove the exstence of sets of ntegers {a 1, a 2,..., a n } for whch the sum S of Problem 1.1 s as small as possble n absolute value. In other words, what could be a good upper bound on S. Qan and Wang [QW06] gave an explct constructon that gves S = O(N n+3/2 ). The ntegers they construct are closely related to the specal class of ntegers that we look at (see Secton 1.2). For Qan and Wang, every nteger a s essentally of the form (X + ) (sutably scaled). They showed that f S = n =0 ( n ) ( 1) X + then S = O(X n+1/2 ) for fxed n 1. The complexty of the DegSLP problem was studed by Allender et al. [ABKPM09], who showed that DegSLP polynomal tme many-one reduces to PosSLP and hence also belongs to the thrd level of CH. Koran and Perfel [KP07] later mproved upon ther result over felds of characterstc p > 0 and obtaned a better bound of co-np ModpP, whch s contaned n the second level of CH. However, over felds of characterstc zero, the thrd level of CH remaned the best known upper bound on the complexty of DegSLP. 1.2 Our contrbuton Our frst contrbuton s an affrmatve answer to Problem 1.2. Usng ths, we prove the followng theorem. Theorem 1.4 (Sum of square root of polynomal ntegers ). Suppose S = n =1 δ a (δ {+1, 1}) such that every postve nteger a s of the form a = X d + b 1 X d b d (d > 0), where X s a postve real number and b j are ntegers. Let B = max({ b j },j, 1) and d = max {d }. If X > (B + 1) p1(n,d), where p 1 (n, d) s a fxed polynomal n n and d, then a non-zero S s lower bounded as, S 1/X p2(n,d), where p 2 (n, d) s another fxed polynomal n n and d. The polynomals p 1 (n, d) and p 2 (n, d) can be taken to be 12 dn 2 log 32d and 6 dn 2, respectvely. Note that the ntegers b j need not be postve. Expressng each a as X d + b 1 X d b d s nothng very unusual - t s lke a base-x representaton of a when X s a postve nteger. What makes the polynomal ntegers specal s the condton that X s exponentally large compared to the b j s; or n other words, all the dgts are small n X-ary representaton. Indeed, f one can prove Theorem 1.4 wthout ths condton then Problem 1.1 would stand solved n ts full generalty by takng X = 2. Fnally, we would lke to note that we have not made an attempt to fnd the best possble expressons for p 1 ( ) and p 2 ( ), our prmary ntenton beng to just show that the functons p 1, p 2 are some fxed polynomals n n and d. For the more general DegSLP problem, we show contanment n the frst level of the countng herarchy (modulo the use of randomzaton), thereby mprovng upon prevous best results [ABKPM09, KP07] for ths problem. More precsely, we show Theorem 1.5. DegSLP s n corp PP. 4

5 The above theorem holds over any underlyng feld. Organzaton - The rest of ths paper s organzed as follows. In Secton 2 we gve a soluton to Problem 1.2 and n Secton 3 we prove Theorem 1.4. The result on the complexty upper bound of DegSLP s presented n Secton 4. 2 Sum of square roots of polynomals In ths secton, we prove the followng theorem. Theorem 2.1. Gven a sum S = n =1 c g (x) f (x) where c F (a feld of characterstc 0), f and g are unvarate polynomals of degree at most d and f (0) 0 for all 1 n, ether S = 0 or the mnmum exponent of x whch has a nonzero coeffcent n the power seres S s bounded by dn 2 + n. The soluton to Problem 1.2 follows mmedately f we take g (x) to be 1 n the above theorem. But, we wll need the slghtly general form, that s, when the g s are not assumed to be 1, to prove Theorem 1.4. We wll see that the proof of Theorem 2.1 s not dffcult n hndsght - t uses a mathematcal object called the Wronskan, whch s used to study lnear dependence of functons. Let us spend some tme to brefly dscuss ths concept. 2.1 The Wronskan and lnear ndependence A set of n functons {h 1, h 2,..., h n } over a feld F s sad to be lnearly dependent f there exst elements c 1, c 2,..., c n F such that the functon (c 1 h 1 + c 2 h c n h n ) s dentcally zero. If each of the h s s n tmes dfferentable, then the Wronskan of ths set, denoted W(h 1,..., h n ) (or, W(h) for short) s defned as the followng determnant. h 1 h 2... h n W(h 1,..., h n ) def h (1) = det 1 h (1) 2... h (1) n.... h (n 1) 1 h (n 1) 2... h (n 1) n where h (j) s the j th dervatve of h. It s a well known functon used n the study of dfferental equatons. It s easy to observe that f the functons h 1,..., h n are F-lnearly dependent then ther Wronskan s dentcally zero. But, the converse need not be true n general. Bôcher [Bˆ00] showed that there are famles of nfntely dfferentable functons whch are lnearly ndependent and yet ther Wronskan vanshes dentcally. However, for analytc functons ths s not the case: a fnte famly of lnearly ndependent (real or complex valued) analytc functons has a non-zero Wronskan. More generally, ths property s true for any famly of formal power seres over any characterstc zero feld. Theorem 2.2 (Wronskan of a famly of power seres). Let F be a feld of characterstc zero. A fnte famly of power seres n F[[x]] has a zero Wronskan f and only f t s F-lnearly dependent. A short and smple proof of the above fact appears n [BD10]. Let us now see how to use ths result to prove Theorem 2.1., 5

6 2.2 Proof of Theorem 2.1 Let S = n =1 c g (x) f (x) be a gven non-zero sum. Assume wthout loss of generalty that f (0) = 1, for all. If ths s not the case then scale f (x) approprately: Express f (x) as f (x) = f (0) f (x)/f (0) and take f (x)/f (0) to be our new f (x) (by reusng symbols). Also, consder an approprate extenson feld that contans f (0) for every 1 n, ths extenson s our base feld F (agan, by reusng symbols). The condton f (0) = 1 s smply to ensure that f can be expressed as a formal power seres n x over F. Denote g f by h. We can also assume that h 1,..., h n are F-lnearly ndependent - f not, smply work wth an F-bass of h 1,..., h n. (Note that, we are not fndng a bass, we are only usng t for the sake of argument.) Suppose, x t dvdes the power seres S, where t s the maxmum possble. Pretend that, n c h = S (2) =1 s a lnear equaton n the varables c 1,..., c n. By takng dervatves of both sdes of Equaton 2 wth respect to x, we have the followng system of lnear equatons n c 1,..., c n, for 0 j n 1, n =1 c h (j) = S (j), (3) where h (j) and S (j) are the j th dervatves of h and S, respectvely. Let C be the coeffcent matrx of the above system of lnear equatons. That s, h 1 h 2... h n h (1) C = 1 h (1) 2... h (1) n..... h (n 1) 1 h (n 1) 2... h (n 1) n Observe that det(c) s the Wronskan W(h). The followng smple clam about W(h) s crucal to the proof. Clam 2.3. The Wronskan W(h) = n 2 det(m), where M s an n n matrx whose every entry s a polynomal n x of degree at most n d. =1 f 2n 3 Proof. Expandng h (j) we get the followng. (Superscrpts ndcate the order of the dervatves.) h (j) = j k=0 ( ) j k g (j k) ( f ) (k) f 2j 1 2 h (j) = j k=0 ( ) j g (j k) k f 2j 1 2 ( f ) (k), multplyng both sdes by f (2j 1)/2. Now notce that, f (2j 1)/2 ( f ) (k) s a polynomal of degree at most j d. Hence, f (2j 1)/2 h (j) s also a polynomal of degree at most (j+1) d, although ndvdually they are power seres n x. Snce j s at max n 1, the statement of the clam follows. Snce S 0, there must be one c whch s nonzero. Let t be c 1. Then, by applyng Cramer s rule, c 1 = det(m 1) W(h) = n =1 f 2n 3 2 det(m 1) det(m) 6 (by Clam 2.3),

