Scalable Symbolic Model Order Reduction

Size: px
Start display at page:

Download "Scalable Symbolic Model Order Reduction"

Transcription

1 Scalable Sybolic Model Order Reduction Yiyu Shi Lei He -J Richard Shi Electrical Engineering Dept, ULA Electrical Engineering Dept, UW Los Angeles, alifornia, 924 Seattle, WA, 985 {yshi, lhe}eeuclaedu {cjshi}uwashingtonedu ASTRAT Sybolic odel order reduction (SMOR) is to reduce the coplexity of a odel with sybolic paraeters It is an iportant proble in analog circuit synthesis and digital circuit odeling with process variations However, existing sybolic odel order reduction (SMOR) ethods do not scale well with the nuber of sybols or with the odel order This paper presents a scalable SMOR algorith, naely S 2 MOR We first separate the original ulti-port ulti-sybol syste into a set of single-port systes by superposition theore, and then integrate the together to for a lower-bordered block diagonal (LD) structured syste Each block is reduced independently, with a stochastic prograing to distribute the given overall odel order between blocks for best accuracy The entire syste is efficiently solved by lowrank update opared with existing SMOR algoriths, given the sae eory space, S 2 MOR iproves accuracy by up to 78% at a siilar reduction tie In addition, the factorization and siulation of the reduced odel by S 2 MOR is up to 7 faster INTRODUTION Sybolic circuit techniques are playing an increasingly iportant role with the advance of design technology, especially when we have entered the nano regie Plenty of algoriths exist in literature discussing how to analyze and siulate those sybolic circuits [ 3] However, all those ethods are practical only if the circuit has a oderate size Unfortunately, to guarantee reasonable odel accuracy, the circuits fro physical extraction usually contain illions of nodes, thus rendering those ethods inefficient Towards this end, nuerous odel order reduction (MOR) techniques have been successfully applied to the reduction of linear large scale circuits over the past decade ( [4, 5],etc) There were also efforts to extend those ethods to nonlinear circuits ( [6, 7], etc) and paraetrized circuits ( [8, 9], etc) However, despite their wide application, unsolved probles do exist when directly extending the to sybolic circuits The idea of sybolic odel order reduction (SMOR) was first introduced in [], which contains three different ethods: sybolic isolation, noinal projection and first order expansion The sybol isolation ethod first reoves all the sybols fro the circuits, and the nodes to which the sybols are connected are odeled as ports As such, the sybolic circuit becoes a sybol-free circuit with assive ports, and can be reduced by any traditional ethods [4,5, 3] However, the size of the reduced circuit is proportional to the port nuber ultiplying the nuber of oents to be atched Accordingly, the tie and space coplexity for the reduced odel increases cubically This work is supported in part by NSF AREER Award R-36682, Defense Advanced Projects Agency under contract W 3P 4θ 5 O R64 and in part by Defense Threat Reduction Agency under contract #F A This work was perfored while Yiyu Shi was an intern at Orora Design Technologies, Inc, Redond, WA with the nuber of ports and thus the nuber of sybols, which renders those ethods of less practical value for circuits with any sybols The noinal projection ethod uses the noinal values of the sybols to copute the projection atrix and is accurate only when the sybol values slightly deviate fro the noinal value The first order expansion ethod uses the first order expansion of the atrix inversion and ultiplication to find the projection atrix, which is first order atrix polynoial wrt all the sybols Again, no large change is allowed for the sybols in order for the ethod to be accurate In addition, all the above three sybolic MOR ethods suffer fro the following three probles: First, they don t scale well with the nuber of sybols or with the reduced order In literature only the experiental results for circuits with less than ten sybols are reported Second, if the original circuit has no special pattern to be explored, the reduced odel is dense and therefore is still tie consuing for siulation Third, when the values of the sybols change, it is difficult to directly update the factorization of the reduced odel instead of redo it fro scratch, especially for the noinal projection and first order expansion ethods A ore detailed analysis of those sybolic MOR ethods are provided in Section 2 The ain contribution of our paper is as follows: A scalable SMOR algorith, naely S 2 MOR is presented We first separate the original ulti-port ulti-sybol syste into a set of single-port systes by superposition theore, then integrate the together to for a lowerbordered block diagonal (LD) structure Each block is reduced independently, with a stochastic prograing to distribute the given overall odel order between blocks for best accuracy And the entire syste is efficiently solved by low-rank update Experients using analog designs show that S 2 MOR handles circuits with up to nodes and 66 sybols in 5 inutes opared with existing SMOR algoriths, S 2 MOR iproves accuracy by up to 78% at a siilar reduction tie and a sae eory space In addition, the factorization and siulation of the reduced odel by S 2 MOR is up to 7 faster The reaining of the paper is organized as follows: Section 2 reviews the existing approach for handling sybolic circuits and its disadvantages Section 3 presents our LD transforation and syste projection algorith We also discuss how to fully utilize the special sparsity pattern of the reduced odel for the efficient siulation and update of the reduced odel Experiental results are presented in Section 4 and concluding rearks are given in Section 5 2 PRELIMINARIES In this section, we briefly review the three SMOR ethods proposed in [], and address their disadvantages by coplexity analysis wrt the reduced order q For later coparison with our ethod, the coplexity for the three different ethods are suarized in the first three rows of Table 2 Sybol Isolation The ain idea of sybol isolation is to reove all the sybols fro

2 the circuit, and odel the nodes to which the sybols are connected as ports Then any traditional MOR ethods can be applied Generally speaking, a circuit with a total nuber of a sybols and p ports is converted to an equivalent sybol-free circuit with p + a ports, assuing all the sybols are two-terinal A general reduction ethod such as PRIMA [4] states that in order to atch the first q oents, the reduced odel will have a size of q() with a cost of O(() 2 q 2 ) for the reduction (the orthonoralization of q(p + a) vectors plus a constant atrix factorization cost) With no special structure, the tie coplexity to factorize the reduced odel is O(q 3 (p + a) 3 ) which increases cubically with a, and the space coplexity to store the reduced odel is O(q 2 (p + a) 2 ) Moreover, to update the reduced odel with a new set of sybol values, the cost is a() 2 q 2 by using the ethod of low rank update Therefore, such a ethod is useful when the circuits contain only a few sybols 22 Noinal Projection The ain idea of noinal projection ethod is to use the noinal values of the sybols to copute the projection atrix The odified nodal analysis (MNA) equation sybolic circuit can be written as (G + G) + s( + ) = u, () where G and are the noinal atrices, and G, contain the staping of the sybols Then the noinal projection ethod coputes the projection atrix V such that V κ q(a, R ), (2) where A = G and R + G, and κ q(a, R ) is the q th order Krylov subspace spanned by A and R This ethod is very efficient as the projection atrix does not depend on the nuber of sybols In order to approxiately atch q oents, the reduced circuit will have a size of pq with the cost of p 2 q 2 for the reduction (the orthonoralization of pq vectors) Accordingly, the tie coplexity for analyzing the reduced odel is O(q 3 p 3 ), and the space coplexity for the reduced odel including the coefficient atrices for all the sybols is O(ap 2 q 2 ) Moreover, to update the reduced odel with a new set of sybol values costs O(p 3 q 3 ) because it requires new factorization of the reduced odel As the values of the sybols deviate fro the noinal values, the projection atrix no longer contains the Krylov subspace spanned by the circuit atrix Accordingly, the odel accuracy decreases [] showed that the error of the reduced odel increases draatically even if the value of the sybols deviates only a few percent fro the noinal value 23 First Order Approxiation Siilar proble exists for the first order approxiation ethod The ethod coputes the projection atrix using first order expansion, ie, when coputing the projection atrix, the following approxiation is used [(G + G) ( + )] k (G + G) k A k R {A k G G A i G k i GA + k A j G A k j }R (3) j= After first order expansion, the Krylov subspace is directly used without orthonoralization to avoid the expensive cost of sybolic coputation which ay render the reduced odel singular Such a ethod can only work when the values of the sybols do not deviate uch fro the noinal values Siilar to the noinal projection ethod, the reduced circuit will have a size of pq if we approxiately atch q oents Note that the reduction cost is O(pq) because no orthonoralization is eployed The reduce odel has the sae space coplexity and tie coplexity as the noinal projection ethod 3 S 2 MOR ALGORITHM 3 Port Separation and Model Reduction The key idea of the S 2 MOR algorith is to separate the original ultiport ulti-sybol syste into a set of single-port systes by superposition theore, and then integrate the together to for a lower-bordered block diagonal (LD) structure We further show that by using the projection atrices fro each of the single-port systes, we can reduced such LD syste with the structure preserved The whole procedure is detailed below We start with the following odified nodal analysis (MNA) equation, which represents an RL circuit with a sybols and p ports (G + s)x + a P is i (P T i x) = u (4) y = L T x, (5) where G and ( R N N ) are the inductance and capacitance atrices, and L ( R N p ) are the incidence atrices for input and output, u ( R p ) is the input current vector The vector P i ( R N ) is the incidence vector for sybol i and it takes the for P i = ` T with at the j th eleent (positive node) at at the k th eleent (negative node) s i is the operator corresponding to the i v relation of the i th sybol Typically, for an resistive sybol, s i = /R is a constant operator For capacitive sybol, we have s i = s and for inductance, s i = /sl Accordingly, the current fro sybol i can be coputed as (6) w i = s i (P T i x), i a (7) Denote i as the i th colun of and u i as the i th eleent of u Then according to superposition theore, Eq (4) becoes p a (G + s)x = iu i P iw i, (8) Eq (8) can be further divided into a set of p + a equations, with each equation in the for of j (G + s)x (i) = iu i, i p P i pw i p, p + i p + a, (9) x = x (i), () i= where x (i) is the state variable for the i th equation, which corresponds to a syste with single port, and w i is the current through sybol i On the other hand, fro Eq () and Eq (7), we have w i = s i ( Pi T x (j) ) = j= j= s i (Pi T x (j) ), () For siplicity of presentation, we assue that the sybol is a two-terinal device However, it is understood that the algorith is readily to be applied to ulti-terinal devices as well

