Renormalised steepest descent converges to a two-point attractor

Size: px
Start display at page:

Download "Renormalised steepest descent converges to a two-point attractor"

Transcription

1 Renormalsed steepest descent converges to a two-pont attractor Luc Pronzato, Henry P. Wynn and Anatoly A. Zhglavsky Abstract The behavour of the standard steepest-descent algorthm for a quadratc functon n IR d s nvestgated. We show that by rescalng the terates to reman always on the unt sphere one can reveal specal features of ths behavour. The renormalzed algorthm s shown to converge to a two-pont cycle on the unt crcle. The cycle depends on the startng pont n a complcated manner (the set of ponts convergng to the same cycle s fractal), but all cycles belong to a partcular plane, gven by certan egenvectors of the Hessan matrx of the obectve functon. The stablty of the attractor s analysed. The rate of convergence of the algorthm s nvestgated. It s shown that the worst value of ths rate s obtaned only for some partcular startng ponts. The ntroducton of a relaxaton coeffcent n the steepest-descent algorthm completely changes ts behavour, whch may become chaotc. Dfferent attractors are presented. We show that relaxaton allows a sgnfcantly mproved rate of convergence. 1 Introducton: an unsolved problem For a general smooth functon f( ) the steepest-descent algorthm s where s the k-th terate of the algorthm, x (k+1) = x (k) α k f(x (k) ), (1) x (k) = ( x (k) 1,..., x (k) ) T d f(x) = ( f,..., f ) T x 1 x d Ths work was supported by a UK Engneerng and Physcal Scences Research Councl for the second author and a grant of the French Mnstère de l Éducaton Natonale (nvtaton trennale, procédure PAST) for the thrd author. Dr., Laboratore I3S, CNRS UPRES-A 6070, Sopha Antpols, Valbonne, France Prof., Dept. of Statstcs, Unversty of Warwck, Coventry CV4 7AL, UK Prof., Dept. of Mathematcs, St. Petersburg Unversty, Bblotechnaya sq. 2, , Russa 1

2 s the gradent of f( ), and α k = arg mn α f(x (k) α f(x (k) ). In [?] the authors studed the asymptotc behavour of ths algorthm for a quadratc functon n IR d : f(x) = 1 2 (x x ) T A(x x ), where x, x IR d and A s a postve defnte d d matrx. Ths seemngly smple problem has some very complex aspects whch are uncovered by the followng renormalzaton dea. Defne v (k) = x(k) x x (k) x, (2) whch has the effect of rescalng the terates to reman always on the unt sphere: v = 1. The authors studed the behavour of the process {v (k) } as k. Except for pathologcal cases, n the lmt the process s found to belong to the two dmensonal space spanned by the egenvectors (u 1, u d ) correspondng to the mnmal and maxmum egenvalues λ 1 and λ d of the matrx A under the assumpton 0 < λ 1 < λ 2... λ d 1 < λ d, (3) where λ s the -th ordered egenvalue of A. Orderng the v (k) compatbly wth (3), the conecture states that v (k) 0 for = 2,..., d 1 as k, and that the algorthm, n ts renormalsed form, converges to a two-pont set on the crcle { v v d 2 = 1} n the plane spanned by u 1 and u d. Numercal smulatons show that the same convergence property holds for a convex dfferentable functon locally quadratc around ts mnmum x. It has been known for many years that the asymptotc convergence rate of the algorthm depends on the rato ρ = λ d /λ 1 (for nstance, see [?], p. 152). Thus for any x (k) IR d ( ) 2 ρ 1 f(x (k+1) ) f(x (k) ) (4) ρ + 1 (see also Secton 4). Despte an ntensve search, no more precse propertes concernng the rate of convergence were found n the lterature. The authors doubts about the exstence of results on the asymptotc behavour of the steepest-descent algorthm were ncreased by the complexty of the behavour of the renormalzed algorthm en route to the lmtng crcle. Ths behavour s fractal n nature, and t stll remans open how the lmtng asymptotc rate depends on the startng values. It s the worst-case rate, gven by (4), not the actual rate, whch smply depends on λ 1 and λ d. However we are able to show that the algorthm attracts to two conugate ponts on the crcle. It s easy to check that startng on the crcle the renormalzed algorthm oscllates between the ponts, but wthout attracton to the crcle ths lmtng behavour s only a conecture. We shall return to the dscusson of these conugate ponts and the correspondng asymptotc rate of convergence n Sectons 3 and 4 respectvely. The behavour of the steepest-descent algorthm wth relaxaton s consdered n Secton 5. 2

3 2 Attracton theorem Theorem 1 Let the obectve functon f be f(x) = 1 2 (x x ) T A(x x ), (5) where A s a postve defnte matrx wth ordered egenvalues 0 < λ 1 < λ 2... λ d 1 < λ d. Let V = span(u 1, u d ) be the two-dmensonal plane generated by the (dstnct orthogonal) egenvectors u 1, u d correspondng to λ 1 and λ d, respectvely. Then for any startng vector x (1) for whch u T 1 v (1) 0 and u T d v (1) 0 (6) the algorthm attracts to the plane V n the followng sense: w T v (k) 0 k for any non-zero vector w V, where v (k) s defned by (2). Moreover, the sequence {v (k) } converges to a two-pont cycle. Proof. For the quadratc functon (5) the gradent of f( ) at x (k) equals and g (k) = f(x (k) ) = A(x (k) x ), α k = (g(k) ) T g (k) (g (k) ) T Ag (k). Therefore, the algorthm (1) can be rewrtten as x (k+1) = x (k) (g(k) ) T g (k) (g (k) ) T Ag (k) g(k). (7) Wthout loss of generalty we can assume that x = 0 and the matrx A s dagonal: A = dag(λ 1,... λ d ) where 0 < λ 1 < λ 2... λ d 1 < λ d. Wth ths assumpton, the teraton of the algorthm (7) can be wrtten as x (k+1) = x (k) d=1 (g (k) ) 2 d=1 λ (g (k) ) 2 g(k) for = 1,..., d. (8) Multplyng both sdes of -th equaton of (8) by λ we obtan g (k+1) = g (k) d=1 (g (k) ) 2 d=1 λ (g (k) ) λ g (k) 2 for = 1,..., d. (9) 3

4 Introduce now the varables and ther renormalsed versons Note that and y (k) = = (g (k) ) 2, 0 for = 1,..., d and d =1 z (k+1) y (k+1) = = 1 λ d=1 y (k) d=1 λ y (k) y (k) d=1. (10) y (k) ( d=1 λ λ ) 2 dl=1 ( d=1 λ 2 = 1. The equatons (9) then mply y (k) for = 1,..., d, λ l ) 2 z (k) l for = 1,..., d. (11) Note that the updatng formula (11) s exactly the same for the weghts and correspondng to equal λ = λ, and therefore these weghts can be summed. It follows that a proof based on the assumpton that all λ are dfferent can easly be extended to the case of tes among λ 2,..., λ d 1. We thus assume that all egenvalues of A are dfferent: 0 < λ 1 < λ 2 <... < λ d. Note also that to prove the convergence to the two-dmensonal plane V for the sequence {x (k) / x (k) }, t s enough to prove the same property for the sequence { }, k. We dvde the rest of the proof nto four parts: we show () that the sequence (11) converges to a two-dmensonal plane, () that the drectons are eventually fxed, and () that they fx at u 1 and u d. Fnally, we show (v) the convergence to a two-pont cycle. It s convenent to consder the dscrete measures π k = { λ1... λ d 1... d whch place the weghts at the ponts λ, respectvely ( = 1,..., d). These measures carry all nformaton about the sequence (11). Let ψ( ) denote the operator correspondng to the applcaton of (11) to a measure π, so that π k+1 = ψ(π k ). (). Defne the moments µ m = µ m (π k ) = λ m dπ k (λ) = d =1 } λ m, m = 0, ±1, ±2,... (12) so that µ 0 = 1, µ 1 = d =1 λ, etc. Defne also the moment matrces M n = M n (π k ) by {M n },l = µ +l 2 (, l = 1,..., n) so that M 2 (π k ) = ( µ0 µ 1 µ 1 µ 2 ) and M 3 (π k ) = 4 µ 0 µ 1 µ 2 µ 1 µ 2 µ 3 µ 2 µ 3 µ 4.