7 where M 1 s the followng matrx, S h 2... h n S M 1 = (1) h (1) 2... h (1) n.... S (n 1) h (n 1) 2... h (n 1) n Note that, Cramer s rule apples here because W(h) 0, as h 1,..., h n are assumed to be lnearly ndependent, whch n turn mples that det(m) 0 (by Clam 2.3). Snce x t dvdes S, x t n+1 must dvde S (j) for every 0 j n 1 and hence x t n+1 dvdes det(m 1 ). Clam 2.4. The maxmum power of x dvdng det(m 1 ) and det(m) must be the same. Proof. Ths s because c 1 0 s an element of the feld F and n =1 f (0) 2n by assumpton. Therefore, t n must be less than the degree of det(m), whch s at most d n 2 (agan by Clam 2.3), and hence t d n 2 + n. Ths proves Theorem 2.1. Wth the polynomal verson of the sum of square roots problem at hand, one wonders as to what can be nferred about the correspondng problem over ntegers. In turns out that ndeed somethng nontrval can be shown about a specal class of ntegers that we have called before as the polynomal ntegers (see Theorem 1.4). Ths consttutes the content of the followng secton.. 3 Sum of square roots of polynomal ntegers Ths secton s devoted to the proof of Theorem 1.4. Let S = n =1 δ a ; δ {+1, 1}, be a gven nonzero sum, where each postve nteger a s of the followng form. a = X d + b 1 X d b d (4) where X s a postve real number and b j s are ntegers (not necessarly postve). Snce X d = ( X) 2d, by takng X to be our new X, we can assume henceforth that every d (1 n) s an even number. Overvew. The overall dea of the proof s to do a Taylor seres expanson for each a so that we get a Taylor expanson for the sum S overall. Usng Theorem 2.1 and the non-zeroness of S, we deduce that we must get a non-zero term very early n the Taylor expanson. That s, there must be some non-zero S l for l relatvely small (see Equatons 5 and 6 below for a precse defnton of S l ). We use the fact that l s small to deduce that such an S l s farly large n absolute value. We then use the fact that each b j s much smaller than X to upper bound each of the remanng terms and thereby deduce that the sum of the remanng terms cannot almost cancel out S l. More specfcally, the sum of the remanng terms s at most 1 2 S l n absolute value. Ths helps us deduce that S tself s farly large n absolute value. Dong the Taylor expanson. From (4) we have a = ( X) d 1 + b 1 X b d X d. 7

8 Addng these expressons together wth the approprate sgn, we get an expresson for S. n S = δ X d /2 1 + b 1 X b d X d. =1 Let y = 1/X and d = max {d }. Then, n S y d/2 = δ y (d d )/2 1 + (b 1 y) (b d y d ). =1 Now notce that, by pretendng that f (y) = 1 + b 1 y b d y d s a polynomal n the formal varable y, the sum S y d/2 s of the form n =1 δ g (y) f (y), where g (y) = y (d d)/2 (snce every d s even, by the adjustment dscussed before, (d d )/2 s an nteger). Therefore, n S y d/2 = δ n y j δ c j (by exchangng the summatons), =1 j 0 c j y j = j 0 =1 where c j s the coeffcent of y j comng from the th power seres g (y) f (y). Ths s the Taylor seres expanson of S that we wll work wth. We gve the name S j to each summand n the expresson above, namely S j = y j n =1 δ c j. Thus S y d/2 = j 0 S j. (5) The Proof Strategy ahead. Applyng Theorem 2.1, the mnmum exponent l of y wth n =1 δ c l 0 s such that l dn 2 + n. Suppose we could show that S l+t S l 1, (6) 2t+1 for every t 1, then from (5) t would follow that S y d/2 Sl S l+1... S l 1 2 S l = 1 2 S l S X d/2 1 2 S l. (7) Ths (potentally) gves us a lower bound on S va a lower bound of S l. But, to satsfy the condton gven by equaton (6), we also need an upper bound on S l+t for every t. Upper bound on S j : S j = y j n δ c j n y j max{ c j } =1 Let us upper bound the quantty c j. Fx any ndex. For the ease of presentaton, we wll avod wrtng the ndex whenever t s clear from the context that we have a specfc n mnd. For example, we wrte c j as the coeffcent of y j comng from the power seres, y d d b 1 y b d y d d d = y 2 8 u k (b 1 y b d y d ) k, k=0

9 where u 0 = 1 and u k = 1 2 ( 1 2 1)...( 1 2 (k 1)) k! = ( 1) k (2(k 1) 1) 2 k k!, for k > 0. Expressed dfferently, ( 2k ) u k = ( 1) k+1 (2k)! (2k 1) (k!) 2 2 2k u k k = (2k 1) 2 2k Now, notce that ( ) 2k k /(2k 1) = 2 Ck 1, where C k s the k th Catalan number 1/(k + 1) (2k ) k. Hence, u k = 2 C k 1 2 2k, and also u k 1. (8) For any j, the coeffcent c j s contrbuted to by those terms of k=0 u k (b 1 y b d y d ) k for whch k s n the range [(j (d d )/2)/d, j (d d )/2]. For any fxed k [j/d, j], the coeffcent of y j n (b 1 y b d y d ) k s exactly, ( ) k v kj = b k 1 1 bk bk d d. k 1 +2k d k d =j, k k d =k k 1,..., k d where k 1,..., k d are non-negatve ntegers. Then, assumng B = max({ b j },j, 1), ( ) k v kj b 1 k1 b 2 k 2... b d k d (B d ) k. k 1,..., k d k k d =k Snce c j+(d d )/2 = k [j/d,j] u k v kj, Lower bound on S j : c j k [0,j] (B d ) k (B d) j+1 (usng Equaton 8) S j n y j (B d) j+1 (9) S j = y j n δ c j Let us lower bound the sum n =1 δ c j. Notce that, n the prevous dscusson on upper boundng S j, the nteger v kj depends on the ndex whereas u k solely depends on k. So, to make the followng dscusson more precse, we swtch to the notaton v kj. For smplcty, the range of k s not specfed n the followng equatons - the approprate range of k s [(j (d d )/2)/d, j (d d )/2]. =1 n δ c j = =1 n δ u k v k(j (d d )/2) = k k =1 n u k δ v k(j (d d )/2). =1 Now notce that, the sum n =1 δ v k(j (d d )/2) s an nteger. Hence, f n =1 δ c j 0 then n =1 δ c j 1/2 2j 1 (by Equaton 8). Therefore, S j yj 2 2j 1 f S j 0. (10) 9

10 Puttng everythng together. Wth the upper and the lower bounds on S j at hand, we are now ready to pnpont the requrement that ensures that condton (6) s satsfed. We want S l+t / S l 1/2 t+1, for all t 1. Hence, by combnng equatons (9) and (10), t s suffcent f, n y l+t (B d) l+t+1 y l /2 2l t+1 X t n 2 2l+t (B d) l+t+1 Therefore, t suffces f X (B + 1) 3 dn2 log 16d (takng nto consderaton that l dn 2 + n). And f ths happens then condton (6) s satsfed and by equaton (7), S X d/2 1 2 S l 1 2 2l X l d/2 (by Equaton 10) Ths mples that S 1/X 6 dn2. Recall that at the begnnng of ths secton, we have adjusted X (by choosng X to be our X) so that d can be assumed to be even whch n turn s used to argue that (d d )/2 s n fact an nteger. Replacng X n place of X and 2d n place of d n the above bounds on X and S, we obtan the bounds X (B + 1) 12 dn2 log 32d and S 1/X 6 dn2 as promsed n Theorem The complexty of DegSLP In ths secton, we consder the algorthmc complexty of the followng problem: gven a polynomal f(x) as an arthmetc crcut, and an nteger d n bnary determne f deg(f) d. Towards ths end, we need to defne another natural computatonal problem. CoeffSLP : Gven an arthmetc crcut computng a polynomal f(x) over ntegers, a monomal X α and a prme p, determne the coeffcent of X α n f(x) modulo p. (The addtonal nput p s to ensure that the output s not too large). Our defnton of problem CoeffSLP s slghtly dfferent from that of [ABKPM09, KP07] and more talored towards our needs. We wll need the followng theorem from [KP07] to prove our result. A smpler, self-contaned proof of the theorem s gven below. Theorem 4.1. [KP07] CoeffSLP s #P-complete. We frst gve some warm-up lemmas. Lemma 4.2. For any m 7, the lcm of the frst m numbers s at least 2 m. Lemma 4.3. For any nteger t Z 1 and prme p, there s a prme r = O(t 2 log p) such that the rng R def = F p [z]/ zr 1 z 1 s the drect sum of fnte felds of sze q > p t. 10

11 Proof. Consder the nteger M := (p 1) (p 2 1)... (p t 1). Then M < p t2. By lemma 4.2, there exsts a prme r < log M such that r does not dvde M. Ths s the prme r that we seek. Let m denote the order of p modulo r,.e. m s the smallest postve nteger such that p m = 1 (mod r). Snce r does not dvde M = [t] (p 1), therefore r does not dvde any (p 1) for 1 t and therefore m > t. Let φ r (z) denote the r-th cyclotomc polynomal, that s φ r (z) def = zr 1 z 1. It s known (cf. Theorem 2.47 of [LN94]) that over F p, φ r (z) factors nto r 1 m rreducble polynomals each of degree m. Thus, R def = F p [z]/ φ r (z) s the drect sum of fnte felds of sze p m > p t. Our next two warmup lemmas pertan to the coeffcent of a monomal x d of a gven polynomal f(x), denoted Coeff(x d, f(x)). The frst lemma s vald when the characterstc of the feld s larger than d and expresses the coeffcent of x d n any shft of f(x) as the evaluaton of the d-th order dervatve of f at a sutable pont. When the sze of the underlyng feld s large enough, our second lemma below expresses Coeff(x d, f(x)) as an exponental sum (wth each summand beng computable n polynomal-tme). Ths wll help us n showng that the computaton of Coeff(x d, f(x)) s n #P. Lemma 4.4. Over a feld of characterstc larger than d, Coeff(x d, f(x + β)) (the coeffcent of x d n f(x + β)) s precsely 1 d! f (d) (β). Proof. By the lnearty of dervatves and substtutons, t s suffcent to show ths for monomals. So let f(x) = a x e. If e < d then f (d) (x) s the zero polynomal and we are done. So let e d. Expandng (x + β) e usng bnomal theorem we get that coeffcent of x d s a (e d) β e d. On the other hand f (d) (x) = a e (e 1)... (e d + 1)x e d. It s now easly verfed that the coeffcent of x d n f(x + β) s 1 d! f (d) (β). Lemma 4.5. For any fnte feld F q and any polynomal f(x) F q [x], we have β d f(β 1 ) = β F q In partcular f (q 1) > deg(f) then d (mod (q 1)) Coeff(x d, f(x)) = β F q Coeff(x, f(x)). β d f(β 1 ). Proof. The proof s a routne computaton n the manner of dong Fourer transforms. In partcular we use the fact that for any nteger k we have { β k 1 f k 0 (mod (q 1)) = (11) 0 otherwse β F q 11

