Markov Chain Monte Carlo Algorithms for Lattice Gaussian Sampling
|
|
- Paula Barber
- 5 years ago
- Views:
Transcription
1 Markov Chain Monte Carlo Algoriths for Lattice Gaussian Sapling Zheng Wang and Cong Ling Departent of EEE Iperial College London London, SW7 2AZ, United Kingdo Eail: z.wang10, Guillaue Hanrot École Norale Supérieure de Lyon LIP (CNRS, ENS Lyon, UCBL, INRIA) 46 Allée d Italie, Lyon Cedex 07, France Eail:Guillaue.Hanrot@ens-lyon.fr arxiv: v1 [cs.it] 7 May 2014 Abstract To be considered for an IEEE Jack Keil Wolf ISIT Student Paper Award. Sapling fro a lattice Gaussian distribution is eerging as an iportant proble in various areas such as coding and cryptography. The default sapling algorith Klein s algorith yields a distribution close to the lattice Gaussian only if the standard deviation is sufficiently large. In this paper, we propose the Markov chain Monte Carlo (MCMC) ethod for lattice Gaussian sapling when this condition is not satisfied. In particular, we present a sapling algorith based on Gibbs sapling, which converges to the target lattice Gaussian distribution for any value of the standard deviation. To iprove the convergence rate, a ore efficient algorith referred to as Gibbs-Klein sapling is proposed, which saples by using Klein s algorith. We show that Gibbs-Klein sapling yields a distribution close to the target lattice Gaussian, under a less stringent condition than that of the original Klein algorith. I. INTRODUCTION The lattice Gaussian distribution is eerging as a coon thee in various areas. In atheatics, Banaszczyk [1] firstly used it to prove the transference theores of lattices. In coding, it iics Shannon s Gaussian rando coding technique, yet perits lattice decoding. Forney applied the lattice Gaussian distribution to obtain the full shaping gain in lattice coding [2] (see also [3]). Recently, it has been used to achieve the capacity of the Gaussian channel [4] and to approach the secrecy capacity of the Gaussian wiretap channel [5], respectively. Sapling fro the lattice Gaussian has also been used in lattice decoding for the ulti-input ulti-output syste [6], [7]. In cryptography, lattice Gaussians have becoe a central tool in the construction of any priitives. Micciancio and Regev used it to propose lattice-based cryptosystes based on the worst-case hardness assuptions [8], and recently, it has underpinned the fully-hooorphic encryption for cloud coputing [9]. The key fact is again that a vector distributed as a lattice Gaussian centered at c with a sall standard deviation is typically very close to c. To illustrate why this ight be useful in cryptography, note that if one knows a short basis of the lattice, one can efficiently produce such a vector [10], while disclosing no inforation on the short basis since the lattice Gaussian distribution does not depend on the particular basis. Thus, in both coding and cryptography, efficient sapling algoriths for the lattice Gaussian as well as a good understanding on how the coplexity depends on the standard deviation is an iportant issue. However, in contrast to sapling fro the continuous Gaussian distribution, it is not at all straightforward to saple fro a discrete Gaussian distribution over a lattice. At present, the default sapling algorith for lattices is due to Klein, originally proposed for boundeddistance decoding [11] (see also [12], [13] for variations and [4] for an algorith for lattices of Construction A). It was shown in [10] that Klein s algorith saples within a negligible statistical distance fro the lattice Gaussian distribution only if the standard deviationσ ω( log n) ax 1 i n b i, where n is the lattice diension and b i s are the Gra- Schidt vectors of the lattice basis. Unfortunately, such a requireent of σ can be excessive, rendering Klein s algorith inapplicable to any cases of interest. Markov chain Monte Carlo (MCMC) ethods attept to saple fro the target distribution of interest by building a Markov chain, which randoly generate the next saple conditioned on the previous saples. As a ajor algorith of MCMC, Gibbs sapling [14] constructs a Markov chain which gradually converges to the target distribution by only considering univariate sapling at each step. In this paper, we introduce the Gibbs algorith into lattice Gaussian sapling and propose a ore efficient -based algorith naed as Gibbs-Klein sapling. In contrast to conventional ed sapling which is coputationally ore deanding, the proposed algorith takes advantages of Klein s algorith as a building. The proposed algoriths are applicable in the scenario σ < ω( log n) ax 1 i n b i. To the best of our knowledge, this is the first tie that MCMC ethods are used in lattice Gaussian distributions. Different fro previous works on Gibbs sapling for signal detection of finite constellations [15] [17], here we are concerned with countably infinite state spaces and with siulating Gaussian distributions over a lattice. It is worth pointing out that although the underlying Markov chain converges to the stationary distribution for all values of σ, the convergence is expected to becoe very slow when σ becoes sall, since for very sall σ we would solve the closest vector proble (CVP) and shortest vector proble (SVP) with high