5 Then the denomnator n the rght hand sde of (11) s D k = d d l=1 =1 λ 2 λ l Usng the updatng formula (11) we have Therefore D k+1 = d d l=1 =1 λ z (k+1) 2 λ l l = µ 2 µ 2 1 = det[m 2 (π k )]. z (k+1) l = d =1 λ 2 z (k+1) d =1 = 1 D k (µ 2 1µ 2 2µ 1 µ 3 + µ 4 ) 1 D 2 k (µ 3 1 2µ 1 µ 2 + µ 3 ) 2. λ z (k+1) D k+1 D k = 2µ 1µ 2 µ 3 + µ 2 µ 4 µ 2 1µ 4 µ 2 3 µ 2 2 (µ 2 µ 2 1) 2 = det[m 3(π k )] {det[m 2 (π k )]} 2. The condtons (6) transfer through the updatng formula (11) to 1 > 0 and d > 0 for every k = 1, 2,... From smple moment theory we have therefore that the matrx M 2 (π k ) s postve defnte for every k = 1, 2,... whch mples D k = det[m 2 (π k )] > 0. Also, M 3 (π k ) s a non-negatve defnte moment matrx and det[m 3 (π k )] 0. Thus D k+1 D k 0 whch means that the sequence {D k } s monotonously non-decreasng. Snce π k s a probablty measure wth bounded support, D k = det[m 2 (π k )] s bounded above by some constant D and therefore the sequence {D k } converges monotoncally to a lmt and D k+1 D k 0 when k. Moreover, det[m 3 (π k )] = (D k+1 D k )D 2 k (D k+1 D k )D 2, so that det[m 3 (π k )] 0 when k. Note that usng the Bnet-Cauchy lemma D k = det[m 2 (π k )] = < det[m 3 (π k )] = <<l (λ λ ) 2 D 1 > 0, 2 l (λ λ ) 2 (λ l λ ) 2 (λ λ l ) 2 0, k. (13) For a fxed teraton k defne the par ( k, k ) whch acheves max <. (If there are several such pars, take the smallest of them n, say, lexcographcal order.) Then for every k we have Therefore δ k D 1 D k k k, δ < k k < (λ λ ) 2. < 1 δ, δ < k < 1 δ, (14) 5

6 where δ = D 1 <(λ λ ) 2 > 0. (15) From (13) we have det[m 3 (π k )] k k (λ k λ k ) 2 (λ λ k ) 2 (λ λ k ) 2. k, k Snce all λ are dstnct, det[m 3 (π k )] 0 and k k δ > 0, we have 0, k. (16) k, k Ths fnshes part () of the proof. The nterpretaton s that although ( k, k ) depends on k the total weght assocated wth all other ponts tends to zero. (). We wll show n addton to (16), that there exsts an teraton number k such that for all k k the par ( k, k ) becomes fxed, that s, ( k, k ) = (, ) for some 1 < d and all k k. Let δd 1 ɛ = (λ d λ 1 ), 2 where δ s defned n (15). Accordng to (16), there exsts k = k (ɛ ) 1 such that the total weght assocated wth all other ponts dfferent from ( k, k ) s smaller than ɛ for all k k. Therefore, for / { k, k } and k k the updatng formula (11) gves From (14), z (k+1) k+1 z (k+1) = ( d=1 λ λ ) 2 < ɛ (λ d λ 1 ) 2 = δ. D k D 1 > δ and z (k+1) k+1 > δ, and therefore / { k, k } mples / { k+1, k+1 }. Ths proves that ( k, k ) = (, ) for some 1 < d and all k k. (). Let us prove that (, ) = (1, d). Recall agan that the assumpton (6) and the updatng formula (11) mply that 1 > 0 and d > 0 for all k. Assume that < d. Denote ɛ = mn{ɛ, δ mn (λ λ )/λ d }, 1 <<d and assume that k = k (ɛ ) k s such that ( k, k ) = (, ) and the total weght assocated wth all other ponts dfferent from (, ) s smaller than ɛ for all k k. The exstence of such a k follows from () and (). For convenence, rewrte the algorthm (11) n the form z (k+1) = (µ 1 λ ) 2 for = 1,..., d, (17) D k 6

7 and note that µ 1 = d =1 λ λ = λ + λ + λ, + λ + λ d λ + λ + ɛ λ d, δλ + (1 δ)λ + ɛ λ d = λ + δ(λ λ ) + ɛ λ d λ δ mn (λ λ ) + ɛ λ d λ. 1 <<d Therefore (17) mples that for all k k z (k+1) d d = (λ d µ 1 ) 2 D k > (λ µ 1 ) 2 D k = z(k+1). We have arrved at a contradcton snce the sequence { } s bounded from below by δ > 0 whle the sequence { d } tends to zero. Thus, = d and analogously = 1. (v). Fnally, let D denote the lmt lm k D k dscussed n (). There are only two dscrete measures π 1, π 2 wth nonzero weghts on λ 1 and λ d and such that namely wth det[m 2 (π 1 )] = det[m 2 (π 2 )] = D, { } π 1 λ1 λ = d, π 2 = p 1 p p = { λ1 λ d 1 p p D (λ d λ 1 ) 2. }, (18) Note that ψ(π 1 ) = π 2 and ψ(π 2 ) = π 1. Therefore, convergence of D k to D mples convergence of π k to the lmtng cycle π 1 π 2 π 1 Remark 1 Theorem 1 obvously generalzes to the case where (6) may not be satsfed. The algorthm then attracts to a two-dmensonal plane V and {v (k) } converges to a twopont cycle. V s defned by the egenvectors u, u assocated wth the smallest and largest egenvalues such that u T v (1) 0, u T v (1) 0. For the sake of smplcty of notatons, we shall assume n what follows that (6) s satsfed, and therefore u = u 1, u = u d. Although the lmtng behavour of the algorthm s smple, ts behavour en route to the attractor s farly complcated and presents a fractal structure. Fgure 1 shows the proecton onto the plane (v 1, v 3 ) of the regon of attracton for v (k) to a small neghbourhood (radus < 0.02) of the pont v = (0.9606, 0, ), wth d = 3, λ 1 = 1, λ 2 = 2, 7

8 λ 3 = 4. It llustrates the dffculty of predctng the lmtng behavour, and thus the aymptotc rate of convergence, see Secton 4, as a functon of the startng pont. Possble locaton of Fgure 1 In Fgure 2, the grey level of startng ponts on the unt sphere depends on the lmtng value of p n (18). Agan, ths llustrates the complexty of the behavour of the algorthm on the way to the attractor. Possble locaton of Fgure 2 3 Stablty of attractors From Theorem 1, the algorthm attracts to two conugate ponts on the crcle { v v d 2 = 1}, characterzed by the dscrete measure (18). However, some values of p correspond to unstable ponts. We shall use the followng defnton of stablty, see [?] p Defnton 1 A fxed pont π for a mappng T ( ) s called stable f ɛ > 0, α > 0 such that π 0 for whch π 0 π < α, T n (π 0 ) π < ɛ for all n > 0. A fxed pont π s unstable f t s not stable. Consder a two-step teraton for, 1 d: = z (k+1) ( d =1 λ z (k+1) λ ) 2 z (k+2) D k+1 = (µ 1 λ ) 2 1 D k D k+1 ( µ 3 1 2µ 1 µ 2 + µ 3 D k λ ) 2, (19) and the correspondng transformaton ψ 2 ( ) defned by π k+2 = ψ 2 (π k ). Snce d =1 = 1 for all k, we substtute 1 d 1 =1 for d n (19). Ths defnes an operator φ( ) : S d 1 S d 1 on the (d 1)-dmensonal canoncal smplex d 1 S d 1 = {z = (z 1,..., z d 1 ) z 0, z 1}, whch maps ( 1,..., d 1 ) to (z(k+2) 1,..., z (k+2) d 1 ). Studyng the propertes of ψ2 ( ) s equvalent to studyng the propertes of φ( ). =1 8