12 Consder the lhs of the dentty we wsh to prove. We have β d Coeff(x, f(x)) β 0 β d f(β 1 ) = β F q β F q = Coeff(x, f(x)) β d β F q 0 = Coeff(x, f(x)) β d 0 β F q = 0 Coeff(x, f(x)) β F q β d = 0 Coeff(x, f(x)) ( 1) δ (d ) 0 (mod (q 1)) (usng (11)) = d (mod (q 1)) Coeff(x, f(x)) As a corollary we get an exact expresson for Coeff(x d, f(x)) as an exponental when the sze of the underlyng fnte feld s larger than the degree of f. Corollary 4.6. Let F q be a fnte feld and f(x) F q [x] be a polynomal. Suppose that d deg(f) < (q 1). Then Coeff(x d, f(x)) = β d f(β 1 ). β F q Note that ths corollary mples that when the sze of the underlyng fnte feld s large enough Coeff(x d, f(x)) can be computed n P #P. To extend ths for arbtrary fnte felds, we need to go to a sutably large extenson of the underlyng fnte feld. However constructng extensons of felds determnstcally s not known but we get around the problem by gong to a sutably rng extenson and ths suffces for our purpose. We collect all ths together nto a proof of theorem 4.1 below. Proof of Theorem 4.1: The #P-hardness of ths problem s well-known and a proof can be found for example n [ABKPM09]. It s suffcent to show ths for unvarate polynomals (by replacng each ndetermnate x by an exponentally ncreasng sequence of monomals, f necessary). That s, our problem now becomes the followng: gven a crcut of sze s computng a unvarate polynomal f(x) and a d Z 0 gven n bnary, compute the coeffcent of x d n f(x). Notce that D def = 2 s s an upper bound on deg(f(x)). Usng lemma 4.3, we obtan an extenson rng R of the form R = F p [z]/ zr 1 z 1 such that r (log D)2 (log p) and R = F q... F q, wth q 1 > D. Combnng ths structure of the rng R wth corollary 4.6 va Chnese remanderng we have Coeff(x d, f(x)) = β R β d f(β 1 ). 12

13 The number of terms n the above summaton s exponentally large but notce that each summand n the above expresson, (β d f(β 1 )), s polynomal-tme computable so that overall ths sum s computable n P #P. Theorem 4.7. DegSLP corp T CoeffSLP. Here we frst prove ths theorem assumng the underlyng feld to be the feld of ratonal numbers. A smlar proof wll go through over other felds of zero characterstc. The theorem s vald even f the underlyng feld has small characterstc. We relegate the proof for small characterstc felds to Secton 4.1. Proof. We frst reduce the multvarate to the unvarate problem by makng a random substtuton of the form g(z) = f(a 1 z, a 2 z,..., a n z). The unvarate verson of DegSLP s n fact equvalent (va determnstc reductons as well) to the multvarate verson. For a proof cf. [ABKPM09]. Clam 4.8. Wth hgh probablty over a random choce of the vector a = (a 1,..., a n ) we have: deg(g(z)) = deg(f(x)). Proof of Clam 4.8: Let f have degree d. We can wrte the polynomal f(x) as d =0 f (X), where each f (X) s a homogeneous polynomal of degree. Applyng the substtuton x := a z, we get that g(z) = f 0 + z f 1 (a) + z 2 f 2 (a) z d f d (a). By the Schwartz-Zppel lemma, f d (a) s non-zero wth hgh probablty so that deg(g(z)) = deg(f(x)) also wth hgh probablty. Our problem thus s the followng: Gven an arthmetc crcut computng a unvarate polynomal g(z) and an nteger d n bnary, we want to determne f deg(g(z)) d. The most natural thng to do s to use the CoeffSLP oracle to determne whether the coeffcent of z d n g(z) s nonzero. (For the oracle call to CoeffSLP, we choose the prme p at random. Ths ensures that wth hgh probablty, the coeffcent α of z d n g(z) s zero as a ratonal number f and only f α s zero modulo p.) If the coeffcent of z d n g(z) happens to be nonzero we have a certfcate that the degree of g s at least d. The converse s easly seen to be false: the coeffcent of z d n g can be zero and yet the degree of g can be larger than d. To fx ths, we take a random shft of g and compute ts degree nstead. Specfcally, we look at the polynomal g(z + β), where β s chosen unformly at random from a large enough subset of F. Clearly, deg(g(z)) = deg(g(z + β)). It suffces then to prove the followng clam: Clam 4.9. The coeffcent of z d n g(z + β) s zero for more than (deg(g) d) choces of β F f and only f deg(g(z)) < d. Proof of Clam 4.9: By lemma 4.4, the coeffcent of z d n g(z + β) s 1 d! g(d) (β), where g (d) (z) denotes as usual the d-th order dervatve of g. Snce the characterstc of the feld s zero, g (d) (z) has degree precsely deg(g) d. In partcular, g (d) (z) s dentcally zero f and only f deg(g) < d. Now the clam follows from an applcaton of the Schwarz-Zppel lemma. Ths completes the proof of the theorem. Combnng Theorems 4.1 and 4.7, we mmedately get: 13

14 Theorem DegSLP s n corp PP for felds of characterstc zero. 4.1 DegSLP over felds of small characterstc In ths secton, we gve the proof of Theorem 4.7 n the general case,.e even when the underlyng feld has small characterstc. We wll need a lemma orgnally due to Edouard Lucas. Lemma [vl99, p.55] Let n, m be postve nte gers whose p-ary representaton s the followng: Then m = m 0 + m 1 p m d p d, : 0 m p 1 n = n 0 + n 1 p n d p d : 0 n p 1. ( ) n = m ( n0 m 0 ) ( n1 m 1 )... ( nd m d ) (mod p) In partcular, for any ntger n 1, the bnomal coeffcent ( n p ) s dvsble by p f and only f the n, the -th dgt n the p-ary representaton of n s zero. Proof of Theorem 4.7: The proof of theorem 4.7 needs to be refned carefully for the low characterstc stuaton because clam 4.9 s no longer true. As before t does hold that deg(g(z)) = deg(g(z + β)) for every β F. In one drecton the clam does ndeed hold true over any feld. Indeed f deg(g(z)) < d then the coeffcent of z d n g(z + β) s zero for every β F. The converse drecton s unfortunately not true. The dea s to prove that a partal converse of clam 4.9 and use t to derve a randomzed reducton from DegSLP to CoeffSLP. Clam Let F be a feld of characterstc p, F ts algebrac closure and g(z) F[z] be a polynomal. Let t 0 be an nteger. If the coeffcent of z pt n g(z + β) s zero for more than (deg(g(z)) p t ) dstnct choces of β F then ether 1. deg(g(z)) < p t or, 2. deg(g(z)) p t+1 Proof of Clam 4.12: g(z + β) s gven by Let the coeffcent of z n g(z) be a F. Then the coeffcent of z d n h(β) = d m deg(g) a m ( ) m β m d. d Note that h(β) s a polynomal functon of β of degree at most (deg(g) d). Snce h(β) s zero for more than (deg(g) d) values of β F t means that h(β) must be dentcally zero as a polynomal functon of β. Thus a m ( ) m = a m d Now from Lucas s lemma 4.11 t follows that ( ) m p t 0 m [p t..(p t+1 1)]. ( ) m p t = 0 m d. (12) 14