2 Algorith 1 Klein s Algorith Input: B, σ, c Output: Bx Λ 1: let B QR and c Q T c 2: for i n,..., 1 do 3: let α i σ r and x i,i i c i 4: saple x i fro D Z,αi, x i 5: end for 6: return Bx n ji+1 ri,jxj probability. The rest of this paper is organized as follows. Section II introduces lattice Gaussian distributions and briefly reviews Klein s algorith. In Section III, the conventional Gibbs and the new Gibbs-Klein sapling algoriths are proposed for lattice Gaussians, followed by a theoretical analysis in Section IV. Section V presents the siulation results. II. LATTICE GAUSSIAN DISTRIBUTIONS Let B [b 1,...,b n ] R n consist of n linearly independent vectors. The n-diensional lattice Λ based on B is defined by Λ L(B) {Bx : x Z n }, (1) where B is known as the lattice basis. We define the Gaussian function centered at c R n for standard deviation σ > 0 as ρ σ,c (z) e z c 2 2σ 2, (2) for all z R n. Then, the discrete Gaussian distribution over Λ is defined as D Λ,σ,c (x) ρ σ,c(bx) ρ σ,c (Λ) 2σ 2 Bx c 2 x Z 1 n e 2σ 2 Bx c 2 (3) for all Bx Λ, where ρ σ,c (Λ) Bx Λ ρ σ,c(bx). An intuition of D Λ,σ,c (x) suggests that the closer lattice point Bx is to c, the higher probability it will be sapled. Thus, lattice Gaussian sapling can be applied to solve the CVP, and Klein s algorith was originally proposed for decoding [11]. As a randoized version of Babai s nearest-plane algorith (i.e., successive interference cancellation), Klein s algorith obtains a vector by sequentially sapling fro a 1-diensional conditional Gaussian distribution. As shown in Algorith 1, its operation has polynoial coplexity O(n 2 ) excluding QR decoposition. The paraeter σ is key to the distribution produced by Klein s algorith. Klein suggested σ in i b i / log n and this was followed/adapted in [6], [7]. In this case, Klein s algorith only yields a distribution that is lower-bounded by the Gaussian distribution. On the other hand, it was deonstrated in [10] that Klein s algorith actually saples fro D Λ,σ,c within a negligible statistical distance if σ ω( log n) ax 1 i n b i. (4) However, Gaussian sapling algoriths are lacking for the range σ < ω( log n) ax i b i. III. MCMC FOR LATTICE GAUSSIAN In this section, we introduce the concept of MCMC into lattice Gaussian sapling for the range of σ where Klein s algorith cannot reach. We further propose a ore efficient sapling algorith naed as Gibbs-Klein sapling to iprove the convergence rate. A. Gibbs Sapling for Lattice Gaussian Lattice Gaussian distribution D Λ,σ,c with σ < ω( log n) ax i b i can be seen as a coplex target distribution lacking direct sapling ethods. MCMC akes use of the conditional distribution as a tractable alternative to work with. Here we apply the Gibbs algorith to saple fro the original joint distribution D Λ,σ,c. Gibbs sapling eploys 1-diensional conditional distributions to construct the Markov chain [14], where all other variables in the distribution are unchanged in each step. In this way, we saple n rando variables fro the corresponding n univariate conditionals in a certain order instead of directly generating an n-diensional vector. Saples drawn fro the target joint distribution will be generated when the Markov chain reaches the stationary distribution. Specifically, in Gibbs sapling, each coordinate of x is sapled fro the following 1-diensional conditional distribution P(x t+1 i x t [ i] ) 2σ 2 Bxt+1 c 2 x t+1 i Z e 1 2σ 2 Bxt+1 c 2, (5) where 1 i n denotes the coordinate index of x, x t [ i] [xt 1,...,x t i 1,xt i+1,...,xt n] T, and t is the tie index of the Markov chain. It is noteworthy that there are any scan schees in Gibbs sapling and we apply the rando-scan in this paper, which eans the index i is randoly chosen at each step. The extension to other scan strategies is possible. By repeating such a procedure, an underlying Markov chain x t+1 [x t 1,...,xt i 1,xt+1 i,x t i+1,...,xt n ]T is induced, whose transition probability between two adjacent states is defined by the univariate Gibbs sapler, P(x t ;x t+1 ) P(x t+1 i x t [ i]). (6) Clearly, every two adjacent states of x differ fro each other by only one coordinate and it is easy to see that D Λ,σ,c stays invariant under such transitions. Algorith 2 gives the operation of Gibbs sapling for lattice Gaussian distributions. The initial rando variable x 0 can be chosen fro Z n arbitrarily or fro the output of a suboptial algorith, while the tie bound T is large enough to reach the stationary distribution D Λ,σ,c. With the transition probabilities (6), we ay for the infinite transition atrix P, whose (i,j)-th entry P(s i ;s j ) represents the probability of transferring to state s j fro the previous state s i. Denote by P t the transition atrix after t steps. We group in the following theore standard results about Gibbs sapling [18].
3 Algorith 2 Gibbs sapling for lattice Gaussian Input: B,σ,c,x 0 Output: x D Λ,σ,c as T 1: for t 1,..., T do 2: randoly choose coordinate indexifro{1,2,...,n} 3: saple x i fro P(x t i xt 1 4: update x t [x t 1 1,...,x t 1 i 1,x i,x t 1 i+1,...,xt 1 5: if Markov chain has reached stationarity then 6: output x t 7: end if 8: end for [ i] ) n ] T Proposition 1. Given the invariant distribution D Λ,σ,c, the Markov chain induced by the Gibbs sapler is irreducible, aperiodic and reversible (hence positive recurrent), and converges to the stationary distribution in the total variation (TV) distance as t : li t Pt (x; ) D Λ,σ,c TV 0, (7) for all states x Z n, where P t (x; ) denotes the row of P t corresponding to initial state x. According to Proposition 1, if tie perits to reach the stationary distribution, the proposed Gibbs sapler will draw saples fro D Λ,σ,c no atter what value σ takes, which eans the obstacle encountered by Klein s algorith is overcoe. B. Gibbs-Klein Sapling for Lattice Gaussian Although the afore-entioned Gibbs sapler will converge to the stationary distribution eventually, the way it functions by individually sapling only one coponent each tie leads to slow convergence. Especially, for lattice bases whose coponents are highly correlated with each other, the Markov chain induced by the standard Gibbs sapling can be trapped for a long tie. To hasten convergence of the Markov chain, a new sapling algorith cobining Gibbs and Klein algoriths is proposed in the sequel. The idea of ed sapling is to saple a of coponents of x at each step [19]. Intuitively, this will lead to a faster convergence rate, which is already shown in [14]. However, sapling a is generally ore costly than coponentwise sapling. We propose to use Klein s algorith for sapling; this leads to the Gibbs-Klein. At each step of the Markov chain, the proposed Gibbs-Klein sapling randoly picks up a of coponents of x to update. For convenience, an n n perutation atrix E is applied before ing so that the s are updated in a fixed order. Specifically, if E is rando, then Gibbs-Klein sapling on randoly chosen coponents will be equivalent to saple consecutive coponents of z in a fixed order, where z E 1 x and B BE. For siplicity, we always