3 Table : Space and tie coplexity coparison between four different ethods wrt the reduced order q Method Mo Matched Reduced Size Space oplexity Tie oplexity Reduction Factorization Update Sybol Isolation q (p + a)q (p + a) 2 q 2 (p + a) 2 q 2 (p + a) 3 q 3 a(p + a) 2 q 2 Noinal Projection Approx q pq ap 2 q 2 p 2 q 2 p 3 q 3 p 3 q 3 st Order Expansion Approx, q pq ap 2 q 2 apq p 3 q 3 p 3 q 3 S 2 MOR q (p + a)q (p + a)q 2 (p + a)q 2 (p + a)q 3 const where we have used the fact that s i is a scalar for RL eleents We introduce and Eq (7) becoes z = ` x ()T x (2)T ()T T x w i = ` s i P T i s i P T i s i P T i (2) z, i a (3) Therefore, Eq (9)-Eq () can be cast into a copact atrix for as (Ĝ + sĉ)z = ˆu, (4) where Ĝ is a lower bordered block diagonal (LD) atrix and Ĉ is a block diagonal atrix as shown in Eq (5) at the top of next page In addition, we have ˆ = p (6) A With the augented syste, the output can be coputed as where y = ˆLz, (7) ˆL = ` L L L T (8) One of the advantages of the augented syste is that a sybol-free projection atrix can be found, which can preserve the special sparse structure of the Ĝ and Ĉ atrices as well This is detailed in the following theore Theore If orthonoralized atrices V i satisfies j κ V i q{g,, i} i p κ q{g,, P i p} p + i p + a, (9) where κ q{g,, i} is the q th order Krylov subspace spanned by G, and i and κ q{g,, P i p} is the q th order Krylov subspace spanned by G, and P i p Then with the projection atrix V V 2 V = A, (2) V the first q oents of the reduced syste (Ĝr + sĉr)zr = ˆ ru (2) y r = ˆL rz r (22) and the original syste are atched, where Ĝr = VT ĜV, Ĉ r = V T ĈV, ˆr = V T r and ˆL r = V T ˆL Due to the space liit, the proof is oitted here Although the projection atrix in Eq (9) can guarantee the atching of the first q oents, it does not always provide the best accuracy depending on the sybol values Later we will show how to find an optial projection atrix in ters of accuracy, under the given reduced size, by changing the order of the Krylov subspaces spanning each V i in Eq (2) To atch q oents, the reduced atrices Ĝr and Ĉr would have a diension of (p + a)q, which is the sae as the that fro PRIMA However, different fro the dense reduced atrices fro PRIMA, it is easy to see that the Ĝr of the reduced syste still keeps the LD structure, and the Ĉr still keeps the block diagonal structure as shown in Eq (5) Note that for efficiency, the atrix G r should never be fored explicitly Instead, it can be cast as a block diagonal atrix plus with a rank q updates, ie, Ĝ r = D + LH T (23) where D ( R ()q ()q ) is a block diagonal atrix G r, D = A, (24) and L, H ( R ()q a ) can be written as G r, P p+ r, L = P p+2, (25) A Pr,a s P r, T s 2 P T sa P T r,a s P r, 2T s 2 P 2T sa P 2T r,a H =, (26) s P ()T s r, 2 P ()T sa P r,a ()T A where only D and L, H are stored and explicitly fored This sparsity structure facilitates the efficient siulation as well as update of the reduced odel, as will be discuss shortly The tie coplexity of the proposed algorith can be decoposed into two portions: the tie to copute p + a Krylov subspaces, each with order q; Note that the factorization of of G can be shared between different subspaces, and we need to do p + a ties of vector orthonoralization, each with q vectors Accordingly, the tie coplexity is O((p + a)q 2 ) Assuing no special structure for G and atrices in (4), the space needed to store the reduced atrices Ĉ r is O(pq 2 ) since it is block diagonal To store Ĝr, we need to store D, L and H, which cost O((p + a)q 3 ), O(aq) and O(a(p + a)q) respectively Keeping the doinant ters in q, and we can see that the total space coplexity is O((p + a)q 2 ) 32 Siulation and Update of the Reduced Model To siulate the reduced odel, either in tie doain for transient analysis or in frequency doain for A analysis, we would face the proble of efficiently factorizing a LD atrix We will concentrate on the QR factorization as it is shown to be the ost stable factorization algorith [4] We will show that instead of the general factorization cost O(() 3 q 3 ) for non-structuralized atrices with diension ()q, our special LD atrix can be factorized at a uch lower cost For

4 Ĝ = G G P s P T G P s P T P s P T P as a P T a G P as a P T a A Ĉ = A (5) the siplicity of presentation, we will siply discuss the factorization of Ĝ r + sĉr atrix for frequency doain analysis Siilar algorith can be applied for the tie doain analysis as the atrix structure will reain the sae Suppose we want to solve (Ĝr + sĉr)x = ˆ ru (27) for any input u and frequency s To start with, fro the atrix inversion lea we know that (Ĝr + sĉr) = (D + s Ĉ r + LH T ) = (D + s Ĉ r) (D + s Ĉ r) L I + H T (D + s Ĉ r) L H T (D + s Ĉ r) (28) Accordingly, an efficient factorization ethod can be obtained by first factorizing E = D+s Ĉ r, which is block diagonal, and then factorizing M = I + H T (D + s Ĉ r)l (M R a a ) Then for any input u, we can first solve and then solve Finally, solve and the solution can be obtained as Ex = ru, (29) Mx = x (3) Ex = Lx (3) x = x x (32) The ain cost for such an algorith lies in the factorization of E and M, which is O((p + a)q 3 ) and O(a 3 ) Keeping the doinant ter in q, we get that the overall tie coplexity for factorization is O((p + a)q 3 ) Moreover, each tie the values of the sybols are changed, we only need to re-factorize M, the cost of which is O(a 3 ) and does not depend on q The coparison of coplexity between the sybol isolation ethod, the noinal projection ethod, the first order expansion ethod as well as the S 2 MOR ethod is presented in Table The reduced circuits fro the four ethods all atch or approxiately atch the first q oents of the original circuit The best of the four ethod is shown in bold Fro the table we can see that the S 2 MOR ethod outperfors the other three ethods in alost all the coplexity coparisons For the sae nuber of oents to be atched, the S 2 MOR algorith has the lowest space and tie coplexity 33 Min-ax Prograing based Projection Order Decision Theore provides a ethod of coputing the projection atrix which can atch the first q oents of the original circuit with a reduced size of d = (p + a)q It would also be interesting to note the following proble: If we are given the overall reduced size d, how to distribute it between each V i in (2), such that the reduced odel has the best accuracy We will show that unifor distribution between the blocks as in Theore does not necessarily provide the best accuracy Specifically, the freedo of choosing the projection atrix lies in the order of the Krylov subspace for each V i in (9) If we choose a Krylov subspace with order q i, then the error introduced due to the isatching of the (q i + ) th oent can be coputed fro Eq (4) If i p, then i = = L T (G ) q i+ G i L T A q i + R i, (33) where A = G, R i = G i and L is the th colun of L If p + i p + a, then i = L T ((G + P i ps i p Pi p) T ) q i+ (G + P i ps i p P T i p) P i p (34) We can expand (34) to the first order by letting G = P is i Pi T k = q i + in (35), ie, i = L T ((G + P i ps i p P T i p) ) q i+ (G +P i ps i p Pi p) T P i p L T q i + q `Ai Ri p (A i + i F (s i ) and q i A j q i F (s i )A i + j i ) R i p, (35)! j= where F (s i ) = G P i ps i p Pi p T and R i p = G P i p Accordingly, the total error is f = p L T q i A + R + i i=p+ L T q i + `Ai Ri p q i q (A i + i F (s i ) A j q i F (s i )A i + j i ) R i p, (36)! j= which is a function of q i ( i p + a) and s i ( i a) Since we know the reduced size is d, we have the constraint q i = d (37) and they ust be non-negative integer q i Z + {}, i p + a (38)