15 Combned wth equaton (12) t means that a m = 0 m [p t..(p t+1 1)]. Ths means that ether deg(g(z)) < p t or that deg(g(z)) p t+1, provng the clam. Ths partal converse allows us to modfy the earler proof sutably n the small characterstc stuaton as follows. Proceedng as before, we can assume wthout loss of generalty that the gven crcut computes a unvarate polynomal g(z). Our problem then s the followng: gven an arthmetc crcut computng a unvarate polynomal g(z) and an nteger d n bnary, we want to determne f deg(g(z)) d. Frst observe that multplyng g(z) wth a sutable power of z, we may assume wthout loss of generalty that d s a power of p, say d = p t. Let the sze of the crcut computng f be s. Then the formal degree of f s bounded by 2 s. Let r = 1 + s log 2 log p. If the underlyng feld s fnte we can go to a sutable extenson n randomzed polynomal tme and thereby ensure that F p 2r. Wth ths preprocessng behnd us, let us see how the partal converse of clam 4.12 allows us to test f deg(g(z)) p t usng an oracle for CoeffSLP. 1. Pck a β F unformly at random from a subset of F of sze at least 2 2s. 2. Usng the CoeffSLP oracle, for each [t..r] compute a p := Coeff(z p, g(z + β)). 3. Accept f and only f a p 0 for some [t..r]. To see why ths works note that f deg(g(z)) < p t, then each a p s zero and therefore the above test always rejects. For the other drecton assume deg(g(z)) p t log deg(g(z)). Let := log p so that p deg(g(z)) < p +1. Then by clam 4.12 a p wll be nonzero (wth hgh probablty over the random choce of β) and hence our test accepts wth hgh probablty. Ths completes the proof of the theorem. Combnng Theorems 4.1 and 4.7, we mmedately get: Theorem DegSLP s n corp PP. 5 Dscusson We have seen that for the class of polynomal ntegers t s possble to compare two sums of square roots by keepng precson of up to polynomally many bts (durng square root computatons). Although, polynomal ntegers form a nontrval class, the condton that X s suffcently large also makes them very restrctve at the same tme. The hope s that t may be possble to explot results smlar to that of Theorem 2.1 to show somethng stronger for the case of ntegers. As a next step, we would be nterested n a smlar result where X s constraned as X poly(n, d) B c (c s a constant), nstead of X (B + 1) poly(n,d) as s the case n our analyss. Could encodng the ntegers as multvarate polynomals be useful n ths regard? 15

16 Nonetheless, polynomal ntegers are perhaps nterestng from one perspectve. A plausble way to make the sum S very small s to assume that the number of + and sgns n S = n =1 δ a are equal and all the ntegers a s are somewhat very close to each other. Notce that, because of the assumpton that X s large, all ntegers of the form X d + b 1 X d b d are reasonably close to X d. Our analyss shows that at least for ths case any non-zero sum S s stll suffcently large. As a fnal remark on the sum of square roots problem, we would lke to note that the proofs and the results presented here generalze n a straghtforward manner to general sums of radcals - lke sums of cube roots or fourth roots of ntegers. We feel that the complexty of the problem DegSLP s not understood well enough. In partcular no hardness results are known for t. We conclude by posng the followng problem: Problem 5.1. (Hardness of DegSLP ): Does there exst an effcent randomzed algorthm for DegSLP?... or, wll the exstence of such an algorthm for DegSLP lead to a collapse of the polynomal herarchy? Acknowledgements We thank Kurt Mehlhorn and Pyush Kurur for some nsghtful dscussons on ths work. Thanks also to the anonymous revewers whose comments have helped us mprove the presentaton of ths paper. 16

17 References [ABKPM09] Erc Allender, Peter Bürgsser, Johan Kjeldgaard-Pedersen, and Peter Bro Mltersen. On the Complexty of Numercal Analyss. SIAM J. Comput., 38(5): , [Bˆ00] [BD10] [BFMS00] [Blö91] M. Bôcher. On lnear dependence of functons of one varables. Bull. Amer. Math. Soc., 7: , Aln Bostan and Phlppe Dumas. Wronskans and lnear ndependence. Amercan Mathematcal Monthly, 117(8): , Chrstoph Burnkel, Rudolf Flescher, Kurt Mehlhorn, and Stefan Schrra. A Strong and Easly Computable Separaton Bound for Arthmetc Expressons Involvng Radcals. Algorthmca, 27(1):87 99, Johannes Blömer. Computng Sums of Radcals n Polynomal Tme. In Proceedngs of the 32nd IEEE Symposum on Foundatons of Computer Scence, FOCS 91, pages , [Blö93] Johannes Blömer. Computng Sums of Radcals n Polynomal Tme Avalable from sbbtex/computngsumsofradcals.pdf. [Bre76] [Che06] [CMSC10] [EHM05] [GGJ76] [Goe98] [IL93] [KP07] [LN94] Rchard P. Brent. Fast Multple-Precson Evaluaton of Elementary Functons. Journal of the ACM, 23(2): , Q Cheng. On Comparng Sums of Square Roots of Small Integers. In MFCS, pages , Q Cheng, Xanmeng Meng, Cel Sun, and Jazhe Chen. Boundng the Sum of Square Roots va Lattce Reducton. Mathematcs of Computaton, 79: , G. Effnger, K. H. Hcks, and G. L. Mullen. Integers and Polynomals: comparng the close cousns Z and F[x]. Mathematcal Intellgencer, (27):26 34, M. R. Garey, Ronald L. Graham, and Davd S. Johnson. Some NP-Complete Geometrc Problems. In STOC, pages 10 22, Mchele Goemans. Semdefnte Programmng and Combnatoral Optmzaton. Documenta Mathematca, Extra Volume ICM 1998, III: , Nel Immerman and Susan Landau. The Smlartes (and Dfferences) between Polynomals and Integers. In Internatonal Conference on Number Theoretc and Algebrac Methods n Computer Scence, pages 57 59, Pascal Koran and Sylvan Perfel. The complexty of two problems on arthmetc crcuts. Theor. Comput. Sc., 389(1-2): , Rudolf Ldl and Harald Nederreter. Introducton to fnte felds and ther applcatons. Cambrdge Unversty Press, 2nd edton,

18 [Mal96] [MR08] [MS00] [QW06] [RT92] [Tw92] G Malojovch. An effectve verson of Kronecker s theorem on smultaneous Dophantne equaton. Techncal report, Cty Unversty of Hong Kong, Wolfgang Mulzer and Günter Rote. Mnmum-weght trangulaton s NP-hard. Journal of the ACM, 55(2), Kurt Mehlhorn and Stefan Schrra. Generalzed and mproved constructve separaton bound for real algebrac expressons. Research Report MPI-I , Max- Planck-Insttut für Informatk, Stuhlsatzenhausweg 85, Saarbrücken, Germany, November Janbo Qan and Cao An Wang. How much precson s needed to compare two sums of square roots of ntegers? Inf. Process. Lett., 100(5): , John H. Ref and Stephen R. Tate. On Threshold Crcuts and Polynomal Computaton. SIAM J. Comput., 21(5): , P. Twar. A problem that s easer to solve on the unt-cost algebrac ram. Journal of Complexty, (8): , [vl99] Jacobus Hendrcus van Lnt. Introducton to codng theory. Sprnger,

1 Example 1: Axis-aligned rectangles

1 Example 1: Axis-aligned rectangles COS 511: Theoretcal Machne Learnng Lecturer: Rob Schapre Lecture # 6 Scrbe: Aaron Schld February 21, 2013 Last class, we dscussed an analogue for Occam s Razor for nfnte hypothess spaces that, n conjuncton

More information

Luby s Alg. for Maximal Independent Sets using Pairwise Independence

Luby s Alg. for Maximal Independent Sets using Pairwise Independence Lecture Notes for Randomzed Algorthms Luby s Alg. for Maxmal Independent Sets usng Parwse Independence Last Updated by Erc Vgoda on February, 006 8. Maxmal Independent Sets For a graph G = (V, E), an ndependent

More information

We are now ready to answer the question: What are the possible cardinalities for finite fields?

We are now ready to answer the question: What are the possible cardinalities for finite fields? Chapter 3 Fnte felds We have seen, n the prevous chapters, some examples of fnte felds. For example, the resdue class rng Z/pZ (when p s a prme) forms a feld wth p elements whch may be dentfed wth the

More information

8.5 UNITARY AND HERMITIAN MATRICES. The conjugate transpose of a complex matrix A, denoted by A*, is given by

8.5 UNITARY AND HERMITIAN MATRICES. The conjugate transpose of a complex matrix A, denoted by A*, is given by 6 CHAPTER 8 COMPLEX VECTOR SPACES 5. Fnd the kernel of the lnear transformaton gven n Exercse 5. In Exercses 55 and 56, fnd the mage of v, for the ndcated composton, where and are gven by the followng

More information

Recurrence. 1 Definitions and main statements

Recurrence. 1 Definitions and main statements Recurrence 1 Defntons and man statements Let X n, n = 0, 1, 2,... be a MC wth the state space S = (1, 2,...), transton probabltes p j = P {X n+1 = j X n = }, and the transton matrx P = (p j ),j S def.

More information

Support Vector Machines

Support Vector Machines Support Vector Machnes Max Wellng Department of Computer Scence Unversty of Toronto 10 Kng s College Road Toronto, M5S 3G5 Canada wellng@cs.toronto.edu Abstract Ths s a note to explan support vector machnes.