consider the fored by the first coponents of z, naely z [z 1,...,z ] T. After QR-decoposition B QR Algorith 3 Gibbs-Klein sapling for lattice Gaussian Input: B,σ,c,,x 0 ; Output: x fro a distribution close to D Λ,σ,c as T 1: for t 1,..., T do 2: randoly generate a perutation atrix E 3: Let B BE and z E 1 x 4: Let B QR and c Q T c 5: for i,..., 1 do 6: let α i σ 7: let z t 1 i c i ji+1 ri,jzt j n 8: saple z t i fro D Z,α i, z t 1 i 9: end for 10: update z t [z t ;zt 1 [ ] ]T 11: return x t Ez t j r +1 i,j zt 1 j 12: if Markov chain has reached stationarity then 13: output x t 14: end if 15: end for and calculating c Q T c, z i in the is sapled fro the following 1-diensional distribution with the backward order fro z to z 1 : i z t [ i] ) D Z,α i, z t i, (8) where α i σ r, i,i zt [ i] [zi+1 t+1,...,zt+1,zt +1,...,zt n ]T and z i t c i ji+1 ri,jzt+1 j n j +1 r i,j zt j. Algorith 3 gives the proposed Gibbs-Klein sapling, where z t+1 [z t+1 ;zt [ ] ] is obtained after each step, and zt [ ] [z+1,...,z t n] t T. The ipleentation given in Algorith 3 is not so efficient due to repeated QR decopositions; Optiizing for better efficiency will be pursued in the future. Note that the extension to other scan strategies is also possible. IV. ANALYSIS OF GIBBS-KLEIN SAMPLING In this section, we show that the proposed Gibbs-Klein sapling algorith can induce a reversible Markov chain within a negligible error. Fro (8) and by induction, the sapling probability of z t+1 conditioned on zt [ ] is given by zt [ ] ) i1 +1 i zt [ (+1 i)]). (9) The following lea gives a closed-for expression of this conditional probability within a negligible error and the proof follows [10]. Lea 1. For a given invariant distribution D Λ,σ,c, the transition probability zt [ ]) of Gibbs-Klein algorith is within negligible statistical distance of the following distribution D 2σ 2 Bz t+1 c 2 (10) z t+1 Z 2σ 2 Bz t+1 c 2
4 if σ ω( log) ax 1 i, where z t+1 [z t+1 ;zt [ ] ]. Proof: According to (8) and (9), we have i1 i1 zt [ ] ) i1 D Z,α+1 i, z t +1 i (zt+1 +1 i ) 2σ 2 i1(c +1 i j+1 i r+1 i,jzt+1 j ) 2 1 z t+1 +1 i Ze 2σ 2 (c +1 i j+1 i r+1 i,jzt+1 j ) 2 2σ 2 c rzt z t+1 +1 i Ze 2σ 2 (r +1 i,+1 iz t+1 +1 i c+1 i+ j+2 i r+1 i,jzt+1 j ) 2 ρ L(r),σ,c (z t+1 ) i1 ρ σ(r +1 i,+1 i Z+ξ), (11) where c i c i n j +1 r i,j zt, c [c j 1,...,c ] T, ξ j+2 i r +1 i,jzj t+1 c i+i and r is the segent of R with r 1,1 to r, in the diagonal. Clearly, the effect of the subvector z t [ ] is hidden in c i. In [20], it has been deonstrated that if σ > η ε (L(r)), then i1 ρ (( ) σ( Z+ξ) 1 ε i1 ρ,1] (12) σ( Z) 1+ε which eans i1 ρ σ( Z + ξ) can be substituted by i1 ρ σ( Z) within negligible errors when ε is sufficiently sall. As shown in [10], η ε (Λ) with negligible ε is upper bounded as η ε (Λ) ω( log n) ax 1 i n b i. Therefore, if σ ω( log ) ax 1 i, zt [ ]) shown in (11) can be rewritten as zt [ ] ) ρ L(r),σ,c(z t+1 ) i1 ρ σ( Z), (13) where represents equality up to a negligible error. Because the denoinator is independent of z t+1, zt [ ] and c, it can be viewed as a constant and the output has a lattice Gaussian distribution D L(r),σ,c (z t+1 ). Then we arrive at the following proposition. Proposition 2. Suppose σ ω( log) ax 1 i bi at each step so that the negligible statistical distance is absorbed by nuerical errors. Then, within nuerical errors, the Markov chain induced by the Gibbs-Klein sapler is irreducible, aperiodic and reversible (hence positive recurrent) and converges to the stationary distribution in the total variation distance as t : for all states x Z n. li t Pt (x; ) D Λ,σ,c TV 0 (14) Proof: Let s i and s j be two adjacent states in Gibbs- Klein sapling. For size, every two adjacent states in Gibbs-Klein sapling differ fro each other by at ost coponents. For convenience, we express the as s i [x (i),x [ ] ] and s j [x (j),x [ ] ], (15) where x (i) and x (j) denote the coponents belonging to s i and s j, respectively. Then, the transition probability of Gibbs-Klein sapling is P(s i ;s j ) P(x t+1 s j x t s i ) P(x t (i) xt+1 (j) xt [ ] ) (a) P(x t+1 (j) xt [ ] ) 2σ 2 Bsj c 2 x t+1 Z 2σ 2 Bxt+1 c 2, (16) where (a) is due to the fact that x t+1 is sapled only conditioned on x t [ ]. To show the Markov chain is irreducible, we note that given a state s one can attain with positive probability in one step any state s which shares > (n ) coponents with s. Now, if s and s have, say, d < n coponents in coon, there is always a positive probability that after each step they get exactly one ore coponent in coon. So we can go in n d steps fro one to the other. But as soon as > 2, we can assue that at the first step we get two ore coponents in coon, and then one at each further step, so we can go with positive probability in n d 1 steps. On the other hand, it is clear to see that the nuber of steps required to ove between any two states (can be the sae state) is arbitrary without any liitation to be a ultiple of soe integer. Put another way, the chain is not forced into soe cycle with fixed period between certain states. Therefore, the Markov chain is aperiodic. As for reversibility, it is no hard to check that the following relationship holds D Λ,σ,c (s i )P(s i ;s j ) D Λ,σ,c (s j )P(s j ;s i ) (17) with the sae expression 2σ 2 Bsi c 2 x Z n 2σ 2 Bx c 2 2σ 2 Bsj c 2 x t+1 Z 2σ 2 Bxt+1 c 2, (18) within negligible errors. Thus, the conclusion follows, copleting the proof. The advantages of Gibbs-Klein sapling are two-fold: copared with the conventional Gibbs sapling which only processes a single variate each tie, it is ore efficient to saple ultiple variates in a, iproving the convergence rate; on the other hand, it overcoes the liitation of Klein s sapling which requires large values of σ and extends lattice Gaussian sapling to the ore general case. V. SIMULATION RESULTS In this section, the perforances of various sapling schees are exeplified in the context of MIMO decoding. Specifically, we exaine the decoding error probabilities to assess the convergence rates. By sapling fro D Λ,σ,c, the closest lattice point will be returned with the highest probability, which iplies an effective approach to lattice decoding.