9 Theorem 2 () Stablty. All ponts n the nterval I S defned by where I S =] 1 2 s(λ ), s(λ )[, s(λ) = (λ d λ) 2 + (λ 1 λ) 2 2(λ d λ 1 ) and s such that λ λ 1+λ d 2 s mnmum over all λ s, = 2,..., d 1, are stable. () Instablty. All ponts n the set I U defned by, are unstable. I U = [0, 1 2 s(λ )[ ]1 2 + s(λ ), 1] Proof. () Stablty. Take p I S and consder the fxed pont z(p) = (p, 0,..., 0) for φ( ), z(p) S d 1. Assume that β 1, = 2,..., d 1 and 1 p β 1, that s z(p) β 1. Then z (k+2) = f(λ, p)[1 + O(β 1 )], β 1 0, (20) where wth f(λ, p) = (µ 1 λ) 2 [(µ 1) 3 2µ 1µ 2 + µ 3 D(p)λ] 2 [D(p)] 4 µ m = λ m 1 p + λ m d (1 p), m = 1, 2, 3, D(p) = p(1 p)(λ d λ 1 ) 2. (21) Then, p I S mples f(λ, p) < 1, = 2,..., d 1. Defne f (p) = max {2,...,d 1} f(λ, p) and let K be any constant such that f (p) < K < 1. Then, from (20): and thus β 0 β < β 0, z (k+2) β < β 0, z (k+2m) K, = 2,..., d 1, K m, = 2,..., d 1, m 1. (22) Now, (13) mples det[m 3 (π k )] Cβ 1, wth C some postve constant. Therefore, D k+1 D k Cβ 1 D 2 k. (23) 9

10 Smlarly, z (k+1) z(1 p) β 2 mples D k+2 D k+1 Cβ 2 D 2 k+1 Cβ 2 D 2 k. (24) Snce z(p) β z (k+1) z(1 p) Hβ, wth H some postve constant, (23,24) mply D k+1 D k C β, D k+2 D k+1 C β when z(p) β, wth C = (C/D 2 k) max{1, H}. Ths gves Choosng β < β 0, one then obtan from (22) Therefore, l > 0, D k+2 D k 2C β. D k+2m+2 D k+2m 2C K m β. D k+2l D k 2C β 1 K = O(β), β 0. Moreover, z(p) β D k D(p) Aβ, wth A a postve constant and D(p) gven by (21). Therefore, l > 0, Ths, together wth (22) mples D k+2l D(p) ( 2C 1 K + A ) l > 0, z (k+2l) 1 p < Lβ, L > 1. (25) Fnally, for any ɛ > 0, defne α = mn{ɛ/l, β 0 /L}, (22) and (25) then mply the stablty of the fxed pont z(p). () Instablty. The Jacoban matrx J φ of φ( ) at ponts z(p) = (p, 0,..., 0) can be computed analytcally, and s gven by: (J φ ) = φ (z) = z z=z(p) 1 f = = 1 (λ λ 1 )(λ d λ ) 2 [2λ d (1 p)+2λ 1 p λ 1 λ ] p(1 p) 2 (λ d λ 1 ) 4 f > 1 and = 1 β. [λ d (1 p)+λ 1 p λ ] 2 [λ d p+λ 1 (1 p) λ ] 2 p 2 (1 p) 2 (λ d λ 1 ) 4 f = 0 otherwse. One can easlly check that for p I U, (J φ ) > 1 for at least one n {2,..., d 1}. Snce the (J φ ) s are egenvalues of J φ, Theorem n [?] ndcates that z(p) s unstable. 10

11 Note that λ 1,..., λ d the stablty nterval I S contans the nterval ] , [ ] , [. 2 When d = 3, numercal smulatons show that for any ntal densty of x (1) n IR d assocated wth a densty of z (1) on S d 1 reasonably spread, the densty of attractors z(p) characterzed by p can be approxmated by the densty ϕ(p) = C log mn{1, (J φ ) 22 } = { C log(jφ ) 22 f p I A 0 otherwse, where C s a normalsaton constant. Fgure 3 shows the emprcal densty of attractors (full lne) together wth ϕ(p) (dashed lne) n the case λ 1 = 1, λ 2 = 4, λ 3 = 10. The support of ths densty concdes wth the stablty nterval I S. Possble locaton of Fgure 3 When d > 3, the densty of attractors depends on the ntal densty of x (1). 4 Rate of convergence Defne the convergence rate at teraton k by r k = f(x(k+1) ) f(x (k) ), (26) where f(x) s gven by (5). Wthout any loss of generalty, one can take f(x) = d =1 λ x 2. Rewrtng r k n terms of defned by (10), one gets r k = 1 1 µ 1 µ 1, where µ m s defned by (12). Defne the asymptotc rate as R = R(x (1), x ) = lm 1/k k r k =1. (27) Generally, R depends on the ntal pont x (1) and the optmal pont x. Theorem 1 mples that for any fxed x and almost all x (1) the asymptotc rate depends only on the attractor (18) and s gven by ( f(x (k+2) ) 1/2 ) R(p) = f(x (k) ) 11

12 where x (k) corresponds to π 1 or π 2, see (18). Ths gves R(p) = p(1 p)(ρ 1) 2 [p + ρ(1 p)][(1 p) + ρp], wth ρ = λ d /λ 1 the condton number of the matrx A. The functon R(p) s symmetrc wth respect to 1/2 and monotonously ncreasng from 0 to 1/2. The worst asymptotc rate s thus obtaned at p = 1/2: R max = ( ) 2 ρ 1, ρ + 1 see (4). Note that from Kantorovch nequalty, see [?], p. 151, µ 1 µ 1 (1 + ρ) 2 /(4ρ), and therefore x (k), r k R max. The worst rate s thus acheved only when x (k) 1 = ±ρx (k) d, x (k) 2 = = x (k) d 1 = 0. Consder now another convergence rate, defned by where R = lm Rewrtng r k n terms of, one gets 1/k k r k =1 r k = x(k+1) 2 x (k) 2. r k = 1 2µ 1 µ 1 µ µ 2 1µ 2. One can easlly check that for almost all x (1) the asymptotc rate R s equal to R(p), where p defnes the attractor. 5 Steepest descent wth relaxaton The ntroducton of a relaxaton coeffcent γ, wth 0 < γ < 1, n the steepest-descent algorthm totally changes ts behavour. The algorthm (7) then becomes x (k+1) = x (k) γ (g(k) ) T g (k) (g (k) ) T Ag (k) g(k). For fxed A, dependng on the value of γ, the renormalzed process ether attracts to perodc orbts (the same for almost all startng ponts) or exhbts a chaotc behavour. Fgures 4 (respectvely 5) presents the classcal perod-doublng phenomenon n the case d = 2 when λ 1 = 1 and λ 2 = 4 (respectvely λ 2 = 10). Fgures 6 (respectvely 7) gve the asymptotc rate (27) as a functon of γ n the same stuaton. 12.