5 Moreover, assue we have the statistical description for the sybols, ie, s i ω i, (39) where ω i is soe statistical description for s i Then we need to solve the following statistical optiization proble in q,q f(q, q ; s, s a ) (4) st q i = d (4) q i Z + {}, i p + a, (42) s i ω i, i a (43) Note that the objective function in Eq (43) is statistical, and we try to iniize its worst case value for all possible values of the sybols, which gives the following constrained in-ax ixed integer optiization proble in q,q st ax f(q, q; s, sa ) (44) s,s a q i = d (45) q i Z + {}, i p + a, (46) s i ω i, i a (47) Generally the in-ax proble is very hard to solve, especially the above proble is ixed-integer based and is nonlinear elow we propose an efficient heuristic algorith to approxiate solve it, based on the key observation that the constraints for s i and q i can be separated (ie, there are no cross ters) Accordingly, we separate the in-ax probles into the following two sub-probles Siilar algorith has been used in [5] for decoupling capacitance budgeting proble (P :) in q,q f(q, q ; s, s a ) (48) st q i = d (49) q i Z + {}, i p + a, (5) which is a integer-iniization proble for fixed s,, s a, and (P2 :) ax s,,s a f(q,, q ; s, s a ) (5) st s i ω i, (52) which is a axiization proble for fixed q,, q Efficient algoriths exist to solve the above two sub-probles [6] and is beyond the scope of this paper We iteratively solve (P) and (P2) until the solution converges In certain cases, the solution ay fail to converge, and we force the algorith to exit after certain iterations to guarantee the convergence This heuristic algorith can provide optial solution only in the case that (P) is convex and (P2) is concave Otherwise, the optiality is not guaranteed However, experiental results show that significant accuracy iproveent is achieved by such heuristic 4 EPERIMENTAL RESULTS In this section, we present experiental results for runtie and accuracy coparison between the S 2 MOR ethod and the sybol isolation, the noinal projection and the first order expansion ethods All the ethods are ipleented in ++ We use the Sparse as the sparse atrix package [7] We run experients on a UNI workstation with Pentiu IV 266G PU and G RAM Finally, we use MOSEK as the linear/quadratic prograing solver [8] for optial projection order decision 4 Algorith Verification Fig illustrates the sparsity of the reduced atrices fro the S 2 MOR ethod The original circuit is a low-noise aplifier (LNA) design with parasitics, which contains 492 nodes, 8 ports sybols The circuit is reduced to order 76 by the S 2 NMOR ethod Fro the figure we can see that although there is no special structure for the original atrices, the reduced Ĝr and Ĉr fro the S2 MOR ethod has LD structure with sparsity 34% and block diagonal structure with sparsity 77% respecitvely nz = 966 (a) Gr nz = 486 (b) r Figure : Sparsity pattern for the atrices Ĝr and Ĉr after reduction fro S 2 MOR Next we verify the effectiveness of our optial projection order decision We perfor Monte arlo siulation on the sybol values for, runs on the sae LNA circuit, and copare the error between unifor projection order decision, assuing the sae projection order for all sybols, and our stochastic prograing based projection order decision The error is defined as the integral of the absolute difference between the original tie doain wavefor and the one of the reduced circuit The error distributions for both ethods are shown in Fig 2 Fro the figure we can see that by our ethod the ean error is reduced by approxiately 3% and the 3σ error by 5%, which verifies the effectiveness of the proposed stochastic prograing based approach ount Stochastic Prograing ased Unifor Error (V*ns) Figure 2: Accuracy oparison between unifor projection order and our stochastic prograing based approach based on the Monte arlo siulation 42 Accuracy oparison In this section, we copare the accuracy of the three existing ethods and the S 2 MOR ethod on soe real industrial designs For fair coparison, we reduce the circuit to different orders for different ethods such that the reduced odels require the sae eory space This

6 Table 2: Testbench inforation ckt nae node # port # sybol # LNA LNA LNA LNA Table 3: Accuracy coparison between sybol isolation (SI), noinal projection (NP), first order expansion (FE) and the S 2 MOR with different variation aount (var) of sybol values All the errors are in the unit of V*ns ckt nae var SI NP FE S 2 MOR LNA % (-24%) 3% (-5%) LNA2 % (-27%) 3% (-78%) LNA3 % (-76%) 3% (-76%) LNA4 % NA 6 (-69%) 3% NA 6 (-69%) iplies that the three existing ethods have the sae reduced order because they are all dense, while the S 2 MOR ethod can have a uch higher reduced order due to its sparsity for the sae eory space We opare the accuracy between the four different ethods on different bencharks, and the results are shown in Table 3 The inforation of the testbenches we used is shown in Table 2 They are all LNA s fro extraction Fro Table 3 we can see that for different circuits and for different variation aount, the S 2 MOR ethod always have the sallest error copared with the other three ethods Specifically, when the variation aount is %, copared with the best one of the other three ethods shown in bold, the S 2 MOR ethod reduces the error by up to 76%; When the variation aount is 3%, the S 2 MOR ethod reduces the error by up to 78% Also note that the first order expansion ethod cannot finish the testbench LNA4 due to its singularity This is a direct result fro the lack of orthonoralization procedure for the projection atrix 43 Runtie oparison In Table 4, we report the runtie for reduction, factorization, and update for the reduced odel on different testbenches All the circuits are reduced to a different size such that the reduced circuits fro different ethods requires the sae eory space Fro the table we can see that the reduction tie is alost the sae for all the ethods The factorization tie is reduced by up to 94 for the S 2 MOR ethod on the largest testbench, although the reduced circuit size of the S 2 MOR ethod is alost 3 larger This speedup coes fro the efficient factorization algorith based on the LD structure of the reduced circuit Finally, the factorization update tie for sybol value changes is reduced by 7 for our ethod copared with the sybol isolation ethod, which uses low rank update This is because we only need to refactorize a atrix with size equal to the nuber of sybols All the above results further verify the correctness of our conclusion in Table 5 ONLUSIONS Sybolic odel order reduction (SMOR) is to reduce the coplexity of a odel with sybolic paraeters It is an iportant proble in analog circuit synthesis and digital circuit odeling with process variations However, existing sybolic odel order reduction (SMOR) ethods do not scale well with the nuber of sybols or with the odel order The literature only reports experiental results for circuits with fewer than ten sybols This paper presents a scalable SMOR algorith, naely S 2 MOR, which handles circuits with up to nodes and 66 sybols in 5 inutes fro experiental results opared with Table 4: Runtie coparison between sybol isolation (SI), noinal projection (NP), first order expansion (FE) and the S 2 MOR ethod The reduced sizes are also reported (size) All units are in seconds ckt nae ethod size reduce factor update LNA SI NP FE S 2 MOR LNA2 SI NP FE S 2 MOR LNA3 SI NP FE S 2 MOR LNA4 SI NP FE S 2 MOR existing SMOR algoriths, given the sae eory space, S 2 MOR iproves accuracy by up to 78% at a siilar reduction tie In addition, the factorization and siulation of the reduced odel by S 2 MOR is up to 7 faster 6 REFERENES [] T Halfann, R Soer, and T Sichann, Application of sybolic circuit analysis: An overview and recent resutls in nonlinear odel generation, in International onference on Electronics, ircuits and Systes, pp , 999 [2] -D Tan and -J Shi, Sybolic circuit-noise analysis and odeling with deterinant decision diagras, in ASPDA, pp , 2 [3] G G Gielen and W M Sansen, Sybolic analysis for autoated design of analog integrated circuits: a tutorial overview Kluwer Acadeic Publisher, 99 [4] A Odabasioglu, M elik, and L Pileggi, PRIMA: Passive reduced-order interconnect acro-odeling algorith, IEEE Trans on AD, pp , 998 [5] Y Su and et al, SAPOR: Second-order arnoldi ethod for passive order reduction of RS circuits, in IEEE/AM IAD, pp 74 79, 24 [6] M Rewienski and J White, A trajectory piecewise-linear approach to odel order reduction and fast siulation of nonlinear circuits and icroachined devices, IEEE Trans on AD, pp 55 7, 23 [7] N ond and L Daniel, A piecewise-linear oent-atching approach to paraeterized odel-order reduction for highly nonlinear systes, IEEE Trans on AD, pp , 27 [8] Y Li, Z ai, Y Su, and Zeng, Paraeterized odel order reduction via a two-directional arnoldi process, in IEEE/AM IAD, pp , 27 [9] Y Shi and L He, Epire: An efficient and copact ultiple-paraeterized odel order reduction ethod for physical optiizaion, in ISPD, pp 5 58, 27 [] G Shi, Hu, and J R Shi, On sybolic odel order reduction, IEEE Trans on AD, pp , 26 [] Y Shi, H Yu, and L He, Sason: A generalized second-order arnoldi ethod for reducing ultiple source linear network with susceptance, in ISPD, pp 25 32, 26 [2] P Li and W-P Shi, Model order reduction of linear networks with assive ports via frequency-dependent port packing, in IEEEAM/ DA, pp , 26 [3] Yan, -D Tan, P Liu, and McGaughy, Sbpor:second-order balanced truncation for passive order reduction of rlc circuits, in IEEEAM/ DA, pp 58 6, 27 [4] G H Golub and F V Loan, Matrix oputations altiore, MD: The Johns Hopkins University Press, 3 ed, 989 [5] Y Shi, J iong, Liu, and L He, Efficient decoupling capacitance budgeting considering operation and process variations, IEEE Trans on AD, pp , 28 [6] S oyd and L Vandenberghe, onvex Optiization abridge University Press, 24 [7] T Davis, Direct Methods for Sparse Linear Systes SIAM, 26 [8] in