5 Bit Error Rate ZF Gibbs Sapling Gibbs Klein Sapling,2 Gibbs Klein Sapling,4 Klein Sapling ML Nuber of Iterations Fig. 1. Bit error rate versus the nuber of iterations for the uncoded 4 4 MIMO syste using 16-QAM. Fig. 1 depicts the bit error rates (BER) of different Gibbs saplers in a 4 4 uncoded MIMO syste with 16-QAM. This corresponds to lattice diension n 8. The perforances of zero-forcing (ZF) and axiu-likelihood (ML) decoding are also shown as bencharks. We assue a flat fading environent with fixed SNR (E b /N 0 15 db). The channel atrix H consists of uncorrelated coplex Gaussian fading gains with unit variance. Hx can be viewed as a lattice point in lattice Λ L(H) and detecting the transitted signal x corresponds to solving the CVP. Due to the finite constellation size, the ipleentation for discrete Gaussian sapling given in [6] is followed. Klein chose σ in 1 i n b i / log n and derived polynoial coplexityo(n Bx c 2 /in i b i 2 ) for his algorith to find the closest lattice point when it is not far fro c [11]. His derivation is essentially based on the assuption of a Gaussian distribution. However, we now know this choice of σ does not satisfy the soothing condition and thus his sapler does not really produce Gaussian saples [10]. Here, we follow Klein s choice of σ and apply the proposed Gibbs and Gibbs-Klein saplers to produce Gaussian saples fro the lattice. For a fair coparison, when the size is, we run sapling for n/ ties, and count this as a full iteration. This corresponds to one run of Klein s original algorith which saples n coponents. As shown in Fig. 1, the decoding perforance of all the sapling schees iprove with the nuber of iterations. With the sae nuber of iterations (hence the sae coplexity), the decoding perforance iproves with the size, which iplies a faster convergence rate. REFERENCES [1] W. Banaszczyk, New bounds in soe transference theores in the geoetry of nubers, Math. Ann., vol. 296, pp , [2] G. Forney and L.-F. Wei, Multidiensional constellations Part II: Voronoi constellations, IEEE J. Sel. Areas Coun., vol. 7, no. 6, pp , Aug [3] F. R. Kschischang and S. Pasupathy, Optial nonunifor signaling for Gaussian channels, IEEE Trans. Infor. Theory, vol. 39, pp , May [4] C. Ling and J.-C. Belfiore, Achieiving the AWGN channel capacity with lattice Gaussian coding, subitted to IEEE Trans. Infor. Theory, Mar. 2012, revised, Nov [Online]. Available: [5] C. Ling, L. Luzzi, J.-C. Belfiore, and D. Stehlé, Seantically secure lattice codes for the Gaussian wiretap channel, subitted to IEEE Trans. Infor. Theory, Oct. 2012, revised, Oct [Online]. Available: [6] S. Liu, C. Ling, and D. Stehle, Decoding by sapling: a randoized lattice algorith for bounded distance decoding, IEEE Trans. Infor. Theory, vol. 57, pp , Sep [7] Z. Wang and C. Ling, Decoding by sapling - part II: derandoization and soft-output decoding, IEEE Trans. Coun., vol. 61, no. 11, pp , Nov [8] D. Micciancio and O. Regev, Worst-case to average-case reuctions based on Gaussian easures, in Proc. Ann. Syp. Found. Coputer Science, Roe, Italy, Oct. 2004, pp [9] C. Gentry, A. Sahai, and B. Waters, Hooorphic encryption fro learning with errors: Conceptually-sipler, asyptotically-faster, attribute-based, in CRYPTO, [10] C. Gentry, C. Peikert, and V. Vaikuntanathan, Trapfoors for hard lattices and new cryptographic constructions, in Proc. 40th Ann. ACM Syp. Theory of Coput., Victoria, Canada, 2008, pp [11] P. Klein, Finding the closest lattice vector when it is unusually close, in ACM-SIAM Syp. Discr. Algoriths, 2000, pp [12] C. Peikert, An efficient and parallel Gaussian sapler for lattices, in CRYPTO, 2010, pp [13] Z. Brakerski, A. Langlois, C. Peikert, O. Regev, and D. Stehlé, Classical hardness of learning with errors, in STOC, 2013, pp [14] J. S. Liu, Monte Carlo Strategies in Scientific Coputing, New York: Springer-Verlag, [15] B. Farhang-Boroujeny, H. Zhu, and Z. Shi, Markov chain Monte Carlo algoriths for CDMA and MIMO counication systes, IEEE Trans. Signal Process., vol. 54, no. 5, pp , [16] B. Hassibi, M. Hansen, A. G. Diakis, H. A. J. Alshaary, and W. Xu, Optiized Markov chain Monte Carlo for signal detection in MIMO systes: an analysis of stationary distribution and ixing tie, [Online]. Available: [17] R. Chen, J. Liu, and X. Wang, Convergence analyses and coparisons of Markov chain Monte Carlo algoriths in digital counications, IEEE Trans. on Signal Process., vol. 50, no. 2, pp , [18] D. A. Levin, Y. Peres, and E. L. Wiler, Markov Chains and Mixing Tie, Aerican Matheatical Society, [19] G. O. Roberts and S. K. Sahu, Updating schees, correlation structure, ing and paraeterization for Gibbs sapler, J. Roy. Statist. Soc. Series B, 59(2): , [20] O. Regev, On lattice, learning with errors, rando linear codes, and cryptography, J. ACM, vol. 56, no. 6, pp. 34:1 34:40, ACKNOWLEDGMENT The authors would like to thank Daien Stehlé for helpful discussions.
III. Valuation Framework for CDS options
III. Valuation Fraework for CDS options In siulation, the underlying asset price is the ost iportant variable. The suitable dynaics is selected to describe the underlying spreads. The relevant paraeters
More informationOptimal Design Of English Auctions With Discrete Bid Levels*
Optial Design Of English Auctions With Discrete Bid Levels* E. David, A. Rogers and N. R. Jennings Electronics and Coputer Science, University of Southapton, Southapton, SO7 BJ, UK. {ed,acr,nrj}@ecs.soton.ac.uk.
More informationAnalysis of the purchase option of computers
Analysis of the of coputers N. Ahituv and I. Borovits Faculty of Manageent, The Leon Recanati Graduate School of Business Adinistration, Tel-Aviv University, University Capus, Raat-Aviv, Tel-Aviv, Israel
More informationThe Least-Squares Method for American Option Pricing
U.U.D.M. Proect Report 29:6 The Least-Squares Method for Aerican Option Pricing Xueun Huang and Xuewen Huang Exaensarbete i ateatik, 3 hp + 5 hp Handledare och exainator: Macie Kliek Septeber 29 Departent
More informationLECTURE 4: MIXED STRATEGIES (CONT D), BIMATRIX GAMES
LECTURE 4: MIXED STRATEGIES (CONT D), BIMATRIX GAMES Mixed Strategies in Matrix Gaes (revision) 2 ixed strategy: the player decides about the probabilities of the alternative strategies (su of the probabilities
More informationMonte Carlo Methods. Monte Carlo methods
ρ θ σ µ Monte Carlo Methos What is a Monte Carlo Metho? Rano walks The Metropolis rule iportance sapling Near neighbor sapling Sapling prior an posterior probability Exaple: gravity inversion The ovie
More informationIntroduction to Risk, Return and the Opportunity Cost of Capital
Introduction to Risk, Return and the Opportunity Cost of Capital Alexander Krüger, 008-09-30 Definitions and Forulas Investent risk There are three basic questions arising when we start thinking about
More informationFinancial Risk: Credit Risk, Lecture 1
Financial Risk: Credit Risk, Lecture 1 Alexander Herbertsson Centre For Finance/Departent of Econoics School of Econoics, Business and Law, University of Gothenburg E-ail: alexander.herbertsson@cff.gu.se
More informationTime Value of Money. Financial Mathematics for Actuaries Downloaded from by on 01/12/18. For personal use only.
Interest Accuulation and Tie Value of Money Fro tie to tie we are faced with probles of aking financial decisions. These ay involve anything fro borrowing a loan fro a bank to purchase a house or a car;
More informationCalculating Value-at-Risk Using the Granularity Adjustment Method in the Portfolio Credit Risk Model with Random Loss Given Default
Journal of Econoics and Manageent, 016, Vol. 1, No., 157-176 Calculating Value-at-Risk Using the Granularity Adjustent Method in the Portfolio Credit Risk Model with Rando Loss Given Default Yi-Ping Chang
More informationA NUMERICAL EXAMPLE FOR PORTFOLIO OPTIMIZATION. In 2003, I collected data on 20 stocks, which are listed below: Berkshire-Hathaway B. Citigroup, Inc.