13 Possble locaton of Fgure 4 Possble locaton of Fgure 5 Possble locaton of Fgure 6 Possble locaton of Fgure 7 We get now nstead of (26): r k (γ) = 1 γ(2 γ) µ 1 µ 1. Note that from Kantorovch nequalty the worst value of the rate s 4ρ 1 γ(2 γ) (1 + ρ) > R 2 max, f γ < 1. However, numercal results show that for γ large enough the asymptotc rate s sgnfcantly better than R max. A detaled analyss of the 2-dmensonal case gves the followng results. () If 0 < γ 2 2ργ, the process attracts to the fxed pont p = 1 and R = R(γ) = 1. ρ+1 ρ+1 2 () If < γ 4ρ, the process attracts to the fxed pont p = 2ρ γ(ρ+1), and ρ+1 (ρ+1) 2 2(ρ 1) R(γ) = R max. () If 4ρ < γ 2( 2+1)ρ (ρ+1) 2 (ρ+1) 2, the process attracts to the 2-pont cycle (p 1, p 2 ), wth p 1,2 = 2ρ γ(ρ + 1) ± γ(γρ 2 4ρ + 2γρ + γ), 2(ρ 1) and R(γ) = 1 γ. (v) For larger values of γ one observes a classcal perod-doublng phenomenon, see Fgures 4 and 5. 13

14 (v) If ρ > , the process attracts agan to a 2-pont cycle for values of γ larger than γ ρ = 8ρ, see Fgure 5. For the lmtng case γ = γ (ρ+1) 2 ρ, the cycle s gven by (p 1, p 2), wth p 1,2 = ρ ( ρ 2 2ρ + 5 ± 2 (ρ 2 2ρ + 5)(5ρ 2 2ρ + 1) ), (ρ 1)(ρ + 1) 3 and the assocated asymptotc rate s R(γ ρ ) = ρ2 6ρ + 1 ρ 2 1 In hgher dmensons, repeated numercal trals show that the process typcally no longer attracts to the 2-dmensonal plane spanned by (u 1, u d ). Fgure 8 presents the proecton of the attractor of on the plane (z 1, z 3 ) n the case d = 3, λ 1 = 1, λ 2 = 2, λ 3 = 4 and γ = Such a pcture s typcal for d > 2. Possble locaton of Fgure 8. 14

15 Fgure captons Fgure 1. Proecton of the regon of attracton for v (k) to a small neghbourhood (radus < 0.02) of the pont v = (0.9606, 0, ) onto the plane (v 1, v 3 ) (d = 3, λ 1 = 1, λ 2 = 2, λ 3 = 4). Fgure 2. Startng ponts on the unt sphere colored a a functon of the lmtng value of p n (18) (d = 3, λ 1 = 1, λ 2 = 2, λ 3 = 4). Fgure 3. Emprcal densty of attractors (full lne) and ϕ(p) (dashed lne) (d = 3, λ 1 = 1, λ 2 = 4, λ 3 = 10). Fgure 4. Attractors for z (1) as a functon of γ (d = 2, λ 1 = 1, λ 2 = 4). Fgure 5. Attractors for z (1) as a functon of γ (d = 2, λ 1 = 1, λ 2 = 10). Fgure 6. Asymptotc rate (27) as a functon of γ (d = 2, λ 1 = 1, λ 2 = 4). Fgure 7. Asymptotc rate (27) as a functon of γ (d = 2, λ 1 = 1, λ 2 = 10). Fgure 8. Proecton of the attractor of on the plane (z 1, z 3 ) (d = 3, λ 1 = 1, λ 2 = 2, λ 3 = 4, γ = 0.97). 15

APPENDIX A Some Linear Algebra

APPENDIX A Some Linear Algebra APPENDIX A Some Lnear Algebra The collecton of m, n matrces A.1 Matrces a 1,1,..., a 1,n A = a m,1,..., a m,n wth real elements a,j s denoted by R m,n. If n = 1 then A s called a column vector. Smlarly,

More information

The equation of motion of a dynamical system is given by a set of differential equations. That is (1)

The equation of motion of a dynamical system is given by a set of differential equations. That is (1) Dynamcal Systems Many engneerng and natural systems are dynamcal systems. For example a pendulum s a dynamcal system. State l The state of the dynamcal system specfes t condtons. For a pendulum n the absence

More information

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X Statstcs 1: Probablty Theory II 37 3 EPECTATION OF SEVERAL RANDOM VARIABLES As n Probablty Theory I, the nterest n most stuatons les not on the actual dstrbuton of a random vector, but rather on a number

More information

Yong Joon Ryang. 1. Introduction Consider the multicommodity transportation problem with convex quadratic cost function. 1 2 (x x0 ) T Q(x x 0 )

Yong Joon Ryang. 1. Introduction Consider the multicommodity transportation problem with convex quadratic cost function. 1 2 (x x0 ) T Q(x x 0 ) Kangweon-Kyungk Math. Jour. 4 1996), No. 1, pp. 7 16 AN ITERATIVE ROW-ACTION METHOD FOR MULTICOMMODITY TRANSPORTATION PROBLEMS Yong Joon Ryang Abstract. The optmzaton problems wth quadratc constrants often

More information

MMA and GCMMA two methods for nonlinear optimization

MMA and GCMMA two methods for nonlinear optimization MMA and GCMMA two methods for nonlnear optmzaton Krster Svanberg Optmzaton and Systems Theory, KTH, Stockholm, Sweden. krlle@math.kth.se Ths note descrbes the algorthms used n the author s 2007 mplementatons

More information

Affine transformations and convexity

Affine transformations and convexity Affne transformatons and convexty The purpose of ths document s to prove some basc propertes of affne transformatons nvolvng convex sets. Here are a few onlne references for background nformaton: http://math.ucr.edu/

More information

Salmon: Lectures on partial differential equations. Consider the general linear, second-order PDE in the form. ,x 2

Salmon: Lectures on partial differential equations. Consider the general linear, second-order PDE in the form. ,x 2 Salmon: Lectures on partal dfferental equatons 5. Classfcaton of second-order equatons There are general methods for classfyng hgher-order partal dfferental equatons. One s very general (applyng even to

More information

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems Numercal Analyss by Dr. Anta Pal Assstant Professor Department of Mathematcs Natonal Insttute of Technology Durgapur Durgapur-713209 emal: anta.bue@gmal.com 1 . Chapter 5 Soluton of System of Lnear Equatons

More information

Lectures - Week 4 Matrix norms, Conditioning, Vector Spaces, Linear Independence, Spanning sets and Basis, Null space and Range of a Matrix

Lectures - Week 4 Matrix norms, Conditioning, Vector Spaces, Linear Independence, Spanning sets and Basis, Null space and Range of a Matrix Lectures - Week 4 Matrx norms, Condtonng, Vector Spaces, Lnear Independence, Spannng sets and Bass, Null space and Range of a Matrx Matrx Norms Now we turn to assocatng a number to each matrx. We could

More information

Assortment Optimization under MNL

Assortment Optimization under MNL Assortment Optmzaton under MNL Haotan Song Aprl 30, 2017 1 Introducton The assortment optmzaton problem ams to fnd the revenue-maxmzng assortment of products to offer when the prces of products are fxed.

More information

Solutions to exam in SF1811 Optimization, Jan 14, 2015

Solutions to exam in SF1811 Optimization, Jan 14, 2015 Solutons to exam n SF8 Optmzaton, Jan 4, 25 3 3 O------O -4 \ / \ / The network: \/ where all lnks go from left to rght. /\ / \ / \ 6 O------O -5 2 4.(a) Let x = ( x 3, x 4, x 23, x 24 ) T, where the varable

More information

More metrics on cartesian products

More metrics on cartesian products More metrcs on cartesan products If (X, d ) are metrc spaces for 1 n, then n Secton II4 of the lecture notes we defned three metrcs on X whose underlyng topologes are the product topology The purpose of

More information

The Order Relation and Trace Inequalities for. Hermitian Operators

The Order Relation and Trace Inequalities for. Hermitian Operators Internatonal Mathematcal Forum, Vol 3, 08, no, 507-57 HIKARI Ltd, wwwm-hkarcom https://doorg/0988/mf088055 The Order Relaton and Trace Inequaltes for Hermtan Operators Y Huang School of Informaton Scence

More information

Vector Norms. Chapter 7 Iterative Techniques in Matrix Algebra. Cauchy-Bunyakovsky-Schwarz Inequality for Sums. Distances. Convergence.