More information

v a 1 b 1 i, a 2 b 2 i,..., a n b n i.

v a 1 b 1 i, a 2 b 2 i,..., a n b n i. SECTION 8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS 455 8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS All the vector spaces we have studed thus far n the text are real vector spaces snce the scalars are

More information

n + d + q = 24 and.05n +.1d +.25q = 2 { n + d + q = 24 (3) n + 2d + 5q = 40 (2)

n + d + q = 24 and.05n +.1d +.25q = 2 { n + d + q = 24 (3) n + 2d + 5q = 40 (2) MATH 16T Exam 1 : Part I (In-Class) Solutons 1. (0 pts) A pggy bank contans 4 cons, all of whch are nckels (5 ), dmes (10 ) or quarters (5 ). The pggy bank also contans a con of each denomnaton. The total

More information

BERNSTEIN POLYNOMIALS

BERNSTEIN POLYNOMIALS On-Lne Geometrc Modelng Notes BERNSTEIN POLYNOMIALS Kenneth I. Joy Vsualzaton and Graphcs Research Group Department of Computer Scence Unversty of Calforna, Davs Overvew Polynomals are ncredbly useful

More information

Ring structure of splines on triangulations

Ring structure of splines on triangulations www.oeaw.ac.at Rng structure of splnes on trangulatons N. Vllamzar RICAM-Report 2014-48 www.rcam.oeaw.ac.at RING STRUCTURE OF SPLINES ON TRIANGULATIONS NELLY VILLAMIZAR Introducton For a trangulated regon

More information

+ + + - - This circuit than can be reduced to a planar circuit

+ + + - - This circuit than can be reduced to a planar circuit MeshCurrent Method The meshcurrent s analog of the nodeoltage method. We sole for a new set of arables, mesh currents, that automatcally satsfy KCLs. As such, meshcurrent method reduces crcut soluton to

More information

Module 2 LOSSLESS IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Module 2 LOSSLESS IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur Module LOSSLESS IMAGE COMPRESSION SYSTEMS Lesson 3 Lossless Compresson: Huffman Codng Instructonal Objectves At the end of ths lesson, the students should be able to:. Defne and measure source entropy..

More information

The OC Curve of Attribute Acceptance Plans

The OC Curve of Attribute Acceptance Plans The OC Curve of Attrbute Acceptance Plans The Operatng Characterstc (OC) curve descrbes the probablty of acceptng a lot as a functon of the lot s qualty. Fgure 1 shows a typcal OC Curve. 10 8 6 4 1 3 4

More information

Quantization Effects in Digital Filters

Quantization Effects in Digital Filters Quantzaton Effects n Dgtal Flters Dstrbuton of Truncaton Errors In two's complement representaton an exact number would have nfntely many bts (n general). When we lmt the number of bts to some fnte value

More information

Institute of Informatics, Faculty of Business and Management, Brno University of Technology,Czech Republic

Institute of Informatics, Faculty of Business and Management, Brno University of Technology,Czech Republic Lagrange Multplers as Quanttatve Indcators n Economcs Ivan Mezník Insttute of Informatcs, Faculty of Busness and Management, Brno Unversty of TechnologCzech Republc Abstract The quanttatve role of Lagrange

More information

where the coordinates are related to those in the old frame as follows.

where the coordinates are related to those in the old frame as follows. Chapter 2 - Cartesan Vectors and Tensors: Ther Algebra Defnton of a vector Examples of vectors Scalar multplcaton Addton of vectors coplanar vectors Unt vectors A bass of non-coplanar vectors Scalar product

More information

benefit is 2, paid if the policyholder dies within the year, and probability of death within the year is ).

benefit is 2, paid if the policyholder dies within the year, and probability of death within the year is ). REVIEW OF RISK MANAGEMENT CONCEPTS LOSS DISTRIBUTIONS AND INSURANCE Loss and nsurance: When someone s subject to the rsk of ncurrng a fnancal loss, the loss s generally modeled usng a random varable or

More information

The Greedy Method. Introduction. 0/1 Knapsack Problem

The Greedy Method. Introduction. 0/1 Knapsack Problem The Greedy Method Introducton We have completed data structures. We now are gong to look at algorthm desgn methods. Often we are lookng at optmzaton problems whose performance s exponental. For an optmzaton

More information

PERRON FROBENIUS THEOREM

PERRON FROBENIUS THEOREM PERRON FROBENIUS THEOREM R. CLARK ROBINSON Defnton. A n n matrx M wth real entres m, s called a stochastc matrx provded () all the entres m satsfy 0 m, () each of the columns sum to one, m = for all, ()

More information

8 Algorithm for Binary Searching in Trees

8 Algorithm for Binary Searching in Trees 8 Algorthm for Bnary Searchng n Trees In ths secton we present our algorthm for bnary searchng n trees. A crucal observaton employed by the algorthm s that ths problem can be effcently solved when the

More information

THE METHOD OF LEAST SQUARES THE METHOD OF LEAST SQUARES

THE METHOD OF LEAST SQUARES THE METHOD OF LEAST SQUARES The goal: to measure (determne) an unknown quantty x (the value of a RV X) Realsaton: n results: y 1, y 2,..., y j,..., y n, (the measured values of Y 1, Y 2,..., Y j,..., Y n ) every result s encumbered

More information

Implementation of Deutsch's Algorithm Using Mathcad

Implementation of Deutsch's Algorithm Using Mathcad Implementaton of Deutsch's Algorthm Usng Mathcad Frank Roux The followng s a Mathcad mplementaton of Davd Deutsch's quantum computer prototype as presented on pages - n "Machnes, Logc and Quantum Physcs"

More information

Linear Circuits Analysis. Superposition, Thevenin /Norton Equivalent circuits

Linear Circuits Analysis. Superposition, Thevenin /Norton Equivalent circuits Lnear Crcuts Analyss. Superposton, Theenn /Norton Equalent crcuts So far we hae explored tmendependent (resste) elements that are also lnear. A tmendependent elements s one for whch we can plot an / cure.

More information

Logical Development Of Vogel s Approximation Method (LD-VAM): An Approach To Find Basic Feasible Solution Of Transportation Problem

Logical Development Of Vogel s Approximation Method (LD-VAM): An Approach To Find Basic Feasible Solution Of Transportation Problem INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME, ISSUE, FEBRUARY ISSN 77-866 Logcal Development Of Vogel s Approxmaton Method (LD- An Approach To Fnd Basc Feasble Soluton Of Transportaton

More information

What is Candidate Sampling

What is Candidate Sampling What s Canddate Samplng Say we have a multclass or mult label problem where each tranng example ( x, T ) conssts of a context x a small (mult)set of target classes T out of a large unverse L of possble

More information

Generalizing the degree sequence problem

Generalizing the degree sequence problem Mddlebury College March 2009 Arzona State Unversty Dscrete Mathematcs Semnar The degree sequence problem Problem: Gven an nteger sequence d = (d 1,...,d n ) determne f there exsts a graph G wth d as ts

More information

1. Fundamentals of probability theory 2. Emergence of communication traffic 3. Stochastic & Markovian Processes (SP & MP)

1. Fundamentals of probability theory 2. Emergence of communication traffic 3. Stochastic & Markovian Processes (SP & MP) 6.3 / -- Communcaton Networks II (Görg) SS20 -- www.comnets.un-bremen.de Communcaton Networks II Contents. Fundamentals of probablty theory 2. Emergence of communcaton traffc 3. Stochastc & Markovan Processes

More information

SPEE Recommended Evaluation Practice #6 Definition of Decline Curve Parameters Background:

SPEE Recommended Evaluation Practice #6 Definition of Decline Curve Parameters Background: SPEE Recommended Evaluaton Practce #6 efnton of eclne Curve Parameters Background: The producton hstores of ol and gas wells can be analyzed to estmate reserves and future ol and gas producton rates and

More information

Embedding lattices in the Kleene degrees

Embedding lattices in the Kleene degrees F U N D A M E N T A MATHEMATICAE 62 (999) Embeddng lattces n the Kleene degrees by Hsato M u r a k (Nagoya) Abstract. Under ZFC+CH, we prove that some lattces whose cardnaltes do not exceed ℵ can be embedded

More information

Matrix Multiplication I

Matrix Multiplication I Matrx Multplcaton I Yuval Flmus February 2, 2012 These notes are based on a lecture gven at the Toronto Student Semnar on February 2, 2012. The materal s taen mostly from the boo Algebrac Complexty Theory

More information

A hybrid global optimization algorithm based on parallel chaos optimization and outlook algorithm

A hybrid global optimization algorithm based on parallel chaos optimization and outlook algorithm Avalable onlne www.ocpr.com Journal of Chemcal and Pharmaceutcal Research, 2014, 6(7):1884-1889 Research Artcle ISSN : 0975-7384 CODEN(USA) : JCPRC5 A hybrd global optmzaton algorthm based on parallel