EE5900 Spring Lecture 4 IC interconnect modeling methods Zhuo Feng

EE5900 Spring Lecture 4 IC interconnect modeling methods Zhuo Feng EE59 Spring Parallel LSI AD Algoriths Lecture I interconnect odeling ethods Zhuo Feng. Z. Feng MTU EE59 So far we ve considered only tie doain analyses We ll soon see that it is soeties preferable to odel

More information

Sharp Time Data Tradeoffs for Linear Inverse Problems

Sharp Time Data Tradeoffs for Linear Inverse Problems Sharp Tie Data Tradeoffs for Linear Inverse Probles Saet Oyak Benjain Recht Mahdi Soltanolkotabi January 016 Abstract In this paper we characterize sharp tie-data tradeoffs for optiization probles used

More information

A Simplified Analytical Approach for Efficiency Evaluation of the Weaving Machines with Automatic Filling Repair

A Simplified Analytical Approach for Efficiency Evaluation of the Weaving Machines with Automatic Filling Repair Proceedings of the 6th SEAS International Conference on Siulation, Modelling and Optiization, Lisbon, Portugal, Septeber -4, 006 0 A Siplified Analytical Approach for Efficiency Evaluation of the eaving

More information

Ch 12: Variations on Backpropagation

Ch 12: Variations on Backpropagation Ch 2: Variations on Backpropagation The basic backpropagation algorith is too slow for ost practical applications. It ay take days or weeks of coputer tie. We deonstrate why the backpropagation algorith

More information

List Scheduling and LPT Oliver Braun (09/05/2017)

List Scheduling and LPT Oliver Braun (09/05/2017) List Scheduling and LPT Oliver Braun (09/05/207) We investigate the classical scheduling proble P ax where a set of n independent jobs has to be processed on 2 parallel and identical processors (achines)

More information

EMPIRICAL COMPLEXITY ANALYSIS OF A MILP-APPROACH FOR OPTIMIZATION OF HYBRID SYSTEMS

EMPIRICAL COMPLEXITY ANALYSIS OF A MILP-APPROACH FOR OPTIMIZATION OF HYBRID SYSTEMS EMPIRICAL COMPLEXITY ANALYSIS OF A MILP-APPROACH FOR OPTIMIZATION OF HYBRID SYSTEMS Jochen Till, Sebastian Engell, Sebastian Panek, and Olaf Stursberg Process Control Lab (CT-AST), University of Dortund,

More information

The Simplex Method is Strongly Polynomial for the Markov Decision Problem with a Fixed Discount Rate

The Simplex Method is Strongly Polynomial for the Markov Decision Problem with a Fixed Discount Rate The Siplex Method is Strongly Polynoial for the Markov Decision Proble with a Fixed Discount Rate Yinyu Ye April 20, 2010 Abstract In this note we prove that the classic siplex ethod with the ost-negativereduced-cost

More information

A Generalized Permanent Estimator and its Application in Computing Multi- Homogeneous Bézout Number

A Generalized Permanent Estimator and its Application in Computing Multi- Homogeneous Bézout Number Research Journal of Applied Sciences, Engineering and Technology 4(23): 5206-52, 202 ISSN: 2040-7467 Maxwell Scientific Organization, 202 Subitted: April 25, 202 Accepted: May 3, 202 Published: Deceber

More information

Lecture 21. Interior Point Methods Setup and Algorithm

Lecture 21. Interior Point Methods Setup and Algorithm Lecture 21 Interior Point Methods In 1984, Kararkar introduced a new weakly polynoial tie algorith for solving LPs [Kar84a], [Kar84b]. His algorith was theoretically faster than the ellipsoid ethod and

More information

Homework 3 Solutions CSE 101 Summer 2017

Homework 3 Solutions CSE 101 Summer 2017 Hoework 3 Solutions CSE 0 Suer 207. Scheduling algoriths The following n = 2 jobs with given processing ties have to be scheduled on = 3 parallel and identical processors with the objective of iniizing

More information

Feature Extraction Techniques

Feature Extraction Techniques Feature Extraction Techniques Unsupervised Learning II Feature Extraction Unsupervised ethods can also be used to find features which can be useful for categorization. There are unsupervised ethods that

More information

Interactive Markov Models of Evolutionary Algorithms

Interactive Markov Models of Evolutionary Algorithms Cleveland State University EngagedScholarship@CSU Electrical Engineering & Coputer Science Faculty Publications Electrical Engineering & Coputer Science Departent 2015 Interactive Markov Models of Evolutionary

More information

A note on the multiplication of sparse matrices

A note on the multiplication of sparse matrices Cent. Eur. J. Cop. Sci. 41) 2014 1-11 DOI: 10.2478/s13537-014-0201-x Central European Journal of Coputer Science A note on the ultiplication of sparse atrices Research Article Keivan Borna 12, Sohrab Aboozarkhani

More information

RESTARTED FULL ORTHOGONALIZATION METHOD FOR SHIFTED LINEAR SYSTEMS

RESTARTED FULL ORTHOGONALIZATION METHOD FOR SHIFTED LINEAR SYSTEMS BIT Nuerical Matheatics 43: 459 466, 2003. 2003 Kluwer Acadeic Publishers. Printed in The Netherlands 459 RESTARTED FULL ORTHOGONALIZATION METHOD FOR SHIFTED LINEAR SYSTEMS V. SIMONCINI Dipartiento di

More information

Hybrid System Identification: An SDP Approach

Hybrid System Identification: An SDP Approach 49th IEEE Conference on Decision and Control Deceber 15-17, 2010 Hilton Atlanta Hotel, Atlanta, GA, USA Hybrid Syste Identification: An SDP Approach C Feng, C M Lagoa, N Ozay and M Sznaier Abstract The

More information

This model assumes that the probability of a gap has size i is proportional to 1/i. i.e., i log m e. j=1. E[gap size] = i P r(i) = N f t.