Vector Norms. Chapter 7 Iterative Techniques in Matrix Algebra. Cauchy-Bunyakovsky-Schwarz Inequality for Sums. Distances. Convergence. Vector Norms Chapter 7 Iteratve Technques n Matrx Algebra Per-Olof Persson persson@berkeley.edu Department of Mathematcs Unversty of Calforna, Berkeley Math 128B Numercal Analyss Defnton A vector norm

More information

Some modelling aspects for the Matlab implementation of MMA

Some modelling aspects for the Matlab implementation of MMA Some modellng aspects for the Matlab mplementaton of MMA Krster Svanberg krlle@math.kth.se Optmzaton and Systems Theory Department of Mathematcs KTH, SE 10044 Stockholm September 2004 1. Consdered optmzaton

More information

Stanford University CS359G: Graph Partitioning and Expanders Handout 4 Luca Trevisan January 13, 2011

Stanford University CS359G: Graph Partitioning and Expanders Handout 4 Luca Trevisan January 13, 2011 Stanford Unversty CS359G: Graph Parttonng and Expanders Handout 4 Luca Trevsan January 3, 0 Lecture 4 In whch we prove the dffcult drecton of Cheeger s nequalty. As n the past lectures, consder an undrected

More information

U.C. Berkeley CS294: Beyond Worst-Case Analysis Luca Trevisan September 5, 2017

U.C. Berkeley CS294: Beyond Worst-Case Analysis Luca Trevisan September 5, 2017 U.C. Berkeley CS94: Beyond Worst-Case Analyss Handout 4s Luca Trevsan September 5, 07 Summary of Lecture 4 In whch we ntroduce semdefnte programmng and apply t to Max Cut. Semdefnte Programmng Recall that

More information

Lecture 12: Discrete Laplacian

Lecture 12: Discrete Laplacian Lecture 12: Dscrete Laplacan Scrbe: Tanye Lu Our goal s to come up wth a dscrete verson of Laplacan operator for trangulated surfaces, so that we can use t n practce to solve related problems We are mostly

More information

The lower and upper bounds on Perron root of nonnegative irreducible matrices

The lower and upper bounds on Perron root of nonnegative irreducible matrices Journal of Computatonal Appled Mathematcs 217 (2008) 259 267 wwwelsevercom/locate/cam The lower upper bounds on Perron root of nonnegatve rreducble matrces Guang-Xn Huang a,, Feng Yn b,keguo a a College

More information

Problem Set 9 Solutions

Problem Set 9 Solutions Desgn and Analyss of Algorthms May 4, 2015 Massachusetts Insttute of Technology 6.046J/18.410J Profs. Erk Demane, Srn Devadas, and Nancy Lynch Problem Set 9 Solutons Problem Set 9 Solutons Ths problem

More information

Lecture Notes on Linear Regression

Lecture Notes on Linear Regression Lecture Notes on Lnear Regresson Feng L fl@sdueducn Shandong Unversty, Chna Lnear Regresson Problem In regresson problem, we am at predct a contnuous target value gven an nput feature vector We assume

More information

Difference Equations

Difference Equations Dfference Equatons c Jan Vrbk 1 Bascs Suppose a sequence of numbers, say a 0,a 1,a,a 3,... s defned by a certan general relatonshp between, say, three consecutve values of the sequence, e.g. a + +3a +1

More information

STAT 309: MATHEMATICAL COMPUTATIONS I FALL 2018 LECTURE 16

STAT 309: MATHEMATICAL COMPUTATIONS I FALL 2018 LECTURE 16 STAT 39: MATHEMATICAL COMPUTATIONS I FALL 218 LECTURE 16 1 why teratve methods f we have a lnear system Ax = b where A s very, very large but s ether sparse or structured (eg, banded, Toepltz, banded plus

More information

The Geometry of Logit and Probit

The Geometry of Logit and Probit The Geometry of Logt and Probt Ths short note s meant as a supplement to Chapters and 3 of Spatal Models of Parlamentary Votng and the notaton and reference to fgures n the text below s to those two chapters.

More information

Graph Reconstruction by Permutations

Graph Reconstruction by Permutations Graph Reconstructon by Permutatons Perre Ille and Wllam Kocay* Insttut de Mathémathques de Lumny CNRS UMR 6206 163 avenue de Lumny, Case 907 13288 Marselle Cedex 9, France e-mal: lle@ml.unv-mrs.fr Computer

More information

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur Module 3 LOSSY IMAGE COMPRESSION SYSTEMS Verson ECE IIT, Kharagpur Lesson 6 Theory of Quantzaton Verson ECE IIT, Kharagpur Instructonal Objectves At the end of ths lesson, the students should be able to:

More information

Randić Energy and Randić Estrada Index of a Graph

Randić Energy and Randić Estrada Index of a Graph EUROPEAN JOURNAL OF PURE AND APPLIED MATHEMATICS Vol. 5, No., 202, 88-96 ISSN 307-5543 www.ejpam.com SPECIAL ISSUE FOR THE INTERNATIONAL CONFERENCE ON APPLIED ANALYSIS AND ALGEBRA 29 JUNE -02JULY 20, ISTANBUL

More information

Foundations of Arithmetic

Foundations of Arithmetic Foundatons of Arthmetc Notaton We shall denote the sum and product of numbers n the usual notaton as a 2 + a 2 + a 3 + + a = a, a 1 a 2 a 3 a = a The notaton a b means a dvdes b,.e. ac = b where c s an

More information

Inductance Calculation for Conductors of Arbitrary Shape

Inductance Calculation for Conductors of Arbitrary Shape CRYO/02/028 Aprl 5, 2002 Inductance Calculaton for Conductors of Arbtrary Shape L. Bottura Dstrbuton: Internal Summary In ths note we descrbe a method for the numercal calculaton of nductances among conductors

More information

Google PageRank with Stochastic Matrix

Google PageRank with Stochastic Matrix Google PageRank wth Stochastc Matrx Md. Sharq, Puranjt Sanyal, Samk Mtra (M.Sc. Applcatons of Mathematcs) Dscrete Tme Markov Chan Let S be a countable set (usually S s a subset of Z or Z d or R or R d

More information

6) Derivatives, gradients and Hessian matrices

6) Derivatives, gradients and Hessian matrices 30C00300 Mathematcal Methods for Economsts (6 cr) 6) Dervatves, gradents and Hessan matrces Smon & Blume chapters: 14, 15 Sldes by: Tmo Kuosmanen 1 Outlne Defnton of dervatve functon Dervatve notatons

More information

Linear Approximation with Regularization and Moving Least Squares

Linear Approximation with Regularization and Moving Least Squares Lnear Approxmaton wth Regularzaton and Movng Least Squares Igor Grešovn May 007 Revson 4.6 (Revson : March 004). 5 4 3 0.5 3 3.5 4 Contents: Lnear Fttng...4. Weghted Least Squares n Functon Approxmaton...

More information

TAIL BOUNDS FOR SUMS OF GEOMETRIC AND EXPONENTIAL VARIABLES

TAIL BOUNDS FOR SUMS OF GEOMETRIC AND EXPONENTIAL VARIABLES TAIL BOUNDS FOR SUMS OF GEOMETRIC AND EXPONENTIAL VARIABLES SVANTE JANSON Abstract. We gve explct bounds for the tal probabltes for sums of ndependent geometrc or exponental varables, possbly wth dfferent

More information

Supplement: Proofs and Technical Details for The Solution Path of the Generalized Lasso

Supplement: Proofs and Technical Details for The Solution Path of the Generalized Lasso Supplement: Proofs and Techncal Detals for The Soluton Path of the Generalzed Lasso Ryan J. Tbshran Jonathan Taylor In ths document we gve supplementary detals to the paper The Soluton Path of the Generalzed

More information

Inexact Newton Methods for Inverse Eigenvalue Problems

Inexact Newton Methods for Inverse Eigenvalue Problems Inexact Newton Methods for Inverse Egenvalue Problems Zheng-jan Ba Abstract In ths paper, we survey some of the latest development n usng nexact Newton-lke methods for solvng nverse egenvalue problems.