More information

Compiling for Parallelism & Locality. Dependence Testing in General. Algorithms for Solving the Dependence Problem. Dependence Testing

Compiling for Parallelism & Locality. Dependence Testing in General. Algorithms for Solving the Dependence Problem. Dependence Testing Complng for Parallelsm & Localty Dependence Testng n General Assgnments Deadlne for proect 4 extended to Dec 1 Last tme Data dependences and loops Today Fnsh data dependence analyss for loops General code

More information

Series Solutions of ODEs 2 the Frobenius method. The basic idea of the Frobenius method is to look for solutions of the form 3

Series Solutions of ODEs 2 the Frobenius method. The basic idea of the Frobenius method is to look for solutions of the form 3 Royal Holloway Unversty of London Department of Physs Seres Solutons of ODEs the Frobenus method Introduton to the Methodology The smple seres expanson method works for dfferental equatons whose solutons

More information

A Probabilistic Theory of Coherence

A Probabilistic Theory of Coherence A Probablstc Theory of Coherence BRANDEN FITELSON. The Coherence Measure C Let E be a set of n propostons E,..., E n. We seek a probablstc measure C(E) of the degree of coherence of E. Intutvely, we want

More information

Multiplication Algorithms for Radix-2 RN-Codings and Two s Complement Numbers

Multiplication Algorithms for Radix-2 RN-Codings and Two s Complement Numbers Multplcaton Algorthms for Radx- RN-Codngs and Two s Complement Numbers Jean-Luc Beuchat Projet Arénare, LIP, ENS Lyon 46, Allée d Itale F 69364 Lyon Cedex 07 jean-luc.beuchat@ens-lyon.fr Jean-Mchel Muller

More information

Project Networks With Mixed-Time Constraints

Project Networks With Mixed-Time Constraints Project Networs Wth Mxed-Tme Constrants L Caccetta and B Wattananon Western Australan Centre of Excellence n Industral Optmsaton (WACEIO) Curtn Unversty of Technology GPO Box U1987 Perth Western Australa

More information

Fisher Markets and Convex Programs

Fisher Markets and Convex Programs Fsher Markets and Convex Programs Nkhl R. Devanur 1 Introducton Convex programmng dualty s usually stated n ts most general form, wth convex objectve functons and convex constrants. (The book by Boyd and

More information

Extending Probabilistic Dynamic Epistemic Logic

Extending Probabilistic Dynamic Epistemic Logic Extendng Probablstc Dynamc Epstemc Logc Joshua Sack May 29, 2008 Probablty Space Defnton A probablty space s a tuple (S, A, µ), where 1 S s a set called the sample space. 2 A P(S) s a σ-algebra: a set

More information

Calculation of Sampling Weights

Calculation of Sampling Weights Perre Foy Statstcs Canada 4 Calculaton of Samplng Weghts 4.1 OVERVIEW The basc sample desgn used n TIMSS Populatons 1 and 2 was a two-stage stratfed cluster desgn. 1 The frst stage conssted of a sample

More information

The Mathematical Derivation of Least Squares

The Mathematical Derivation of Least Squares Pscholog 885 Prof. Federco The Mathematcal Dervaton of Least Squares Back when the powers that e forced ou to learn matr algera and calculus, I et ou all asked ourself the age-old queston: When the hell

More information

Section C2: BJT Structure and Operational Modes

Section C2: BJT Structure and Operational Modes Secton 2: JT Structure and Operatonal Modes Recall that the semconductor dode s smply a pn juncton. Dependng on how the juncton s based, current may easly flow between the dode termnals (forward bas, v

More information

PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 12

PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 12 14 The Ch-squared dstrbuton PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 1 If a normal varable X, havng mean µ and varance σ, s standardsed, the new varable Z has a mean 0 and varance 1. When ths standardsed

More information

The Development of Web Log Mining Based on Improve-K-Means Clustering Analysis

The Development of Web Log Mining Based on Improve-K-Means Clustering Analysis The Development of Web Log Mnng Based on Improve-K-Means Clusterng Analyss TngZhong Wang * College of Informaton Technology, Luoyang Normal Unversty, Luoyang, 471022, Chna wangtngzhong2@sna.cn Abstract.

More information

Using Series to Analyze Financial Situations: Present Value

Using Series to Analyze Financial Situations: Present Value 2.8 Usng Seres to Analyze Fnancal Stuatons: Present Value In the prevous secton, you learned how to calculate the amount, or future value, of an ordnary smple annuty. The amount s the sum of the accumulated

More information

Section 5.4 Annuities, Present Value, and Amortization

Section 5.4 Annuities, Present Value, and Amortization Secton 5.4 Annutes, Present Value, and Amortzaton Present Value In Secton 5.2, we saw that the present value of A dollars at nterest rate per perod for n perods s the amount that must be deposted today

More information

HÜCKEL MOLECULAR ORBITAL THEORY

HÜCKEL MOLECULAR ORBITAL THEORY 1 HÜCKEL MOLECULAR ORBITAL THEORY In general, the vast maorty polyatomc molecules can be thought of as consstng of a collecton of two electron bonds between pars of atoms. So the qualtatve pcture of σ

More information

7.5. Present Value of an Annuity. Investigate

7.5. Present Value of an Annuity. Investigate 7.5 Present Value of an Annuty Owen and Anna are approachng retrement and are puttng ther fnances n order. They have worked hard and nvested ther earnngs so that they now have a large amount of money on

More information

Joe Pimbley, unpublished, 2005. Yield Curve Calculations

Joe Pimbley, unpublished, 2005. Yield Curve Calculations Joe Pmbley, unpublshed, 005. Yeld Curve Calculatons Background: Everythng s dscount factors Yeld curve calculatons nclude valuaton of forward rate agreements (FRAs), swaps, nterest rate optons, and forward

More information

Feature selection for intrusion detection. Slobodan Petrović NISlab, Gjøvik University College

Feature selection for intrusion detection. Slobodan Petrović NISlab, Gjøvik University College Feature selecton for ntruson detecton Slobodan Petrovć NISlab, Gjøvk Unversty College Contents The feature selecton problem Intruson detecton Traffc features relevant for IDS The CFS measure The mrmr measure

More information

NON-CONSTANT SUM RED-AND-BLACK GAMES WITH BET-DEPENDENT WIN PROBABILITY FUNCTION LAURA PONTIGGIA, University of the Sciences in Philadelphia

NON-CONSTANT SUM RED-AND-BLACK GAMES WITH BET-DEPENDENT WIN PROBABILITY FUNCTION LAURA PONTIGGIA, University of the Sciences in Philadelphia To appear n Journal o Appled Probablty June 2007 O-COSTAT SUM RED-AD-BLACK GAMES WITH BET-DEPEDET WI PROBABILITY FUCTIO LAURA POTIGGIA, Unversty o the Scences n Phladelpha Abstract In ths paper we nvestgate

More information

DEFINING %COMPLETE IN MICROSOFT PROJECT

DEFINING %COMPLETE IN MICROSOFT PROJECT CelersSystems DEFINING %COMPLETE IN MICROSOFT PROJECT PREPARED BY James E Aksel, PMP, PMI-SP, MVP For Addtonal Informaton about Earned Value Management Systems and reportng, please contact: CelersSystems,

More information

"Research Note" APPLICATION OF CHARGE SIMULATION METHOD TO ELECTRIC FIELD CALCULATION IN THE POWER CABLES *

Research Note APPLICATION OF CHARGE SIMULATION METHOD TO ELECTRIC FIELD CALCULATION IN THE POWER CABLES * Iranan Journal of Scence & Technology, Transacton B, Engneerng, ol. 30, No. B6, 789-794 rnted n The Islamc Republc of Iran, 006 Shraz Unversty "Research Note" ALICATION OF CHARGE SIMULATION METHOD TO ELECTRIC

More information

How Sets of Coherent Probabilities May Serve as Models for Degrees of Incoherence

How Sets of Coherent Probabilities May Serve as Models for Degrees of Incoherence 1 st Internatonal Symposum on Imprecse Probabltes and Ther Applcatons, Ghent, Belgum, 29 June 2 July 1999 How Sets of Coherent Probabltes May Serve as Models for Degrees of Incoherence Mar J. Schervsh

More information

1. Measuring association using correlation and regression

1. Measuring association using correlation and regression How to measure assocaton I: Correlaton. 1. Measurng assocaton usng correlaton and regresson We often would lke to know how one varable, such as a mother's weght, s related to another varable, such as a

More information

Texas Instruments 30X IIS Calculator

Texas Instruments 30X IIS Calculator Texas Instruments 30X IIS Calculator Keystrokes for the TI-30X IIS are shown for a few topcs n whch keystrokes are unque. Start by readng the Quk Start secton. Then, before begnnng a specfc unt of the