A NUMERICAL EXAMPLE FOR PORTFOLIO OPTIMIZATION In 3, I collected data on stocks, which are listed below: Sybol ADBE AMZN BA BRKB C CAT CSCO DD FDX GE GLW GM INTC JNJ KO MO MSFT RTN SBC Nae Adobe Systes
More informationProject selection by using AHP and Bernardo Techniques
International Journal of Huanities and Applied Sciences (IJHAS) Vol. 5, No., 06 ISSN 77 4386 Project selection by using AHP and Bernardo Techniques Aza Keshavarz Haddadha, Ali Naazian, Siaak Haji Yakhchali
More informationRecursive Inspection Games
Recursive Inspection Gaes Bernhard von Stengel February 7, 2016 arxiv:1412.0129v2 [cs.gt] 7 Feb 2016 Abstract We consider a sequential inspection gae where an inspector uses a liited nuber of inspections
More informationTruthful Randomized Mechanisms for Combinatorial Auctions
Truthful Randoized Mechaniss for Cobinatorial Auctions Shahar Dobzinski Noa Nisan Michael Schapira March 20, 2006 Abstract We design two coputationally-efficient incentive-copatible echaniss for cobinatorial
More informationPRODUCTION COSTS MANAGEMENT BY MEANS OF INDIRECT COST ALLOCATED MODEL
PRODUCTION COSTS MANAGEMENT BY MEANS OF INDIRECT COST ALLOCATED MODEL Berislav Bolfek 1, Jasna Vujčić 2 1 Polytechnic Slavonski Brod, Croatia, berislav.bolfek@vusb.hr 2 High school ''Matija Antun Reljković'',
More informationReceived May 27, 2009; accepted January 14, 2011
MATHEMATICAL COMMUNICATIONS 53 Math. Coun. 6(20), 53 538. I σ -Convergence Fatih Nuray,, Hafize Gök and Uǧur Ulusu Departent of Matheatics, Afyon Kocatepe University, 03200 Afyonkarahisar, Turkey Received
More informationHistorical Yield Curve Scenarios Generation without Resorting to Variance Reduction Techniques
Working Paper Series National Centre of Copetence in Research Financial Valuation and Risk Manageent Working Paper No. 136 Historical Yield Curve Scenarios Generation without Resorting to Variance Reduction
More informationMachine Learning in Computer Vision Markov Random Fields Part II
Machine Learning in Computer Vision Markov Random Fields Part II Oren Freifeld Computer Science, Ben-Gurion University March 22, 2018 Mar 22, 2018 1 / 40 1 Some MRF Computations 2 Mar 22, 2018 2 / 40 Few
More informationAn alternative route to performance hypothesis testing Received (in revised form): 7th November, 2003
An alternative route to perforance hypothesis testing Received (in revised for): 7th Noveber, 3 Bernd Scherer heads Research for Deutsche Asset Manageent in Europe. Before joining Deutsche, he worked at
More informationOptimal Dynamic Pricing for Assembly Product Supply Chain
Aerican Journal of Operations Manageent and Inforation Systes 7; (: 54-7 http://www.sciencepublishinggroup.co/j/ajois doi:.648/j.ajois.7.8 Optial Dynaic Pricing for Assebly Product Supply Chain Yufang
More informationNeural Network Model of Pricing Health Care Insurance
Neural Network Model of Pricing Health Care Insurance Abstract To pricing health insurance plan statisticians use atheatical odels to analysis custoer s future health condition. General Addictive Model
More informationImplementation of MADM Methods in Solving Group Decision Support System on Gene Mutations Detection Simulation
Ipleentation of MADM Methods in Solving Group Decision Support Syste on Gene Mutations Detection Siulation Eratita *1, Sri Hartati *2, Retantyo Wardoyo *2, Agus Harjoko *2 *1 Departent of Inforation Syste,
More informationResearch Article Analysis on the Impact of the Fluctuation of the International Gold Prices on the Chinese Gold Stocks
Discrete Dynaics in Nature and Society, Article ID 308626, 6 pages http://dx.doi.org/10.1155/2014/308626 Research Article Analysis on the Ipact of the Fluctuation of the International Gold Prices on the
More informationPerformance Analysis of Online Anticipatory Algorithms for Large Multistage Stochastic Integer Programs
Perforance Analysis of Online Anticipatory Algoriths for Large Multistage Stochastic Integer Progras Luc Mercier and Pascal Van Hentenryck Brown University {ercier,pvh}@cs.brown.edu Abstract Despite significant
More informationEvaluation on the Growth of Listed SMEs Based on Improved Principal Component Projection Method
Proceedings of the 7th International Conference on Innovation & Manageent 519 Evaluation on the Growth of Listed SMEs Based on Iproved Principal Coponent Projection Method Li Li, Ci Jinfeng Shenzhen Graduate
More informationResearch on the Management Strategy from the Perspective of Profit and Loss Balance
ISSN: 2278-3369 International Journal of Advances in Manageent and Econoics Available online at: www.anageentjournal.info RESEARCH ARTICLE Research on the Manageent Strategy fro the Perspective of Profit
More informationCHAPTER 2: FUTURES MARKETS AND THE USE OF FUTURES FOR HEDGING
CHAPER : FUURES MARKES AND HE USE OF FUURES FOR HEDGING Futures contracts are agreeents to buy or sell an asset in the future for a certain price. Unlike forward contracts, they are usually traded on an
More informationResearch on Entrepreneur Environment Management Evaluation Method Derived from Advantage Structure
Research Journal of Applied Sciences, Engineering and Technology 6(1): 160-164, 2013 ISSN: 2040-7459; e-issn: 2040-7467 Maxwell Scientific Organization, 2013 Subitted: Noveber 08, 2012 Accepted: Deceber
More informationResearch Article A Two-Stage Model for Project Optimization in Transportation Infrastructure Management System
Matheatical Probles in Engineering, Article ID 914515, 8 pages http://dx.doi.org/10.1155/2014/914515 Research Article A wo-stage Model for Project Optiization in ransportation Infrastructure Manageent
More informationCorrective Taxation versus Liability
Aerican Econoic Review: Papers & Proceedings 2011, 101:3, 273 276 http://www.aeaweb.org/articles.php?doi=10.1257/aer.101.3.273 Law and Econoics Corrective Taxation versus Liability By Steven Shavell* Since
More informationImprecise Probabilities in Non-cooperative Games
7th International Syposiu on Iprecise Probability: Theories and Applications, Innsbruck, Austria, 2011 Iprecise Probabilities in Non-cooperative Gaes Robert Nau Fuqua School of Business Duke University
More informationm-string Prediction
Figure 1. An =3 strategy. -string Prediction 000 0 001 1 010 1 011 0 100 0 101 1 110 0 111 1 10 Figure 2: N=101 s=1 9 8 7 6 σ 5 4 3 2 1 0 0 2 4 6 8 10 12 14 16 42 Figure 3: N=101 s=2 15 10 σ 5 0 0 2 4
More informationMAT 3788 Lecture 3, Feb