More information

Lecture 3. Ax x i a i. i i

Lecture 3. Ax x i a i. i i 18.409 The Behavor of Algorthms n Practce 2/14/2 Lecturer: Dan Spelman Lecture 3 Scrbe: Arvnd Sankar 1 Largest sngular value In order to bound the condton number, we need an upper bound on the largest

More information

MATH 241B FUNCTIONAL ANALYSIS - NOTES EXAMPLES OF C ALGEBRAS

MATH 241B FUNCTIONAL ANALYSIS - NOTES EXAMPLES OF C ALGEBRAS MATH 241B FUNCTIONAL ANALYSIS - NOTES EXAMPLES OF C ALGEBRAS These are nformal notes whch cover some of the materal whch s not n the course book. The man purpose s to gve a number of nontrval examples

More information

Appendix B. The Finite Difference Scheme

Appendix B. The Finite Difference Scheme 140 APPENDIXES Appendx B. The Fnte Dfference Scheme In ths appendx we present numercal technques whch are used to approxmate solutons of system 3.1 3.3. A comprehensve treatment of theoretcal and mplementaton

More information

2.3 Nilpotent endomorphisms

2.3 Nilpotent endomorphisms s a block dagonal matrx, wth A Mat dm U (C) In fact, we can assume that B = B 1 B k, wth B an ordered bass of U, and that A = [f U ] B, where f U : U U s the restrcton of f to U 40 23 Nlpotent endomorphsms

More information

Inner Product. Euclidean Space. Orthonormal Basis. Orthogonal

Inner Product. Euclidean Space. Orthonormal Basis. Orthogonal Inner Product Defnton 1 () A Eucldean space s a fnte-dmensonal vector space over the reals R, wth an nner product,. Defnton 2 (Inner Product) An nner product, on a real vector space X s a symmetrc, blnear,

More information

Spectral Graph Theory and its Applications September 16, Lecture 5

Spectral Graph Theory and its Applications September 16, Lecture 5 Spectral Graph Theory and ts Applcatons September 16, 2004 Lecturer: Danel A. Spelman Lecture 5 5.1 Introducton In ths lecture, we wll prove the followng theorem: Theorem 5.1.1. Let G be a planar graph

More information

ρ some λ THE INVERSE POWER METHOD (or INVERSE ITERATION) , for , or (more usually) to

ρ some λ THE INVERSE POWER METHOD (or INVERSE ITERATION) , for , or (more usually) to THE INVERSE POWER METHOD (or INVERSE ITERATION) -- applcaton of the Power method to A some fxed constant ρ (whch s called a shft), x λ ρ If the egenpars of A are { ( λ, x ) } ( ), or (more usually) to,

More information

General viscosity iterative method for a sequence of quasi-nonexpansive mappings

General viscosity iterative method for a sequence of quasi-nonexpansive mappings Avalable onlne at www.tjnsa.com J. Nonlnear Sc. Appl. 9 (2016), 5672 5682 Research Artcle General vscosty teratve method for a sequence of quas-nonexpansve mappngs Cuje Zhang, Ynan Wang College of Scence,

More information

Singular Value Decomposition: Theory and Applications

Singular Value Decomposition: Theory and Applications Sngular Value Decomposton: Theory and Applcatons Danel Khashab Sprng 2015 Last Update: March 2, 2015 1 Introducton A = UDV where columns of U and V are orthonormal and matrx D s dagonal wth postve real

More information

Curvature and isoperimetric inequality

Curvature and isoperimetric inequality urvature and sopermetrc nequalty Julà ufí, Agustí Reventós, arlos J Rodríguez Abstract We prove an nequalty nvolvng the length of a plane curve and the ntegral of ts radus of curvature, that has as a consequence

More information

Week3, Chapter 4. Position and Displacement. Motion in Two Dimensions. Instantaneous Velocity. Average Velocity

Week3, Chapter 4. Position and Displacement. Motion in Two Dimensions. Instantaneous Velocity. Average Velocity Week3, Chapter 4 Moton n Two Dmensons Lecture Quz A partcle confned to moton along the x axs moves wth constant acceleraton from x =.0 m to x = 8.0 m durng a 1-s tme nterval. The velocty of the partcle

More information

U.C. Berkeley CS294: Spectral Methods and Expanders Handout 8 Luca Trevisan February 17, 2016

U.C. Berkeley CS294: Spectral Methods and Expanders Handout 8 Luca Trevisan February 17, 2016 U.C. Berkeley CS94: Spectral Methods and Expanders Handout 8 Luca Trevsan February 7, 06 Lecture 8: Spectral Algorthms Wrap-up In whch we talk about even more generalzatons of Cheeger s nequaltes, and

More information

Module 9. Lecture 6. Duality in Assignment Problems

Module 9. Lecture 6. Duality in Assignment Problems Module 9 1 Lecture 6 Dualty n Assgnment Problems In ths lecture we attempt to answer few other mportant questons posed n earler lecture for (AP) and see how some of them can be explaned through the concept

More information

A New Refinement of Jacobi Method for Solution of Linear System Equations AX=b

A New Refinement of Jacobi Method for Solution of Linear System Equations AX=b Int J Contemp Math Scences, Vol 3, 28, no 17, 819-827 A New Refnement of Jacob Method for Soluton of Lnear System Equatons AX=b F Naem Dafchah Department of Mathematcs, Faculty of Scences Unversty of Gulan,

More information

Convexity preserving interpolation by splines of arbitrary degree

Convexity preserving interpolation by splines of arbitrary degree Computer Scence Journal of Moldova, vol.18, no.1(52), 2010 Convexty preservng nterpolaton by splnes of arbtrary degree Igor Verlan Abstract In the present paper an algorthm of C 2 nterpolaton of dscrete

More information

Using T.O.M to Estimate Parameter of distributions that have not Single Exponential Family

Using T.O.M to Estimate Parameter of distributions that have not Single Exponential Family IOSR Journal of Mathematcs IOSR-JM) ISSN: 2278-5728. Volume 3, Issue 3 Sep-Oct. 202), PP 44-48 www.osrjournals.org Usng T.O.M to Estmate Parameter of dstrbutons that have not Sngle Exponental Famly Jubran

More information

Perron Vectors of an Irreducible Nonnegative Interval Matrix

Perron Vectors of an Irreducible Nonnegative Interval Matrix Perron Vectors of an Irreducble Nonnegatve Interval Matrx Jr Rohn August 4 2005 Abstract As s well known an rreducble nonnegatve matrx possesses a unquely determned Perron vector. As the man result of

More information

Feature Selection: Part 1

Feature Selection: Part 1 CSE 546: Machne Learnng Lecture 5 Feature Selecton: Part 1 Instructor: Sham Kakade 1 Regresson n the hgh dmensonal settng How do we learn when the number of features d s greater than the sample sze n?

More information

Approximate D-optimal designs of experiments on the convex hull of a finite set of information matrices

Approximate D-optimal designs of experiments on the convex hull of a finite set of information matrices Approxmate D-optmal desgns of experments on the convex hull of a fnte set of nformaton matrces Radoslav Harman, Mára Trnovská Department of Appled Mathematcs and Statstcs Faculty of Mathematcs, Physcs

More information

Random Walks on Digraphs

Random Walks on Digraphs Random Walks on Dgraphs J. J. P. Veerman October 23, 27 Introducton Let V = {, n} be a vertex set and S a non-negatve row-stochastc matrx (.e. rows sum to ). V and S defne a dgraph G = G(V, S) and a drected

More information

Errors for Linear Systems

Errors for Linear Systems Errors for Lnear Systems When we solve a lnear system Ax b we often do not know A and b exactly, but have only approxmatons  and ˆb avalable. Then the best thng we can do s to solve ˆx ˆb exactly whch

More information

CHALMERS, GÖTEBORGS UNIVERSITET. SOLUTIONS to RE-EXAM for ARTIFICIAL NEURAL NETWORKS. COURSE CODES: FFR 135, FIM 720 GU, PhD

CHALMERS, GÖTEBORGS UNIVERSITET. SOLUTIONS to RE-EXAM for ARTIFICIAL NEURAL NETWORKS. COURSE CODES: FFR 135, FIM 720 GU, PhD CHALMERS, GÖTEBORGS UNIVERSITET SOLUTIONS to RE-EXAM for ARTIFICIAL NEURAL NETWORKS COURSE CODES: FFR 35, FIM 72 GU, PhD Tme: Place: Teachers: Allowed materal: Not allowed: January 2, 28, at 8 3 2 3 SB