This model assumes that the probability of a gap has size i is proportional to 1/i. i.e., i log m e. j=1. E[gap size] = i P r(i) = N f t. CS 493: Algoriths for Massive Data Sets Feb 2, 2002 Local Models, Bloo Filter Scribe: Qin Lv Local Models In global odels, every inverted file entry is copressed with the sae odel. This work wells when

More information

Using a De-Convolution Window for Operating Modal Analysis

Using a De-Convolution Window for Operating Modal Analysis Using a De-Convolution Window for Operating Modal Analysis Brian Schwarz Vibrant Technology, Inc. Scotts Valley, CA Mark Richardson Vibrant Technology, Inc. Scotts Valley, CA Abstract Operating Modal Analysis

More information

Chapter 6 1-D Continuous Groups

Chapter 6 1-D Continuous Groups Chapter 6 1-D Continuous Groups Continuous groups consist of group eleents labelled by one or ore continuous variables, say a 1, a 2,, a r, where each variable has a well- defined range. This chapter explores:

More information

Finding Rightmost Eigenvalues of Large Sparse. Non-symmetric Parameterized Eigenvalue Problems. Abstract. Introduction

Finding Rightmost Eigenvalues of Large Sparse. Non-symmetric Parameterized Eigenvalue Problems. Abstract. Introduction Finding Rightost Eigenvalues of Large Sparse Non-syetric Paraeterized Eigenvalue Probles Applied Matheatics and Scientific Coputation Progra Departent of Matheatics University of Maryland, College Par,

More information

ESTIMATING AND FORMING CONFIDENCE INTERVALS FOR EXTREMA OF RANDOM POLYNOMIALS. A Thesis. Presented to. The Faculty of the Department of Mathematics

ESTIMATING AND FORMING CONFIDENCE INTERVALS FOR EXTREMA OF RANDOM POLYNOMIALS. A Thesis. Presented to. The Faculty of the Department of Mathematics ESTIMATING AND FORMING CONFIDENCE INTERVALS FOR EXTREMA OF RANDOM POLYNOMIALS A Thesis Presented to The Faculty of the Departent of Matheatics San Jose State University In Partial Fulfillent of the Requireents

More information

Block designs and statistics

Block designs and statistics Bloc designs and statistics Notes for Math 447 May 3, 2011 The ain paraeters of a bloc design are nuber of varieties v, bloc size, nuber of blocs b. A design is built on a set of v eleents. Each eleent

More information

Randomized Recovery for Boolean Compressed Sensing

Randomized Recovery for Boolean Compressed Sensing Randoized Recovery for Boolean Copressed Sensing Mitra Fatei and Martin Vetterli Laboratory of Audiovisual Counication École Polytechnique Fédéral de Lausanne (EPFL) Eail: {itra.fatei, artin.vetterli}@epfl.ch

More information

Lecture 9 November 23, 2015

Lecture 9 November 23, 2015 CSC244: Discrepancy Theory in Coputer Science Fall 25 Aleksandar Nikolov Lecture 9 Noveber 23, 25 Scribe: Nick Spooner Properties of γ 2 Recall that γ 2 (A) is defined for A R n as follows: γ 2 (A) = in{r(u)

More information

Multi-Scale/Multi-Resolution: Wavelet Transform

Multi-Scale/Multi-Resolution: Wavelet Transform Multi-Scale/Multi-Resolution: Wavelet Transfor Proble with Fourier Fourier analysis -- breaks down a signal into constituent sinusoids of different frequencies. A serious drawback in transforing to the

More information

13.2 Fully Polynomial Randomized Approximation Scheme for Permanent of Random 0-1 Matrices

13.2 Fully Polynomial Randomized Approximation Scheme for Permanent of Random 0-1 Matrices CS71 Randoness & Coputation Spring 018 Instructor: Alistair Sinclair Lecture 13: February 7 Disclaier: These notes have not been subjected to the usual scrutiny accorded to foral publications. They ay

More information

Non-Parametric Non-Line-of-Sight Identification 1

Non-Parametric Non-Line-of-Sight Identification 1 Non-Paraetric Non-Line-of-Sight Identification Sinan Gezici, Hisashi Kobayashi and H. Vincent Poor Departent of Electrical Engineering School of Engineering and Applied Science Princeton University, Princeton,

More information

CS Lecture 13. More Maximum Likelihood

CS Lecture 13. More Maximum Likelihood CS 6347 Lecture 13 More Maxiu Likelihood Recap Last tie: Introduction to axiu likelihood estiation MLE for Bayesian networks Optial CPTs correspond to epirical counts Today: MLE for CRFs 2 Maxiu Likelihood

More information

Statistical Logic Cell Delay Analysis Using a Current-based Model

Statistical Logic Cell Delay Analysis Using a Current-based Model Statistical Logic Cell Delay Analysis Using a Current-based Model Hanif Fatei Shahin Nazarian Massoud Pedra Dept. of EE-Systes, University of Southern California, Los Angeles, CA 90089 {fatei, shahin,

More information

Design of Spatially Coupled LDPC Codes over GF(q) for Windowed Decoding

Design of Spatially Coupled LDPC Codes over GF(q) for Windowed Decoding IEEE TRANSACTIONS ON INFORMATION THEORY (SUBMITTED PAPER) 1 Design of Spatially Coupled LDPC Codes over GF(q) for Windowed Decoding Lai Wei, Student Meber, IEEE, David G. M. Mitchell, Meber, IEEE, Thoas

More information

Department of Electronic and Optical Engineering, Ordnance Engineering College, Shijiazhuang, , China

Department of Electronic and Optical Engineering, Ordnance Engineering College, Shijiazhuang, , China 6th International Conference on Machinery, Materials, Environent, Biotechnology and Coputer (MMEBC 06) Solving Multi-Sensor Multi-Target Assignent Proble Based on Copositive Cobat Efficiency and QPSO Algorith

More information

Experimental Design For Model Discrimination And Precise Parameter Estimation In WDS Analysis

Experimental Design For Model Discrimination And Precise Parameter Estimation In WDS Analysis City University of New York (CUNY) CUNY Acadeic Works International Conference on Hydroinforatics 8-1-2014 Experiental Design For Model Discriination And Precise Paraeter Estiation In WDS Analysis Giovanna

More information

About the definition of parameters and regimes of active two-port networks with variable loads on the basis of projective geometry

About the definition of parameters and regimes of active two-port networks with variable loads on the basis of projective geometry About the definition of paraeters and regies of active two-port networks with variable loads on the basis of projective geoetry PENN ALEXANDR nstitute of Electronic Engineering and Nanotechnologies "D

More information

An improved self-adaptive harmony search algorithm for joint replenishment problems

An improved self-adaptive harmony search algorithm for joint replenishment problems An iproved self-adaptive harony search algorith for joint replenishent probles Lin Wang School of Manageent, Huazhong University of Science & Technology zhoulearner@gail.co Xiaojian Zhou School of Manageent,

More information

Extension of CSRSM for the Parametric Study of the Face Stability of Pressurized Tunnels

Extension of CSRSM for the Parametric Study of the Face Stability of Pressurized Tunnels Extension of CSRSM for the Paraetric Study of the Face Stability of Pressurized Tunnels Guilhe Mollon 1, Daniel Dias 2, and Abdul-Haid Soubra 3, M.ASCE 1 LGCIE, INSA Lyon, Université de Lyon, Doaine scientifique

More information

On the Use of A Priori Information for Sparse Signal Approximations

On the Use of A Priori Information for Sparse Signal Approximations ITS TECHNICAL REPORT NO. 3/4 On the Use of A Priori Inforation for Sparse Signal Approxiations Oscar Divorra Escoda, Lorenzo Granai and Pierre Vandergheynst Signal Processing Institute ITS) Ecole Polytechnique

More information

Page 1 Lab 1 Elementary Matrix and Linear Algebra Spring 2011

Page 1 Lab 1 Elementary Matrix and Linear Algebra Spring 2011 Page Lab Eleentary Matri and Linear Algebra Spring 0 Nae Due /03/0 Score /5 Probles through 4 are each worth 4 points.. Go to the Linear Algebra oolkit site ransforing a atri to reduced row echelon for