The Tie Value of Money MAT 3788 Lecture 3, Feb 010 The Tie Value of Money and Interest Rates Prof. Boyan Kostadinov, City Tech of CUNY Everyone is failiar with the saying "tie is oney" and in finance there
More information... About Higher Moments
WHAT PRACTITIONERS NEED TO KNOW...... About Higher Moents Mark P. Kritzan In financial analysis, a return distribution is coonly described by its expected return and standard deviation. For exaple, the
More informationMADM Methods in Solving Group Decision Support System on Gene Mutations Detection Simulation
MADM Methods in Solving Group Decision Support Syste on Gene Mutations Detection Siulation Eratita *1, Sri Hartati *2, Retantyo Wardoyo *2, Agus Harjoko *2 *1 Departent of Inforation Syste, Coputer Science
More informationThe finite-dimensional Witsenhausen. counterexample
The finite-diensional Witsenhausen counterexaple Pulkit rover, Se Yong Park and Anant Sahai Departent of EECS, University of California at Berkeley, CA-947, USA {pulkit, sahai, separk}@eecs.berkeley.edu
More informationExpert Advisor (EA) Evaluation System Using Web-based ELECTRE Method in Foreign Exchange (Forex) Market
The 2 nd International Conference on Energy, Environent and Inforation Syste (ICENIS 2017) 15 th 16 th August 2017, Universitas Diponegoro, Searang, Indonesia Expert Advisor (EA) Evaluation Syste Using
More informationPuerto Rico, US, Dec 2013: 5-year sentence for pricefixing
Dynaic oligopoly theory Collusion price coordination Illegal in ost countries - Explicit collusion not feasible - Legal exeptions Recent EU cases - Banking approx..7 billion Euros in fines (03) - Cathodic
More informationEVOLVING PARAMETERS OF LOGIT MODEL USING GENETIC ALGORITHMS
EVOLVING PARAMETERS OF LOGIT MODEL USING GENETIC ALGORITHMS Ming Zhong, University of New Brunswick Pawan Lingras and Will Blades, Saint Mary s University John Douglas Hunt, University of Calgary Introduction
More informationLinear Dispersion Over Time and Frequency
Linear Dispersion Over Time and Frequency Jinsong Wu and Steven D Blostein Department of Electrical and Computer Engineering Queen s University, Kingston, Ontario, Canada, K7L3N6 Email: {jwu, sdb@eequeensuca
More informationWhy Do Large Investors Disclose Their Information?
Why Do Large Investors Disclose Their Inforation? Ying Liu Noveber 7, 2017 Abstract Large investors often advertise private inforation at private talks or in the edia. To analyse the incentives for inforation
More informationNon-Linear Programming Approach for Optimization of Construction Project s Control System
Non-Linear Prograing Approach for Optiization of Construction Project s Control Syste Yusrizal Lubis 1 and Zuhri 1,2 Sekolah Tinggi Teknik Harapan Medan Indonesia, Sekolah Tinggi Ilu Manajeen Suka Medan
More informationPing Cao School of Economic and Management, Shanghai Institute of Technology, Shanghai, China
doi:10.21311/001.39.7.27 Payent Scheduling Probles of Software Projects fro a Bilateral Perspective Ping Cao School of Econoic and Manageent, Shanghai Institute of Technology, Shanghai, China Jian Zhang*
More informationThe finite-dimensional Witsenhausen. counterexample
The finite-diensional Witsenhausen counterexaple Pulkit rover, Anant Sahai and Se Yong Park Departent of EECS, University of California at Berkeley, CA-947, USA {pulkit, sahai, separk}@eecs.berkeley.edu
More informationASSESSING CREDIT LOSS DISTRIBUTIONS FOR INDIVIDUAL BORROWERS AND CREDIT PORTFOLIOS. BAYESIAN MULTI-PERIOD MODEL VS. BASEL II MODEL.
ASSESSING CREIT LOSS ISTRIBUTIONS FOR INIVIUAL BORROWERS AN CREIT PORTFOLIOS. BAYESIAN ULTI-PERIO OEL VS. BASEL II OEL. Leonid V. Philosophov,. Sc., Professor, oscow Coittee of Bankruptcy Affairs. 33 47
More informationRisk Sharing, Risk Shifting and the Role of Convertible Debt
Risk Sharing, Risk Shifting and the Role of Convertible Debt Saltuk Ozerturk Departent of Econoics, Southern Methodist University Abstract This paper considers a financial contracting proble between a
More informationVariance Swaps and Non-Constant Vega
Variance Swaps and Non-Constant Vega David E. Kuenzi Head of Risk anageent and Quantitative Research Glenwood Capital Investents, LLC 3 N. Wacker Drive, Suite 8 Chicago, IL 666 dkuenzi@glenwood.co Phone
More information\Notes" Yuri Y. Boykov. 4 August Analytic approximation of. In this chapter we apply the method of lines to approximate values of several
\Notes" Yuri Y. Boyov 4 August 1996 Part II Analytic approxiation of soe exotic options 1 Introduction In this chapter we apply the ethod of lines to approxiate values of several options of both European
More informationOptimal resource allocation among transit agencies for fleet management
Optial resource allocation aong transit agencies for fleet anageent To V Mathew a, Snehaay Khasnabis b, Sabyasachee Mishra b a Departent of Civil Engineering, Indian Institute of Technology Bobay, Powai,
More informationResearch Article Concession Period Decision Models for Public Infrastructure Projects Based on Option Games
Matheatical Probles in Engineering Volue 2015, Article ID 671764, 9 pages http://dx.doi.org/10.1155/2015/671764 Research Article Concession Period Decision Models for Public Infrastructure Projects Based
More informationRealized Variance and IID Market Microstructure Noise
Realized Variance and IID Market Microstructure Noise Peter R. Hansen a, Asger Lunde b a Brown University, Departent of Econoics, Box B,Providence, RI 02912, USA b Aarhus School of Business, Departent
More informationThe Social Accounting Matrix (SAM)
Università degli Studi di Roa "Tor Vergata The Social Accounting Matrix (SAM) Methodology and Web site Federica Alfani 17 Maggio 2009 The Social Accounting Matrix (SAM) Iportant aspects related to this
More informationA New Lattice-Based Cryptosystem Mixed with a Knapsack
A New Lattice-Based Cryptosystem Mixed with a Knapsack Yanbin Pan and Yingpu Deng and Yupeng Jiang and Ziran Tu Key Laboratory of Mathematics Mechanization Academy of Mathematics and Systems Science,Chinese
More informationPROBE A multicriteria decision support system for portfolio robustness evaluation
ISSN 2041-4668 (Online) PROBE A ulticriteria decision support syste for portfolio robustness evaluation João Carlos Lourenço 1 and Carlos A. Bana e Costa 1,2 1 CEG-IST, Centre for Manageent Studies of
More informationStaff Memo N O 2005/11. Documentation of the method used by Norges Bank for estimating implied forward interest rates.