More information

Time-Varying Systems and Computations Lecture 6

Time-Varying Systems and Computations Lecture 6 Tme-Varyng Systems and Computatons Lecture 6 Klaus Depold 14. Januar 2014 The Kalman Flter The Kalman estmaton flter attempts to estmate the actual state of an unknown dscrete dynamcal system, gven nosy

More information

Another converse of Jensen s inequality

Another converse of Jensen s inequality Another converse of Jensen s nequalty Slavko Smc Abstract. We gve the best possble global bounds for a form of dscrete Jensen s nequalty. By some examples ts frutfulness s shown. 1. Introducton Throughout

More information

find (x): given element x, return the canonical element of the set containing x;

find (x): given element x, return the canonical element of the set containing x; COS 43 Sprng, 009 Dsjont Set Unon Problem: Mantan a collecton of dsjont sets. Two operatons: fnd the set contanng a gven element; unte two sets nto one (destructvely). Approach: Canoncal element method:

More information

Appendix B. Criterion of Riemann-Stieltjes Integrability

Appendix B. Criterion of Riemann-Stieltjes Integrability Appendx B. Crteron of Remann-Steltes Integrablty Ths note s complementary to [R, Ch. 6] and [T, Sec. 3.5]. The man result of ths note s Theorem B.3, whch provdes the necessary and suffcent condtons for

More information

On the Interval Zoro Symmetric Single-step Procedure for Simultaneous Finding of Polynomial Zeros

On the Interval Zoro Symmetric Single-step Procedure for Simultaneous Finding of Polynomial Zeros Appled Mathematcal Scences, Vol. 5, 2011, no. 75, 3693-3706 On the Interval Zoro Symmetrc Sngle-step Procedure for Smultaneous Fndng of Polynomal Zeros S. F. M. Rusl, M. Mons, M. A. Hassan and W. J. Leong

More information

Additional Codes using Finite Difference Method. 1 HJB Equation for Consumption-Saving Problem Without Uncertainty

Additional Codes using Finite Difference Method. 1 HJB Equation for Consumption-Saving Problem Without Uncertainty Addtonal Codes usng Fnte Dfference Method Benamn Moll 1 HJB Equaton for Consumpton-Savng Problem Wthout Uncertanty Before consderng the case wth stochastc ncome n http://www.prnceton.edu/~moll/ HACTproect/HACT_Numercal_Appendx.pdf,

More information

Dynamic Systems on Graphs

Dynamic Systems on Graphs Prepared by F.L. Lews Updated: Saturday, February 06, 200 Dynamc Systems on Graphs Control Graphs and Consensus A network s a set of nodes that collaborates to acheve what each cannot acheve alone. A network,

More information

Composite Hypotheses testing

Composite Hypotheses testing Composte ypotheses testng In many hypothess testng problems there are many possble dstrbutons that can occur under each of the hypotheses. The output of the source s a set of parameters (ponts n a parameter

More information

The Second Eigenvalue of Planar Graphs

The Second Eigenvalue of Planar Graphs Spectral Graph Theory Lecture 20 The Second Egenvalue of Planar Graphs Danel A. Spelman November 11, 2015 Dsclamer These notes are not necessarly an accurate representaton of what happened n class. The

More information

Introduction. - The Second Lyapunov Method. - The First Lyapunov Method

Introduction. - The Second Lyapunov Method. - The First Lyapunov Method Stablty Analyss A. Khak Sedgh Control Systems Group Faculty of Electrcal and Computer Engneerng K. N. Toos Unversty of Technology February 2009 1 Introducton Stablty s the most promnent characterstc of

More information

CSCE 790S Background Results

CSCE 790S Background Results CSCE 790S Background Results Stephen A. Fenner September 8, 011 Abstract These results are background to the course CSCE 790S/CSCE 790B, Quantum Computaton and Informaton (Sprng 007 and Fall 011). Each

More information

CHAPTER-5 INFORMATION MEASURE OF FUZZY MATRIX AND FUZZY BINARY RELATION

CHAPTER-5 INFORMATION MEASURE OF FUZZY MATRIX AND FUZZY BINARY RELATION CAPTER- INFORMATION MEASURE OF FUZZY MATRI AN FUZZY BINARY RELATION Introducton The basc concept of the fuzz matr theor s ver smple and can be appled to socal and natural stuatons A branch of fuzz matr

More information

Maximizing the number of nonnegative subsets

Maximizing the number of nonnegative subsets Maxmzng the number of nonnegatve subsets Noga Alon Hao Huang December 1, 213 Abstract Gven a set of n real numbers, f the sum of elements of every subset of sze larger than k s negatve, what s the maxmum

More information

CHAPTER 14 GENERAL PERTURBATION THEORY

CHAPTER 14 GENERAL PERTURBATION THEORY CHAPTER 4 GENERAL PERTURBATION THEORY 4 Introducton A partcle n orbt around a pont mass or a sphercally symmetrc mass dstrbuton s movng n a gravtatonal potental of the form GM / r In ths potental t moves

More information

Eigenvalues of Random Graphs

Eigenvalues of Random Graphs Spectral Graph Theory Lecture 2 Egenvalues of Random Graphs Danel A. Spelman November 4, 202 2. Introducton In ths lecture, we consder a random graph on n vertces n whch each edge s chosen to be n the

More information

Week 9 Chapter 10 Section 1-5

Week 9 Chapter 10 Section 1-5 Week 9 Chapter 10 Secton 1-5 Rotaton Rgd Object A rgd object s one that s nondeformable The relatve locatons of all partcles makng up the object reman constant All real objects are deformable to some extent,

More information

1 GSW Iterative Techniques for y = Ax

1 GSW Iterative Techniques for y = Ax 1 for y = A I m gong to cheat here. here are a lot of teratve technques that can be used to solve the general case of a set of smultaneous equatons (wrtten n the matr form as y = A), but ths chapter sn

More information

Power law and dimension of the maximum value for belief distribution with the max Deng entropy

Power law and dimension of the maximum value for belief distribution with the max Deng entropy Power law and dmenson of the maxmum value for belef dstrbuton wth the max Deng entropy Bngy Kang a, a College of Informaton Engneerng, Northwest A&F Unversty, Yanglng, Shaanx, 712100, Chna. Abstract Deng

More information

Chapter 22. Mathematical Chaos Stable and Unstable Manifolds Saddle Fixed Point

Chapter 22. Mathematical Chaos Stable and Unstable Manifolds Saddle Fixed Point Chapter 22 Mathematcal Chaos The sets generated as the long tme attractors of physcal dynamcal systems descrbed by ordnary dfferental equatons or dscrete evoluton equatons n the case of maps such as the

More information

A new construction of 3-separable matrices via an improved decoding of Macula s construction

A new construction of 3-separable matrices via an improved decoding of Macula s construction Dscrete Optmzaton 5 008 700 704 Contents lsts avalable at ScenceDrect Dscrete Optmzaton journal homepage: wwwelsevercom/locate/dsopt A new constructon of 3-separable matrces va an mproved decodng of Macula

More information

ELASTIC WAVE PROPAGATION IN A CONTINUOUS MEDIUM

ELASTIC WAVE PROPAGATION IN A CONTINUOUS MEDIUM ELASTIC WAVE PROPAGATION IN A CONTINUOUS MEDIUM An elastc wave s a deformaton of the body that travels throughout the body n all drectons. We can examne the deformaton over a perod of tme by fxng our look

More information

10.34 Fall 2015 Metropolis Monte Carlo Algorithm

10.34 Fall 2015 Metropolis Monte Carlo Algorithm 10.34 Fall 2015 Metropols Monte Carlo Algorthm The Metropols Monte Carlo method s very useful for calculatng manydmensonal ntegraton. For e.g. n statstcal mechancs n order to calculate the prospertes of

More information

ANSWERS. Problem 1. and the moment generating function (mgf) by. defined for any real t. Use this to show that E( U) var( U)

ANSWERS. Problem 1. and the moment generating function (mgf) by. defined for any real t. Use this to show that E( U) var( U) Econ 413 Exam 13 H ANSWERS Settet er nndelt 9 deloppgaver, A,B,C, som alle anbefales å telle lkt for å gøre det ltt lettere å stå. Svar er gtt . Unfortunately, there s a prntng error n the hnt of

More information

BOUNDEDNESS OF THE RIESZ TRANSFORM WITH MATRIX A 2 WEIGHTS

BOUNDEDNESS OF THE RIESZ TRANSFORM WITH MATRIX A 2 WEIGHTS BOUNDEDNESS OF THE IESZ TANSFOM WITH MATIX A WEIGHTS Introducton Let L = L ( n, be the functon space wth norm (ˆ f L = f(x C dx d < For a d d matrx valued functon W : wth W (x postve sem-defnte for all

More information

Complete subgraphs in multipartite graphs

Complete subgraphs in multipartite graphs Complete subgraphs n multpartte graphs FLORIAN PFENDER Unverstät Rostock, Insttut für Mathematk D-18057 Rostock, Germany Floran.Pfender@un-rostock.de Abstract Turán s Theorem states that every graph G

More information

Linear, affine, and convex sets and hulls In the sequel, unless otherwise specified, X will denote a real vector space.