More information

Lecture 2: Single Layer Perceptrons Kevin Swingler

Lecture 2: Single Layer Perceptrons Kevin Swingler Lecture 2: Sngle Layer Perceptrons Kevn Sngler kms@cs.str.ac.uk Recap: McCulloch-Ptts Neuron Ths vastly smplfed model of real neurons s also knon as a Threshold Logc Unt: W 2 A Y 3 n W n. A set of synapses

More information

An Alternative Way to Measure Private Equity Performance

An Alternative Way to Measure Private Equity Performance An Alternatve Way to Measure Prvate Equty Performance Peter Todd Parlux Investment Technology LLC Summary Internal Rate of Return (IRR) s probably the most common way to measure the performance of prvate

More information

Nonbinary Quantum Error-Correcting Codes from Algebraic Curves

Nonbinary Quantum Error-Correcting Codes from Algebraic Curves Nonbnary Quantum Error-Correctng Codes from Algebrac Curves Jon-Lark Km and Judy Walker Department of Mathematcs Unversty of Nebraska-Lncoln, Lncoln, NE 68588-0130 USA e-mal: {jlkm, jwalker}@math.unl.edu

More information

Forecasting the Direction and Strength of Stock Market Movement

Forecasting the Direction and Strength of Stock Market Movement Forecastng the Drecton and Strength of Stock Market Movement Jngwe Chen Mng Chen Nan Ye cjngwe@stanford.edu mchen5@stanford.edu nanye@stanford.edu Abstract - Stock market s one of the most complcated systems

More information

Production. 2. Y is closed A set is closed if it contains its boundary. We need this for the solution existence in the profit maximization problem.

Production. 2. Y is closed A set is closed if it contains its boundary. We need this for the solution existence in the profit maximization problem. Producer Theory Producton ASSUMPTION 2.1 Propertes of the Producton Set The producton set Y satsfes the followng propertes 1. Y s non-empty If Y s empty, we have nothng to talk about 2. Y s closed A set

More information

CS 2750 Machine Learning. Lecture 3. Density estimation. CS 2750 Machine Learning. Announcements

CS 2750 Machine Learning. Lecture 3. Density estimation. CS 2750 Machine Learning. Announcements Lecture 3 Densty estmaton Mlos Hauskrecht mlos@cs.ptt.edu 5329 Sennott Square Next lecture: Matlab tutoral Announcements Rules for attendng the class: Regstered for credt Regstered for audt (only f there

More information

INTERPRETING TRUE ARITHMETIC IN THE LOCAL STRUCTURE OF THE ENUMERATION DEGREES.

INTERPRETING TRUE ARITHMETIC IN THE LOCAL STRUCTURE OF THE ENUMERATION DEGREES. INTERPRETING TRUE ARITHMETIC IN THE LOCAL STRUCTURE OF THE ENUMERATION DEGREES. HRISTO GANCHEV AND MARIYA SOSKOVA 1. Introducton Degree theory studes mathematcal structures, whch arse from a formal noton

More information

THE DISTRIBUTION OF LOAN PORTFOLIO VALUE * Oldrich Alfons Vasicek

THE DISTRIBUTION OF LOAN PORTFOLIO VALUE * Oldrich Alfons Vasicek HE DISRIBUION OF LOAN PORFOLIO VALUE * Oldrch Alfons Vascek he amount of captal necessary to support a portfolo of debt securtes depends on the probablty dstrbuton of the portfolo loss. Consder a portfolo

More information

How To Calculate The Accountng Perod Of Nequalty

How To Calculate The Accountng Perod Of Nequalty Inequalty and The Accountng Perod Quentn Wodon and Shlomo Ytzha World Ban and Hebrew Unversty September Abstract Income nequalty typcally declnes wth the length of tme taen nto account for measurement.

More information

Complete Fairness in Secure Two-Party Computation

Complete Fairness in Secure Two-Party Computation Complete Farness n Secure Two-Party Computaton S. Dov Gordon Carmt Hazay Jonathan Katz Yehuda Lndell Abstract In the settng of secure two-party computaton, two mutually dstrustng partes wsh to compute

More information

To Fill or not to Fill: The Gas Station Problem

To Fill or not to Fill: The Gas Station Problem To Fll or not to Fll: The Gas Staton Problem Samr Khuller Azarakhsh Malekan Julán Mestre Abstract In ths paper we study several routng problems that generalze shortest paths and the Travelng Salesman Problem.

More information

Loop Parallelization

Loop Parallelization - - Loop Parallelzaton C-52 Complaton steps: nested loops operatng on arrays, sequentell executon of teraton space DECLARE B[..,..+] FOR I :=.. FOR J :=.. I B[I,J] := B[I-,J]+B[I-,J-] ED FOR ED FOR analyze

More information

The descriptive complexity of the family of Banach spaces with the π-property

The descriptive complexity of the family of Banach spaces with the π-property Arab. J. Math. (2015) 4:35 39 DOI 10.1007/s40065-014-0116-3 Araban Journal of Mathematcs Ghadeer Ghawadrah The descrptve complexty of the famly of Banach spaces wth the π-property Receved: 25 March 2014

More information

Minimal Coding Network With Combinatorial Structure For Instantaneous Recovery From Edge Failures

Minimal Coding Network With Combinatorial Structure For Instantaneous Recovery From Edge Failures Mnmal Codng Network Wth Combnatoral Structure For Instantaneous Recovery From Edge Falures Ashly Joseph 1, Mr.M.Sadsh Sendl 2, Dr.S.Karthk 3 1 Fnal Year ME CSE Student Department of Computer Scence Engneerng

More information

Finite Math Chapter 10: Study Guide and Solution to Problems

Finite Math Chapter 10: Study Guide and Solution to Problems Fnte Math Chapter 10: Study Gude and Soluton to Problems Basc Formulas and Concepts 10.1 Interest Basc Concepts Interest A fee a bank pays you for money you depost nto a savngs account. Prncpal P The amount

More information

Simple Interest Loans (Section 5.1) :

Simple Interest Loans (Section 5.1) : Chapter 5 Fnance The frst part of ths revew wll explan the dfferent nterest and nvestment equatons you learned n secton 5.1 through 5.4 of your textbook and go through several examples. The second part

More information

Logistic Regression. Lecture 4: More classifiers and classes. Logistic regression. Adaboost. Optimization. Multiple class classification

Logistic Regression. Lecture 4: More classifiers and classes. Logistic regression. Adaboost. Optimization. Multiple class classification Lecture 4: More classfers and classes C4B Machne Learnng Hlary 20 A. Zsserman Logstc regresson Loss functons revsted Adaboost Loss functons revsted Optmzaton Multple class classfcaton Logstc Regresson

More information

On the Optimal Control of a Cascade of Hydro-Electric Power Stations

On the Optimal Control of a Cascade of Hydro-Electric Power Stations On the Optmal Control of a Cascade of Hydro-Electrc Power Statons M.C.M. Guedes a, A.F. Rbero a, G.V. Smrnov b and S. Vlela c a Department of Mathematcs, School of Scences, Unversty of Porto, Portugal;

More information

A Novel Methodology of Working Capital Management for Large. Public Constructions by Using Fuzzy S-curve Regression

A Novel Methodology of Working Capital Management for Large. Public Constructions by Using Fuzzy S-curve Regression Novel Methodology of Workng Captal Management for Large Publc Constructons by Usng Fuzzy S-curve Regresson Cheng-Wu Chen, Morrs H. L. Wang and Tng-Ya Hseh Department of Cvl Engneerng, Natonal Central Unversty,

More information

Energies of Network Nastsemble

Energies of Network Nastsemble Supplementary materal: Assessng the relevance of node features for network structure Gnestra Bancon, 1 Paolo Pn,, 3 and Matteo Marsl 1 1 The Abdus Salam Internatonal Center for Theoretcal Physcs, Strada

More information

The Full-Wave Rectifier

The Full-Wave Rectifier 9/3/2005 The Full Wae ectfer.doc /0 The Full-Wae ectfer Consder the followng juncton dode crcut: s (t) Power Lne s (t) 2 Note that we are usng a transformer n ths crcut. The job of ths transformer s to

More information

J. Parallel Distrib. Comput.

J. Parallel Distrib. Comput. J. Parallel Dstrb. Comput. 71 (2011) 62 76 Contents lsts avalable at ScenceDrect J. Parallel Dstrb. Comput. journal homepage: www.elsever.com/locate/jpdc Optmzng server placement n dstrbuted systems n

More information

Causal, Explanatory Forecasting. Analysis. Regression Analysis. Simple Linear Regression. Which is Independent? Forecasting

Causal, Explanatory Forecasting. Analysis. Regression Analysis. Simple Linear Regression. Which is Independent? Forecasting Causal, Explanatory Forecastng Assumes cause-and-effect relatonshp between system nputs and ts output Forecastng wth Regresson Analyss Rchard S. Barr Inputs System Cause + Effect Relatonshp The job of

More information

TENSOR GAUGE FIELDS OF DEGREE THREE

TENSOR GAUGE FIELDS OF DEGREE THREE TENSOR GAUGE FIELDS OF DEGREE THREE E.M. CIOROIANU Department of Physcs, Unversty of Craova, A. I. Cuza 13, 2585, Craova, Romana, EU E-mal: manache@central.ucv.ro Receved February 2, 213 Startng from a