N O 005/ Oslo Noveber 4, 005 Staff Meo Departent for Market Operations and Analysis Docuentation of the ethod used by Norges Bank for estiating iplied forward interest rates by Gaute Myklebust Publications
More informationDepartment of Econometrics and Business Statistics
ISSN 440-77X Australia Departent of Econoetrics and Business Statistics http://www.buseco.onash.edu.au/depts/ebs/pubs/wpapers/ Applications of Inforation Measures to Assess Convergence in the Central Liit
More informationIntroduction to the Lattice Crypto Day
MAYA Introduction to the Lattice Crypto Day Phong Nguyễn http://www.di.ens.fr/~pnguyen May 2010 Summary History of Lattice-based Crypto Background on Lattices Lattice-based Crypto vs. Classical PKC Program
More informationWho Gains and Who Loses from the 2011 Debit Card Interchange Fee Reform?
No. 12-6 Who Gains and Who Loses fro the 2011 Debit Card Interchange Fee Refor? Abstract: Oz Shy In October 2011, new rules governing debit card interchange fees becae effective in the United States. These
More information"Inflation, Wealth And The Real Rate Of Interest"
Econoic Staff Paper Series Econoics 3-1975 "Inflation, Wealth And The Real Rate Of Interest" Walter Enders Iowa State University Follow this and additional works at: http://lib.dr.iastate.edu/econ_las_staffpapers
More informationFinancialTimeSeriesRecentTrendsinEconometrics
Global Journal of Manageent and Business Research Finance Volue 13 Issue 5 Version 1.0 Year 013 Type: Double Blind Peer Reviewed International Research Journal Publisher: Global Journals Inc. (USA) Online
More informationA Complete Example of an Optimal. Two-Bracket Income Tax
A Coplete Exaple of an Optial Two-Bracket Incoe Tax Jean-François Wen Departent of Econoics University of Calgary March 6, 2014 Abstract I provide a siple odel that is solved analytically to yield tidy
More informationProduction, Process Investment and the Survival of Debt Financed Startup Firms
Babson College Digital Knowledge at Babson Babson Faculty Research Fund Working Papers Babson Faculty Research Fund 00 Production, Process Investent and the Survival of Debt Financed Startup Firs S. Sinan
More informationOn Complexity of Multistage Stochastic Programs
On Complexity of Multistage Stochastic Programs Alexander Shapiro School of Industrial and Systems Engineering, Georgia Institute of Technology, Atlanta, Georgia 30332-0205, USA e-mail: ashapiro@isye.gatech.edu
More informationA Multi-Objective Optimization Model for Transit Fleet Resource Allocation
Mishra et al. 0 0 0 0 0 A Multi-Objective Optiization Model for Transit Fleet Resource Allocation By Sabyasachee Mishra, Ph.D., P.E. Assistant Professor Departent of Civil Engineering University of Mephis
More informationLattice Coding and its Applications in Communications
Lattice Coding and its Applications in Communications Alister Burr University of York alister.burr@york.ac.uk Introduction to lattices Definition; Sphere packings; Basis vectors; Matrix description Codes
More informationUncertain Efficiency Gains and Merger Policy
Uncertain Efficiency Gains and Merger Policy Mariana Cunha Paula Sarento Hélder Vasconcelos February 17, 2014 Abstract This paper studies the role of uncertainty in erger control and in erger decisions.
More informationLattice Cryptography: Introduction and Open Problems
Lattice Cryptography: Introduction and Open Problems Daniele Micciancio Department of Computer Science and Engineering University of California, San Diego August 2015 Daniele Micciancio (UCSD) Lattice
More informationCatastrophe Insurance Products in Markov Jump Diffusion Model
Catastrophe Insurance Products in Markov Jup Diffusion Model (Topic of paper: Risk anageent of an insurance enterprise) in Shih-Kuei Assistant Professor Departent of Finance National University of Kaohsiung
More informationAN ANALYSIS OF EQUITY IN INSURANCE. THE MATHEMATICAL APPROACH OF RISK OF RUIN FOR INSURERS
Iulian Mircea AN ANALYSIS OF EQUITY IN INSURANCE. THE MATHEMATICAL APPROACH OF RISK OF RUIN FOR INSURERS A.S.E. Bucure ti, CSIE, Str.Mihail Moxa nr. 5-7, irceaiulian9@yahoo.co, Tel.074.0.0.38 Paul T n
More informationEstimating Nonlinear Models With Multiply Imputed Data
Estiating onlinear Models With Multiply Iputed Data Catherine Phillips Montalto 1 and Yoonkyung Yuh 2 Repeated-iputation inference (RII) techniques for estiating nonlinear odels with ultiply iputed data
More informationOptimal rebalancing of portfolios with transaction costs assuming constant risk aversion
Optimal rebalancing of portfolios with transaction costs assuming constant risk aversion Lars Holden PhD, Managing director t: +47 22852672 Norwegian Computing Center, P. O. Box 114 Blindern, NO 0314 Oslo,
More informationSTOCK PRICE AND EXCHANGE RATE CAUSALITY: THE CASE OF FOUR ASEAN COUNTRIES
Stock Price and Exchange Rate Causality: The Case of Four Asean Countries STOCK PRICE AND EXCHANGE RATE CAUSALITY: THE CASE OF FOUR ASEAN COUNTRIES D. Agus Harjito, Indonesian Islaic University Carl B.