More information

Deflation of the I-O Series Some Technical Aspects. Giorgio Rampa University of Genoa April 2007

Deflation of the I-O Series Some Technical Aspects. Giorgio Rampa University of Genoa April 2007 Deflation of the I-O Series 1959-2. Soe Technical Aspects Giorgio Rapa University of Genoa g.rapa@unige.it April 27 1. Introduction The nuber of sectors is 42 for the period 1965-2 and 38 for the initial

More information

Ştefan ŞTEFĂNESCU * is the minimum global value for the function h (x)

Ştefan ŞTEFĂNESCU * is the minimum global value for the function h (x) 7Applying Nelder Mead s Optiization Algorith APPLYING NELDER MEAD S OPTIMIZATION ALGORITHM FOR MULTIPLE GLOBAL MINIMA Abstract Ştefan ŞTEFĂNESCU * The iterative deterinistic optiization ethod could not

More information

Support Vector Machines. Goals for the lecture

Support Vector Machines. Goals for the lecture Support Vector Machines Mark Craven and David Page Coputer Sciences 760 Spring 2018 www.biostat.wisc.edu/~craven/cs760/ Soe of the slides in these lectures have been adapted/borrowed fro aterials developed

More information

Model Fitting. CURM Background Material, Fall 2014 Dr. Doreen De Leon

Model Fitting. CURM Background Material, Fall 2014 Dr. Doreen De Leon Model Fitting CURM Background Material, Fall 014 Dr. Doreen De Leon 1 Introduction Given a set of data points, we often want to fit a selected odel or type to the data (e.g., we suspect an exponential

More information

On Constant Power Water-filling

On Constant Power Water-filling On Constant Power Water-filling Wei Yu and John M. Cioffi Electrical Engineering Departent Stanford University, Stanford, CA94305, U.S.A. eails: {weiyu,cioffi}@stanford.edu Abstract This paper derives

More information

COS 424: Interacting with Data. Written Exercises

COS 424: Interacting with Data. Written Exercises COS 424: Interacting with Data Hoework #4 Spring 2007 Regression Due: Wednesday, April 18 Written Exercises See the course website for iportant inforation about collaboration and late policies, as well

More information

RECOVERY OF A DENSITY FROM THE EIGENVALUES OF A NONHOMOGENEOUS MEMBRANE

RECOVERY OF A DENSITY FROM THE EIGENVALUES OF A NONHOMOGENEOUS MEMBRANE Proceedings of ICIPE rd International Conference on Inverse Probles in Engineering: Theory and Practice June -8, 999, Port Ludlow, Washington, USA : RECOVERY OF A DENSITY FROM THE EIGENVALUES OF A NONHOMOGENEOUS

More information

Support Vector Machine Classification of Uncertain and Imbalanced data using Robust Optimization

Support Vector Machine Classification of Uncertain and Imbalanced data using Robust Optimization Recent Researches in Coputer Science Support Vector Machine Classification of Uncertain and Ibalanced data using Robust Optiization RAGHAV PAT, THEODORE B. TRAFALIS, KASH BARKER School of Industrial Engineering

More information

Optimal Jamming Over Additive Noise: Vector Source-Channel Case

Optimal Jamming Over Additive Noise: Vector Source-Channel Case Fifty-first Annual Allerton Conference Allerton House, UIUC, Illinois, USA October 2-3, 2013 Optial Jaing Over Additive Noise: Vector Source-Channel Case Erah Akyol and Kenneth Rose Abstract This paper

More information

Reducing Vibration and Providing Robustness with Multi-Input Shapers

Reducing Vibration and Providing Robustness with Multi-Input Shapers 29 Aerican Control Conference Hyatt Regency Riverfront, St. Louis, MO, USA June -2, 29 WeA6.4 Reducing Vibration and Providing Robustness with Multi-Input Shapers Joshua Vaughan and Willia Singhose Abstract

More information

A method to determine relative stroke detection efficiencies from multiplicity distributions

A method to determine relative stroke detection efficiencies from multiplicity distributions A ethod to deterine relative stroke detection eiciencies ro ultiplicity distributions Schulz W. and Cuins K. 2. Austrian Lightning Detection and Inoration Syste (ALDIS), Kahlenberger Str.2A, 90 Vienna,

More information

Gary J. Balas Aerospace Engineering and Mechanics, University of Minnesota, Minneapolis, MN USA

Gary J. Balas Aerospace Engineering and Mechanics, University of Minnesota, Minneapolis, MN USA μ-synthesis Gary J. Balas Aerospace Engineering and Mechanics, University of Minnesota, Minneapolis, MN 55455 USA Keywords: Robust control, ultivariable control, linear fractional transforation (LFT),

More information

Boosting with log-loss

Boosting with log-loss Boosting with log-loss Marco Cusuano-Towner Septeber 2, 202 The proble Suppose we have data exaples {x i, y i ) i =... } for a two-class proble with y i {, }. Let F x) be the predictor function with the

More information

Fast Montgomery-like Square Root Computation over GF(2 m ) for All Trinomials

Fast Montgomery-like Square Root Computation over GF(2 m ) for All Trinomials Fast Montgoery-like Square Root Coputation over GF( ) for All Trinoials Yin Li a, Yu Zhang a, a Departent of Coputer Science and Technology, Xinyang Noral University, Henan, P.R.China Abstract This letter

More information

Analyzing Simulation Results

Analyzing Simulation Results Analyzing Siulation Results Dr. John Mellor-Cruey Departent of Coputer Science Rice University johnc@cs.rice.edu COMP 528 Lecture 20 31 March 2005 Topics for Today Model verification Model validation Transient

More information

A Simple Regression Problem

A Simple Regression Problem A Siple Regression Proble R. M. Castro March 23, 2 In this brief note a siple regression proble will be introduced, illustrating clearly the bias-variance tradeoff. Let Y i f(x i ) + W i, i,..., n, where

More information

Course Notes for EE227C (Spring 2018): Convex Optimization and Approximation

Course Notes for EE227C (Spring 2018): Convex Optimization and Approximation Course Notes for EE7C (Spring 018: Convex Optiization and Approxiation Instructor: Moritz Hardt Eail: hardt+ee7c@berkeley.edu Graduate Instructor: Max Sichowitz Eail: sichow+ee7c@berkeley.edu October 15,

More information

Recovering Data from Underdetermined Quadratic Measurements (CS 229a Project: Final Writeup)

Recovering Data from Underdetermined Quadratic Measurements (CS 229a Project: Final Writeup) Recovering Data fro Underdeterined Quadratic Measureents (CS 229a Project: Final Writeup) Mahdi Soltanolkotabi Deceber 16, 2011 1 Introduction Data that arises fro engineering applications often contains

More information

Ensemble Based on Data Envelopment Analysis

Ensemble Based on Data Envelopment Analysis Enseble Based on Data Envelopent Analysis So Young Sohn & Hong Choi Departent of Coputer Science & Industrial Systes Engineering, Yonsei University, Seoul, Korea Tel) 82-2-223-404, Fax) 82-2- 364-7807

More information

Ph 20.3 Numerical Solution of Ordinary Differential Equations

Ph 20.3 Numerical Solution of Ordinary Differential Equations Ph 20.3 Nuerical Solution of Ordinary Differential Equations Due: Week 5 -v20170314- This Assignent So far, your assignents have tried to failiarize you with the hardware and software in the Physics Coputing

More information

Stability Analysis of the Matrix-Free Linearly Implicit 2 Euler Method 3 UNCORRECTED PROOF

Stability Analysis of the Matrix-Free Linearly Implicit 2 Euler Method 3 UNCORRECTED PROOF 1 Stability Analysis of the Matrix-Free Linearly Iplicit 2 Euler Method 3 Adrian Sandu 1 andaikst-cyr 2 4 1 Coputational Science Laboratory, Departent of Coputer Science, Virginia 5 Polytechnic Institute,

More information

Convexity-Based Optimization for Power-Delay Tradeoff using Transistor Sizing

Convexity-Based Optimization for Power-Delay Tradeoff using Transistor Sizing Convexity-Based Optiization for Power-Delay Tradeoff using Transistor Sizing Mahesh Ketkar, and Sachin S. Sapatnekar Departent of Electrical and Coputer Engineering University of Minnesota, Minneapolis,

More information

Vulnerability of MRD-Code-Based Universal Secure Error-Correcting Network Codes under Time-Varying Jamming Links