More information

F-Rational Rings and the Integral Closures of Ideals

F-Rational Rings and the Integral Closures of Ideals Mchgan Math. J. 49 (2001) F-Ratonal Rngs and the Integral Closures of Ideals Ian M. Aberbach & Crag Huneke 1. Introducton The hstory of the Brançon Skoda theorem and ts ensung avatars n commutatve algebra

More information

Power-of-Two Policies for Single- Warehouse Multi-Retailer Inventory Systems with Order Frequency Discounts

Power-of-Two Policies for Single- Warehouse Multi-Retailer Inventory Systems with Order Frequency Discounts Power-of-wo Polces for Sngle- Warehouse Mult-Retaler Inventory Systems wth Order Frequency Dscounts José A. Ventura Pennsylvana State Unversty (USA) Yale. Herer echnon Israel Insttute of echnology (Israel)

More information

QUESTIONS, How can quantum computers do the amazing things that they are able to do, such. cryptography quantum computers

QUESTIONS, How can quantum computers do the amazing things that they are able to do, such. cryptography quantum computers 2O cryptography quantum computers cryptography quantum computers QUESTIONS, Quantum Computers, and Cryptography A mathematcal metaphor for the power of quantum algorthms Mark Ettnger How can quantum computers

More information

On Lockett pairs and Lockett conjecture for π-soluble Fitting classes

On Lockett pairs and Lockett conjecture for π-soluble Fitting classes On Lockett pars and Lockett conjecture for π-soluble Fttng classes Lujn Zhu Department of Mathematcs, Yangzhou Unversty, Yangzhou 225002, P.R. Chna E-mal: ljzhu@yzu.edu.cn Nanyng Yang School of Mathematcs

More information

Answer: A). There is a flatter IS curve in the high MPC economy. Original LM LM after increase in M. IS curve for low MPC economy

Answer: A). There is a flatter IS curve in the high MPC economy. Original LM LM after increase in M. IS curve for low MPC economy 4.02 Quz Solutons Fall 2004 Multple-Choce Questons (30/00 ponts) Please, crcle the correct answer for each of the followng 0 multple-choce questons. For each queston, only one of the answers s correct.

More information

An Analysis of Central Processor Scheduling in Multiprogrammed Computer Systems

An Analysis of Central Processor Scheduling in Multiprogrammed Computer Systems STAN-CS-73-355 I SU-SE-73-013 An Analyss of Central Processor Schedulng n Multprogrammed Computer Systems (Dgest Edton) by Thomas G. Prce October 1972 Techncal Report No. 57 Reproducton n whole or n part

More information

FINANCIAL MATHEMATICS. A Practical Guide for Actuaries. and other Business Professionals

FINANCIAL MATHEMATICS. A Practical Guide for Actuaries. and other Business Professionals FINANCIAL MATHEMATICS A Practcal Gude for Actuares and other Busness Professonals Second Edton CHRIS RUCKMAN, FSA, MAAA JOE FRANCIS, FSA, MAAA, CFA Study Notes Prepared by Kevn Shand, FSA, FCIA Assstant

More information

Lecture 3: Force of Interest, Real Interest Rate, Annuity

Lecture 3: Force of Interest, Real Interest Rate, Annuity Lecture 3: Force of Interest, Real Interest Rate, Annuty Goals: Study contnuous compoundng and force of nterest Dscuss real nterest rate Learn annuty-mmedate, and ts present value Study annuty-due, and

More information

Descriptive Models. Cluster Analysis. Example. General Applications of Clustering. Examples of Clustering Applications

Descriptive Models. Cluster Analysis. Example. General Applications of Clustering. Examples of Clustering Applications CMSC828G Prncples of Data Mnng Lecture #9 Today s Readng: HMS, chapter 9 Today s Lecture: Descrptve Modelng Clusterng Algorthms Descrptve Models model presents the man features of the data, a global summary

More information

Enabling P2P One-view Multi-party Video Conferencing

Enabling P2P One-view Multi-party Video Conferencing Enablng P2P One-vew Mult-party Vdeo Conferencng Yongxang Zhao, Yong Lu, Changja Chen, and JanYn Zhang Abstract Mult-Party Vdeo Conferencng (MPVC) facltates realtme group nteracton between users. Whle P2P

More information

How To Know The Components Of Mean Squared Error Of Herarchcal Estmator S

How To Know The Components Of Mean Squared Error Of Herarchcal Estmator S S C H E D A E I N F O R M A T I C A E VOLUME 0 0 On Mean Squared Error of Herarchcal Estmator Stans law Brodowsk Faculty of Physcs, Astronomy, and Appled Computer Scence, Jagellonan Unversty, Reymonta

More information

Fast degree elevation and knot insertion for B-spline curves

Fast degree elevation and knot insertion for B-spline curves Computer Aded Geometrc Desgn 22 (2005) 183 197 www.elsever.com/locate/cagd Fast degree elevaton and knot nserton for B-splne curves Q-Xng Huang a,sh-mnhu a,, Ralph R. Martn b a Department of Computer Scence

More information

Period and Deadline Selection for Schedulability in Real-Time Systems

Period and Deadline Selection for Schedulability in Real-Time Systems Perod and Deadlne Selecton for Schedulablty n Real-Tme Systems Thdapat Chantem, Xaofeng Wang, M.D. Lemmon, and X. Sharon Hu Department of Computer Scence and Engneerng, Department of Electrcal Engneerng

More information

How To Understand The Results Of The German Meris Cloud And Water Vapour Product

How To Understand The Results Of The German Meris Cloud And Water Vapour Product Ttel: Project: Doc. No.: MERIS level 3 cloud and water vapour products MAPP MAPP-ATBD-ClWVL3 Issue: 1 Revson: 0 Date: 9.12.1998 Functon Name Organsaton Sgnature Date Author: Bennartz FUB Preusker FUB Schüller

More information

Combinatorial Agency of Threshold Functions

Combinatorial Agency of Threshold Functions Combnatoral Agency of Threshold Functons Shal Jan Computer Scence Department Yale Unversty New Haven, CT 06520 shal.jan@yale.edu Davd C. Parkes School of Engneerng and Appled Scences Harvard Unversty Cambrdge,

More information

Jacobian hits circuits: Hitting-sets, lower bounds for depth-d occur-k formulas & depth-3 transcendence degree-k circuits

Jacobian hits circuits: Hitting-sets, lower bounds for depth-d occur-k formulas & depth-3 transcendence degree-k circuits Jacoban hts crcuts: Httng-sets, lower bounds for depth-d occur-k formulas & depth-3 transcendence degree-k crcuts Manndra Agrawal Chandan Saha Ramprasad Saptharsh Ntn Saxena Abstract We present a sngle

More information

Number of Levels Cumulative Annual operating Income per year construction costs costs ($) ($) ($) 1 600,000 35,000 100,000 2 2,200,000 60,000 350,000

Number of Levels Cumulative Annual operating Income per year construction costs costs ($) ($) ($) 1 600,000 35,000 100,000 2 2,200,000 60,000 350,000 Problem Set 5 Solutons 1 MIT s consderng buldng a new car park near Kendall Square. o unversty funds are avalable (overhead rates are under pressure and the new faclty would have to pay for tself from

More information

NPAR TESTS. One-Sample Chi-Square Test. Cell Specification. Observed Frequencies 1O i 6. Expected Frequencies 1EXP i 6

NPAR TESTS. One-Sample Chi-Square Test. Cell Specification. Observed Frequencies 1O i 6. Expected Frequencies 1EXP i 6 PAR TESTS If a WEIGHT varable s specfed, t s used to replcate a case as many tmes as ndcated by the weght value rounded to the nearest nteger. If the workspace requrements are exceeded and samplng has

More information

Multiple stage amplifiers

Multiple stage amplifiers Multple stage amplfers Ams: Examne a few common 2-transstor amplfers: -- Dfferental amplfers -- Cascode amplfers -- Darlngton pars -- current mrrors Introduce formal methods for exactly analysng multple

More information

L10: Linear discriminants analysis

L10: Linear discriminants analysis L0: Lnear dscrmnants analyss Lnear dscrmnant analyss, two classes Lnear dscrmnant analyss, C classes LDA vs. PCA Lmtatons of LDA Varants of LDA Other dmensonalty reducton methods CSCE 666 Pattern Analyss

More information

CHOLESTEROL REFERENCE METHOD LABORATORY NETWORK. Sample Stability Protocol

CHOLESTEROL REFERENCE METHOD LABORATORY NETWORK. Sample Stability Protocol CHOLESTEROL REFERENCE METHOD LABORATORY NETWORK Sample Stablty Protocol Background The Cholesterol Reference Method Laboratory Network (CRMLN) developed certfcaton protocols for total cholesterol, HDL

More information