More informationCross-Packing Lattices for the Rician Fading Channel
Cross-Packing Lattices for the Rician Fading Channel Amin Sakzad, Anna-Lena Trautmann, and Emanuele Viterbo Department of Electrical and Computer Systems Engineering, Monash University. Abstract We introduce
More informationCOS 513: Gibbs Sampling
COS 513: Gibbs Sampling Matthew Salesi December 6, 2010 1 Overview Concluding the coverage of Markov chain Monte Carlo (MCMC) sampling methods, we look today at Gibbs sampling. Gibbs sampling is a simple
More informationNBER WORKING PAPER SERIES THE LEVERAGE EFFECT PUZZLE: DISENTANGLING SOURCES OF BIAS AT HIGH FREQUENCY. Yacine Ait-Sahalia Jianqing Fan Yingying Li
NBER WORKING PAPER SERIES THE LEVERAGE EFFECT PUZZLE: DISENTANGLING SOURCES OF BIAS AT HIGH FREQUENCY Yacine Ait-Sahalia Jianqing Fan Yingying Li Working Paper 17592 http://www.nber.org/papers/w17592 NATIONAL
More information1102 IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 51, NO. 3, MARCH Genyuan Wang and Xiang-Gen Xia, Senior Member, IEEE
1102 IEEE TRANSACTIONS ON INFORMATION THEORY, VOL 51, NO 3, MARCH 2005 On Optimal Multilayer Cyclotomic Space Time Code Designs Genyuan Wang Xiang-Gen Xia, Senior Member, IEEE Abstract High rate large
More informationAn Analytical Solution to Reasonable Royalty Rate Calculations a
-0- An Analytical Solution to Reasonable Royalty Rate Calculations a Willia Choi b Roy Weinstein c July 000 Abstract The courts are increasingly encouraging use of ore rigorous, scientific approaches to
More informationCPSC 540: Machine Learning
CPSC 540: Machine Learning Monte Carlo Methods Mark Schmidt University of British Columbia Winter 2019 Last Time: Markov Chains We can use Markov chains for density estimation, d p(x) = p(x 1 ) p(x }{{}
More informationConstruction and behavior of Multinomial Markov random field models
Graduate Theses and Dissertations Iowa State University Capstones, Theses and Dissertations 2010 Construction and behavior of Multinomial Markov random field models Kim Mueller Iowa State University Follow
More informationAn Index Number Formula Problem: the Aggregation of Broadly Comparable Items
An Index Nuber Forula Proble: the Aggregation of Broadly Coparable Ites Mick Silver* International Monetary Fund Presentation to the (Ottawa) International Working Group on Price Indices (May 27 29, 2009)
More informationSection on Survey Research Methods
Using the Statistics of Incoe Division s Saple Data to Reduce Measureent and Processing Error in Sall Area Estiates Produced fro Adinistrative Tax Records Kiberly Henry, Partha Lahiri, and Robin Fisher
More informationIntroduction Ideal lattices Ring-SIS Ring-LWE Other algebraic lattices Conclusion. Ideal Lattices. Damien Stehlé. ENS de Lyon. Berkeley, 07/07/2015
Ideal Lattices Damien Stehlé ENS de Lyon Berkeley, 07/07/2015 Damien Stehlé Ideal Lattices 07/07/2015 1/32 Lattice-based cryptography: elegant but impractical Lattice-based cryptography is fascinating:
More informationStochastic Analysis of Life Insurance Surplus
Stochastic Analysis of Life Insurance Surplus Natalia Lysenko, Gary Parker Abstract The behaviour of insurance surplus over tie for a portfolio of hoogeneous life policies in an environent of stochastic
More informationStudy on the Risk Transmission Mechanism of Rural Banks in China *
06 nd International Conference on Modern Education and Social Science (MESS 06) ISBN: 978--60595-346-5 Study on the isk Transission Mechanis of ural Banks in China * Shan-Shan WANG,,a, Jian-Guo WEI,b,
More informationOPTIMIZATION APPROACHES IN RISK MANAGEMENT: APPLICATIONS IN FINANCE AND AGRICULTURE
OPTIMIZATION APPROACHES IN RISK MANAGEMENT: APPLICATIONS IN FINANCE AND AGRICULTURE By CHUNG-JUI WANG A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
More informationWhat should be weights on the three Titlecurrencies for a common currency ba Asia?
What should be weights on the the Titlecurncies for a coon curncy ba Asia? Author(s) Ogawa, Eiji; Kawasaki, Kentaro Citation Issue 2003-12 Date Type Technical Report Text Version publisher URL http://hdl.handle.net/10086/14062
More informationMix-nets for long-term privacy
Mix-nets for long-term privacy October 2017 Núria Costa nuria.costa@scytl.com Index 1. Introdution: Previous work 2. Mix-nets 3. Lattice-based cryptography 4. Proof of a shuffle for lattice-based cryptography
More informationARTICLE IN PRESS. Journal of Mathematical Economics xxx (2008) xxx xxx. Contents lists available at ScienceDirect. Journal of Mathematical Economics
Journal of Matheatical Econoics xxx (28) xxx xxx Contents lists available at ScienceDirect Journal of Matheatical Econoics journal hoepage: www.elsevier.co/locate/jateco 1 1 2 2 3 4 5 6 7 8 9 1 11 12 13
More informationPortfolio decision analysis with PROBE: Addressing costs of not financing projects
Portfolio decision analysis with PROBE: Addressing costs of not financing proects JOÃO CARLOS LOURENÇO, CARLOS A. BANA E COSTA, JOÃO OLIVEIRA SOARES Departent of Engineering and Manageent (DEG) and Centre
More informationEarnings per share up by 21.2% to EUR 6.53, allowing increased dividend of EUR 2.40 per share
Press Release Mülhei an der Ruhr, March 21, 2013 Brenntag once again reports significantly iproved earnings and exceeds iddle of the guidance range Gross profit increased to 1,925.7 illion Growth in operating
More informationFIT5124 Advanced Topics in Security. Lecture 1: Lattice-Based Crypto. I
FIT5124 Advanced Topics in Security Lecture 1: Lattice-Based Crypto. I Ron Steinfeld Clayton School of IT Monash University March 2016 Acknowledgements: Some figures sourced from Oded Regev s Lecture Notes
More informationAn agent-based analysis of main cross-border balancing arrangements for Northern Europe
1 An agent-based analysis of ain cross-border balancing arrangeents for Northern Europe R. A. C. van der Vee A. Abbasy, and R. A. Hakvoort Abstract The topic of electricity balancing arket integration
More informationTotal PS TG. Budgeted production levels can be calculated as follows:
U. ;' cn '.:. \.' >>.:---"--^ '-.'" * i--.'. * ::-;.v>"--:'i.-^ -7 -..=../.-' "-. " '.:.' Ill all it.;? s Solution Total PS TG Sales units 6,000 5,000 1,000 Sales value $605,000 $475,000 $130,000 Workings
More informationImprovement and Efficient Implementation of a Lattice-based Signature scheme
Improvement and Efficient Implementation of a Lattice-based Signature scheme, Johannes Buchmann Technische Universität Darmstadt TU Darmstadt August 2013 Lattice-based Signatures1 Outline Introduction
More informationCPSC 540: Machine Learning
CPSC 540: Machine Learning Monte Carlo Methods Mark Schmidt University of British Columbia Winter 2018 Last Time: Markov Chains We can use Markov chains for density estimation, p(x) = p(x 1 ) }{{} d p(x
More informationPractical example of an Economic Scenario Generator
Practical example of an Economic Scenario Generator Martin Schenk Actuarial & Insurance Solutions SAV 7 March 2014 Agenda Introduction Deterministic vs. stochastic approach Mathematical model Application
More informationFree-riding in International Environmental Agreements:
School of Econoic Sciences Working Paper Series WP 2009-08 Free-riding in International Environental Agreeents: A Signaling Approach to Non- Enforceable Treaties By Ana Espínola Arredondo and Félix Muñoz
More informationA class of coherent risk measures based on one-sided moments
A class of coherent risk measures based on one-sided moments T. Fischer Darmstadt University of Technology November 11, 2003 Abstract This brief paper explains how to obtain upper boundaries of shortfall
More informationWe can also develop useful upper bounds. The value of the arithmetic Asian option is
Aian option Introduction Aian option are popular in currency and coodity arket becaue they offer a cheaper ethod of hedging expoure to regular periodic cah flow they are le uceptible to anipulation of
More informationBERMUDA NATIONAL PENSION SCHEME (GENERAL) REGULATIONS 1999 BR 82 / 1999
QUO FA T A F U E R N T BERMUDA NATIONAL PENSION SCHEME (GENERAL) REGULATIONS 1999 BR 82 / 1999 TABLE OF CONTENTS 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Citation Interpretation PART 1 PRELIMINARY PART II REGISTRATION
More information