Vulnerability of MRD-Code-Based Universal Secure Error-Correcting Network Codes under Time-Varying Jamming Links Vulnerability of MRD-Code-Based Universal Secure Error-Correcting Network Codes under Tie-Varying Jaing Links Jun Kurihara KDDI R&D Laboratories, Inc 2 5 Ohara, Fujiino, Saitaa, 356 8502 Japan Eail: kurihara@kddilabsjp

More information

Approximation in Stochastic Scheduling: The Power of LP-Based Priority Policies

Approximation in Stochastic Scheduling: The Power of LP-Based Priority Policies Approxiation in Stochastic Scheduling: The Power of -Based Priority Policies Rolf Möhring, Andreas Schulz, Marc Uetz Setting (A P p stoch, r E( w and (B P p stoch E( w We will assue that the processing

More information

Distributed Subgradient Methods for Multi-agent Optimization

Distributed Subgradient Methods for Multi-agent Optimization 1 Distributed Subgradient Methods for Multi-agent Optiization Angelia Nedić and Asuan Ozdaglar October 29, 2007 Abstract We study a distributed coputation odel for optiizing a su of convex objective functions

More information

A model reduction approach to numerical inversion for a parabolic partial differential equation

A model reduction approach to numerical inversion for a parabolic partial differential equation Inverse Probles Inverse Probles 30 (204) 250 (33pp) doi:0.088/0266-56/30/2/250 A odel reduction approach to nuerical inversion for a parabolic partial differential equation Liliana Borcea, Vladiir Drusin

More information

The Wilson Model of Cortical Neurons Richard B. Wells

The Wilson Model of Cortical Neurons Richard B. Wells The Wilson Model of Cortical Neurons Richard B. Wells I. Refineents on the odgkin-uxley Model The years since odgkin s and uxley s pioneering work have produced a nuber of derivative odgkin-uxley-like

More information

Introduction to Machine Learning. Recitation 11

Introduction to Machine Learning. Recitation 11 Introduction to Machine Learning Lecturer: Regev Schweiger Recitation Fall Seester Scribe: Regev Schweiger. Kernel Ridge Regression We now take on the task of kernel-izing ridge regression. Let x,...,

More information

e-companion ONLY AVAILABLE IN ELECTRONIC FORM

e-companion ONLY AVAILABLE IN ELECTRONIC FORM OPERATIONS RESEARCH doi 10.1287/opre.1070.0427ec pp. ec1 ec5 e-copanion ONLY AVAILABLE IN ELECTRONIC FORM infors 07 INFORMS Electronic Copanion A Learning Approach for Interactive Marketing to a Custoer

More information

The Weierstrass Approximation Theorem

The Weierstrass Approximation Theorem 36 The Weierstrass Approxiation Theore Recall that the fundaental idea underlying the construction of the real nubers is approxiation by the sipler rational nubers. Firstly, nubers are often deterined

More information

Testing equality of variances for multiple univariate normal populations

Testing equality of variances for multiple univariate normal populations University of Wollongong Research Online Centre for Statistical & Survey Methodology Working Paper Series Faculty of Engineering and Inforation Sciences 0 esting equality of variances for ultiple univariate

More information

Quantum algorithms (CO 781, Winter 2008) Prof. Andrew Childs, University of Waterloo LECTURE 15: Unstructured search and spatial search

Quantum algorithms (CO 781, Winter 2008) Prof. Andrew Childs, University of Waterloo LECTURE 15: Unstructured search and spatial search Quantu algoriths (CO 781, Winter 2008) Prof Andrew Childs, University of Waterloo LECTURE 15: Unstructured search and spatial search ow we begin to discuss applications of quantu walks to search algoriths

More information

Tight Bounds for Maximal Identifiability of Failure Nodes in Boolean Network Tomography

Tight Bounds for Maximal Identifiability of Failure Nodes in Boolean Network Tomography Tight Bounds for axial Identifiability of Failure Nodes in Boolean Network Toography Nicola Galesi Sapienza Università di Roa nicola.galesi@uniroa1.it Fariba Ranjbar Sapienza Università di Roa fariba.ranjbar@uniroa1.it

More information

Intelligent Systems: Reasoning and Recognition. Perceptrons and Support Vector Machines

Intelligent Systems: Reasoning and Recognition. Perceptrons and Support Vector Machines Intelligent Systes: Reasoning and Recognition Jaes L. Crowley osig 1 Winter Seester 2018 Lesson 6 27 February 2018 Outline Perceptrons and Support Vector achines Notation...2 Linear odels...3 Lines, Planes

More information

Estimation of ADC Nonlinearities from the Measurement in Input Voltage Intervals

Estimation of ADC Nonlinearities from the Measurement in Input Voltage Intervals Estiation of ADC Nonlinearities fro the Measureent in Input Voltage Intervals M. Godla, L. Michaeli, 3 J. Šaliga, 4 R. Palenčár,,3 Deptartent of Electronics and Multiedia Counications, FEI TU of Košice,

More information

Multi-Dimensional Hegselmann-Krause Dynamics

Multi-Dimensional Hegselmann-Krause Dynamics Multi-Diensional Hegselann-Krause Dynaics A. Nedić Industrial and Enterprise Systes Engineering Dept. University of Illinois Urbana, IL 680 angelia@illinois.edu B. Touri Coordinated Science Laboratory

More information

Convex Programming for Scheduling Unrelated Parallel Machines

Convex Programming for Scheduling Unrelated Parallel Machines Convex Prograing for Scheduling Unrelated Parallel Machines Yossi Azar Air Epstein Abstract We consider the classical proble of scheduling parallel unrelated achines. Each job is to be processed by exactly

More information

Randomized Accuracy-Aware Program Transformations For Efficient Approximate Computations

Randomized Accuracy-Aware Program Transformations For Efficient Approximate Computations Randoized Accuracy-Aware Progra Transforations For Efficient Approxiate Coputations Zeyuan Allen Zhu Sasa Misailovic Jonathan A. Kelner Martin Rinard MIT CSAIL zeyuan@csail.it.edu isailo@it.edu kelner@it.edu

More information

HIGH RESOLUTION NEAR-FIELD MULTIPLE TARGET DETECTION AND LOCALIZATION USING SUPPORT VECTOR MACHINES

HIGH RESOLUTION NEAR-FIELD MULTIPLE TARGET DETECTION AND LOCALIZATION USING SUPPORT VECTOR MACHINES ICONIC 2007 St. Louis, O, USA June 27-29, 2007 HIGH RESOLUTION NEAR-FIELD ULTIPLE TARGET DETECTION AND LOCALIZATION USING SUPPORT VECTOR ACHINES A. Randazzo,. A. Abou-Khousa 2,.Pastorino, and R. Zoughi

More information

Lower Bounds for Quantized Matrix Completion

Lower Bounds for Quantized Matrix Completion Lower Bounds for Quantized Matrix Copletion Mary Wootters and Yaniv Plan Departent of Matheatics University of Michigan Ann Arbor, MI Eail: wootters, yplan}@uich.edu Mark A. Davenport School of Elec. &

More information

When Short Runs Beat Long Runs

When Short Runs Beat Long Runs When Short Runs Beat Long Runs Sean Luke George Mason University http://www.cs.gu.edu/ sean/ Abstract What will yield the best results: doing one run n generations long or doing runs n/ generations long

More information

Optimal Resource Allocation in Multicast Device-to-Device Communications Underlaying LTE Networks

Optimal Resource Allocation in Multicast Device-to-Device Communications Underlaying LTE Networks 1 Optial Resource Allocation in Multicast Device-to-Device Counications Underlaying LTE Networks Hadi Meshgi 1, Dongei Zhao 1 and Rong Zheng 2 1 Departent of Electrical and Coputer Engineering, McMaster

More information

Using EM To Estimate A Probablity Density With A Mixture Of Gaussians

Using EM To Estimate A Probablity Density With A Mixture Of Gaussians Using EM To Estiate A Probablity Density With A Mixture Of Gaussians Aaron A. D Souza adsouza@usc.edu Introduction The proble we are trying to address in this note is siple. Given a set of data points

More information

Iterative Linear Solvers and Jacobian-free Newton-Krylov Methods

Iterative Linear Solvers and Jacobian-free Newton-Krylov Methods Eric de Sturler Iterative Linear Solvers and Jacobian-free Newton-Krylov Methods Eric de Sturler Departent of Matheatics, Virginia Tech www.ath.vt.edu/people/sturler/index.htl sturler@vt.edu Efficient