Linear, affine, and convex sets and hulls In the sequel, unless otherwise specified, X will denote a real vector space. Lnear, affne, and convex sets and hulls In the sequel, unless otherwse specfed, X wll denote a real vector space. Lnes and segments. Gven two ponts x, y X, we defne xy = {x + t(y x) : t R} = {(1 t)x +

More information

STAT 309: MATHEMATICAL COMPUTATIONS I FALL 2018 LECTURE 17. a ij x (k) b i. a ij x (k+1) (D + L)x (k+1) = b Ux (k)

STAT 309: MATHEMATICAL COMPUTATIONS I FALL 2018 LECTURE 17. a ij x (k) b i. a ij x (k+1) (D + L)x (k+1) = b Ux (k) STAT 309: MATHEMATICAL COMPUTATIONS I FALL 08 LECTURE 7. sor method remnder: n coordnatewse form, Jacob method s = [ b a x (k) a and Gauss Sedel method s = [ b a = = remnder: n matrx form, Jacob method

More information

Statistical Mechanics and Combinatorics : Lecture III

Statistical Mechanics and Combinatorics : Lecture III Statstcal Mechancs and Combnatorcs : Lecture III Dmer Model Dmer defntons Defnton A dmer coverng (perfect matchng) of a fnte graph s a set of edges whch covers every vertex exactly once, e every vertex

More information

Topic 5: Non-Linear Regression

Topic 5: Non-Linear Regression Topc 5: Non-Lnear Regresson The models we ve worked wth so far have been lnear n the parameters. They ve been of the form: y = Xβ + ε Many models based on economc theory are actually non-lnear n the parameters.

More information

Section 8.3 Polar Form of Complex Numbers

Section 8.3 Polar Form of Complex Numbers 80 Chapter 8 Secton 8 Polar Form of Complex Numbers From prevous classes, you may have encountered magnary numbers the square roots of negatve numbers and, more generally, complex numbers whch are the

More information

Notes on Frequency Estimation in Data Streams

Notes on Frequency Estimation in Data Streams Notes on Frequency Estmaton n Data Streams In (one of) the data streamng model(s), the data s a sequence of arrvals a 1, a 2,..., a m of the form a j = (, v) where s the dentty of the tem and belongs to

More information

NUMERICAL DIFFERENTIATION

NUMERICAL DIFFERENTIATION NUMERICAL DIFFERENTIATION 1 Introducton Dfferentaton s a method to compute the rate at whch a dependent output y changes wth respect to the change n the ndependent nput x. Ths rate of change s called the

More information

College of Computer & Information Science Fall 2009 Northeastern University 20 October 2009

College of Computer & Information Science Fall 2009 Northeastern University 20 October 2009 College of Computer & Informaton Scence Fall 2009 Northeastern Unversty 20 October 2009 CS7880: Algorthmc Power Tools Scrbe: Jan Wen and Laura Poplawsk Lecture Outlne: Prmal-dual schema Network Desgn:

More information

Structure and Drive Paul A. Jensen Copyright July 20, 2003

Structure and Drive Paul A. Jensen Copyright July 20, 2003 Structure and Drve Paul A. Jensen Copyrght July 20, 2003 A system s made up of several operatons wth flow passng between them. The structure of the system descrbes the flow paths from nputs to outputs.

More information

Global Sensitivity. Tuesday 20 th February, 2018

Global Sensitivity. Tuesday 20 th February, 2018 Global Senstvty Tuesday 2 th February, 28 ) Local Senstvty Most senstvty analyses [] are based on local estmates of senstvty, typcally by expandng the response n a Taylor seres about some specfc values

More information

Lecture 5.8 Flux Vector Splitting

Lecture 5.8 Flux Vector Splitting Lecture 5.8 Flux Vector Splttng 1 Flux Vector Splttng The vector E n (5.7.) can be rewrtten as E = AU (5.8.1) (wth A as gven n (5.7.4) or (5.7.6) ) whenever, the equaton of state s of the separable form

More information

Lecture 21: Numerical methods for pricing American type derivatives

Lecture 21: Numerical methods for pricing American type derivatives Lecture 21: Numercal methods for prcng Amercan type dervatves Xaoguang Wang STAT 598W Aprl 10th, 2014 (STAT 598W) Lecture 21 1 / 26 Outlne 1 Fnte Dfference Method Explct Method Penalty Method (STAT 598W)

More information

A CHARACTERIZATION OF ADDITIVE DERIVATIONS ON VON NEUMANN ALGEBRAS

A CHARACTERIZATION OF ADDITIVE DERIVATIONS ON VON NEUMANN ALGEBRAS Journal of Mathematcal Scences: Advances and Applcatons Volume 25, 2014, Pages 1-12 A CHARACTERIZATION OF ADDITIVE DERIVATIONS ON VON NEUMANN ALGEBRAS JIA JI, WEN ZHANG and XIAOFEI QI Department of Mathematcs

More information

THE WEIGHTED WEAK TYPE INEQUALITY FOR THE STRONG MAXIMAL FUNCTION

THE WEIGHTED WEAK TYPE INEQUALITY FOR THE STRONG MAXIMAL FUNCTION THE WEIGHTED WEAK TYPE INEQUALITY FO THE STONG MAXIMAL FUNCTION THEMIS MITSIS Abstract. We prove the natural Fefferman-Sten weak type nequalty for the strong maxmal functon n the plane, under the assumpton

More information

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity LINEAR REGRESSION ANALYSIS MODULE IX Lecture - 31 Multcollnearty Dr. Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur 6. Rdge regresson The OLSE s the best lnear unbased

More information

j) = 1 (note sigma notation) ii. Continuous random variable (e.g. Normal distribution) 1. density function: f ( x) 0 and f ( x) dx = 1

j) = 1 (note sigma notation) ii. Continuous random variable (e.g. Normal distribution) 1. density function: f ( x) 0 and f ( x) dx = 1 Random varables Measure of central tendences and varablty (means and varances) Jont densty functons and ndependence Measures of assocaton (covarance and correlaton) Interestng result Condtonal dstrbutons

More information

Hongyi Miao, College of Science, Nanjing Forestry University, Nanjing ,China. (Received 20 June 2013, accepted 11 March 2014) I)ϕ (k)

Hongyi Miao, College of Science, Nanjing Forestry University, Nanjing ,China. (Received 20 June 2013, accepted 11 March 2014) I)ϕ (k) ISSN 1749-3889 (prnt), 1749-3897 (onlne) Internatonal Journal of Nonlnear Scence Vol.17(2014) No.2,pp.188-192 Modfed Block Jacob-Davdson Method for Solvng Large Sparse Egenproblems Hongy Mao, College of

More information

Density matrix. c α (t)φ α (q)

Density matrix. c α (t)φ α (q) Densty matrx Note: ths s supplementary materal. I strongly recommend that you read t for your own nterest. I beleve t wll help wth understandng the quantum ensembles, but t s not necessary to know t n

More information