More information

Fast Structural Similarity Search of Noncoding RNAs Based on Matched Filtering of Stem Patterns

Fast Structural Similarity Search of Noncoding RNAs Based on Matched Filtering of Stem Patterns Fast Structural Siilarity Search of Noncoding RNs Based on Matched Filtering of Ste Patterns Byung-Jun Yoon Dept. of Electrical Engineering alifornia Institute of Technology Pasadena, 91125, S Eail: bjyoon@caltech.edu

More information

Correlated Bayesian Model Fusion: Efficient Performance Modeling of Large-Scale Tunable Analog/RF Integrated Circuits

Correlated Bayesian Model Fusion: Efficient Performance Modeling of Large-Scale Tunable Analog/RF Integrated Circuits Correlated Bayesian odel Fusion: Efficient Perforance odeling of Large-Scale unable Analog/RF Integrated Circuits Fa Wang and Xin Li ECE Departent, Carnegie ellon University, Pittsburgh, PA 53 {fwang,

More information

Use of PSO in Parameter Estimation of Robot Dynamics; Part One: No Need for Parameterization

Use of PSO in Parameter Estimation of Robot Dynamics; Part One: No Need for Parameterization Use of PSO in Paraeter Estiation of Robot Dynaics; Part One: No Need for Paraeterization Hossein Jahandideh, Mehrzad Navar Abstract Offline procedures for estiating paraeters of robot dynaics are practically

More information

Stochastic Subgradient Methods

Stochastic Subgradient Methods Stochastic Subgradient Methods Lingjie Weng Yutian Chen Bren School of Inforation and Coputer Science University of California, Irvine {wengl, yutianc}@ics.uci.edu Abstract Stochastic subgradient ethods

More information

Kernel Methods and Support Vector Machines

Kernel Methods and Support Vector Machines Intelligent Systes: Reasoning and Recognition Jaes L. Crowley ENSIAG 2 / osig 1 Second Seester 2012/2013 Lesson 20 2 ay 2013 Kernel ethods and Support Vector achines Contents Kernel Functions...2 Quadratic

More information

The linear sampling method and the MUSIC algorithm

The linear sampling method and the MUSIC algorithm INSTITUTE OF PHYSICS PUBLISHING INVERSE PROBLEMS Inverse Probles 17 (2001) 591 595 www.iop.org/journals/ip PII: S0266-5611(01)16989-3 The linear sapling ethod and the MUSIC algorith Margaret Cheney Departent

More information

Algorithms for parallel processor scheduling with distinct due windows and unit-time jobs

Algorithms for parallel processor scheduling with distinct due windows and unit-time jobs BULLETIN OF THE POLISH ACADEMY OF SCIENCES TECHNICAL SCIENCES Vol. 57, No. 3, 2009 Algoriths for parallel processor scheduling with distinct due windows and unit-tie obs A. JANIAK 1, W.A. JANIAK 2, and

More information

time time δ jobs jobs

time time δ jobs jobs Approxiating Total Flow Tie on Parallel Machines Stefano Leonardi Danny Raz y Abstract We consider the proble of optiizing the total ow tie of a strea of jobs that are released over tie in a ultiprocessor

More information

The proofs of Theorem 1-3 are along the lines of Wied and Galeano (2013).

The proofs of Theorem 1-3 are along the lines of Wied and Galeano (2013). A Appendix: Proofs The proofs of Theore 1-3 are along the lines of Wied and Galeano (2013) Proof of Theore 1 Let D[d 1, d 2 ] be the space of càdlàg functions on the interval [d 1, d 2 ] equipped with

More information

. The univariate situation. It is well-known for a long tie that denoinators of Pade approxiants can be considered as orthogonal polynoials with respe

. The univariate situation. It is well-known for a long tie that denoinators of Pade approxiants can be considered as orthogonal polynoials with respe PROPERTIES OF MULTIVARIATE HOMOGENEOUS ORTHOGONAL POLYNOMIALS Brahi Benouahane y Annie Cuyt? Keywords Abstract It is well-known that the denoinators of Pade approxiants can be considered as orthogonal

More information

CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION Vol. IX Uncertainty Models For Robustness Analysis - A. Garulli, A. Tesi and A. Vicino

CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION Vol. IX Uncertainty Models For Robustness Analysis - A. Garulli, A. Tesi and A. Vicino UNCERTAINTY MODELS FOR ROBUSTNESS ANALYSIS A. Garulli Dipartiento di Ingegneria dell Inforazione, Università di Siena, Italy A. Tesi Dipartiento di Sistei e Inforatica, Università di Firenze, Italy A.

More information

Determining the Robot-to-Robot Relative Pose Using Range-only Measurements

Determining the Robot-to-Robot Relative Pose Using Range-only Measurements Deterining the Robot-to-Robot Relative Pose Using Range-only Measureents Xun S Zhou and Stergios I Roueliotis Abstract In this paper we address the proble of deterining the relative pose of pairs robots

More information

Defect-Aware SOC Test Scheduling

Defect-Aware SOC Test Scheduling Defect-Aware SOC Test Scheduling Erik Larsson +, Julien Pouget*, and Zebo Peng + Ebedded Systes Laboratory + LIRMM* Departent of Coputer Science Montpellier 2 University Linköpings universitet CNRS Sweden

More information

ASSUME a source over an alphabet size m, from which a sequence of n independent samples are drawn. The classical

ASSUME a source over an alphabet size m, from which a sequence of n independent samples are drawn. The classical IEEE TRANSACTIONS ON INFORMATION THEORY Large Alphabet Source Coding using Independent Coponent Analysis Aichai Painsky, Meber, IEEE, Saharon Rosset and Meir Feder, Fellow, IEEE arxiv:67.7v [cs.it] Jul

More information

Fixed-to-Variable Length Distribution Matching

Fixed-to-Variable Length Distribution Matching Fixed-to-Variable Length Distribution Matching Rana Ali Ajad and Georg Böcherer Institute for Counications Engineering Technische Universität München, Gerany Eail: raa2463@gail.co,georg.boecherer@tu.de

More information

Compressive Distilled Sensing: Sparse Recovery Using Adaptivity in Compressive Measurements

Compressive Distilled Sensing: Sparse Recovery Using Adaptivity in Compressive Measurements 1 Copressive Distilled Sensing: Sparse Recovery Using Adaptivity in Copressive Measureents Jarvis D. Haupt 1 Richard G. Baraniuk 1 Rui M. Castro 2 and Robert D. Nowak 3 1 Dept. of Electrical and Coputer

More information

On the Inapproximability of Vertex Cover on k-partite k-uniform Hypergraphs

On the Inapproximability of Vertex Cover on k-partite k-uniform Hypergraphs On the Inapproxiability of Vertex Cover on k-partite k-unifor Hypergraphs Venkatesan Guruswai and Rishi Saket Coputer Science Departent Carnegie Mellon University Pittsburgh, PA 1513. Abstract. Coputing

More information

REDUCTION OF FINITE ELEMENT MODELS BY PARAMETER IDENTIFICATION

REDUCTION OF FINITE ELEMENT MODELS BY PARAMETER IDENTIFICATION ISSN 139 14X INFORMATION TECHNOLOGY AND CONTROL, 008, Vol.37, No.3 REDUCTION OF FINITE ELEMENT MODELS BY PARAMETER IDENTIFICATION Riantas Barauskas, Vidantas Riavičius Departent of Syste Analysis, Kaunas

More information

On the Communication Complexity of Lipschitzian Optimization for the Coordinated Model of Computation

On the Communication Complexity of Lipschitzian Optimization for the Coordinated Model of Computation journal of coplexity 6, 459473 (2000) doi:0.006jco.2000.0544, available online at http:www.idealibrary.co on On the Counication Coplexity of Lipschitzian Optiization for the Coordinated Model of Coputation

More information

Scheduling Contract Algorithms on Multiple Processors

Scheduling Contract Algorithms on Multiple Processors Fro: AAAI Technical Report FS-0-04. Copilation copyright 200, AAAI (www.aaai.org). All rights reserved. Scheduling Contract Algoriths on Multiple Processors Daniel S. Bernstein, Theodore. Perkins, Shloo

More information