Sistema konkurrenteak simulatzeko kernel bat: multiprozesadore sistemak

Jarraian, hainbat hilabetetan zehar garatutako proiektuaren deskribapena biltzen duen memoria dugu eskuragarri. Proiektu hau, sistema konkurrenteen simulazioan zentratzen da eta horretarako, mota honetako sistemen arloan hain erabiliak diren Petri Sareak lantzeaz gain, simulatzaile bat programatzeko informazio nahikoa ere barneratzen ditu. Gertaera diskretuko simulatzaile estatistiko batean oinarrituko da proiektuaren garapena, helburua izanik Petri Sareen bidez formalizatzen diren sistemak simulatzeko softwarea osatzea. Proiektuaren helburua da objektuetara zuzendutako hizkuntzaren bidez, Java hizkuntzaren bidez alegia, simulatzailearen programazioa erraztea eta ingurune honen baliabideak erabiltzea, bereziki XML teknologiari lotutakoak. Proiektu hau, bi zati nagusitan banatzen dela esan daiteke. Lehenengo zatiari dagokionez, konputazio munduan simulazioa aurkeztu eta honi buruzko behar adina informazio emango da. Hau, oso erabilgarria izango da programatuko den simulatzailearen nondik norakoak ulertu eta klase desberdinen inplementazioa egin ahal izateko. Horrez gain, zorizko aldagaiak eta hauen simulazioa ere islatzen dira, simulazio prozesu hori ahalik eta era errealean gauzatzeko helburuarekin. Ondoren, Petri Sareak aurkeztuko dira, hauen ezaugarri eta sailkapen desberdinak goraipatuz. Gainera, Petri Sareak definitzeko XML lengoaia erabiliko denez, mota honetako dokumentu eta eskemak aztertuko dira, hauek, garatuko den aplikazioaren oinarri izango direlarik. Bestalde, aplikazioaren muin izango diren klaseen diseinu eta inplementazioak bildu dira azken aurreko kapituluan. Alde batetik, erabili den DOM egituraren inguruko informazioa islatzen da eta bestetik, XML-tik habiatuz lortuko diren PetriNet instantziak maneiatzeko ezinbestekoak diren Java klaseen kodeak erakusten dira. Amaitzeko, egileak ateratako ondorioez gain, proiektuaren garapen prozesuan erabili den bibliografiaren berri ere ematen da.

Some properties of the coefficients of cyclotomic polynomials

An explicit formula is obtained for the coefficients of the cyclotomic polynomial F_n(x), where n is the product of two distinct odd primes. A recursion formula and a lower bound and an improvement of Bang’s upper bound for the coefficients of F_n(x) are also obtained, where n is the product of three distinct primes. The cyclotomic coefficients are also studied when n is the product of four distinct odd primes. A recursion formula and upper bounds for its coefficients are obtained. The last chapter includes a different approach to the cyclotomic coefficients. A connection is obtained between a certain partition function and the cyclotomic coefficients when n is the product of an arbitrary number of distinct odd primes. Finally, an upper bound for the coefficients is derived when n is the product of an arbitrary number of distinct and odd primes.

Contributions to the efficient use of general purpose coprocessors: kernel density estimation as case study

142 p.

Copula-based Kernel Dependency Measures

The paper presents a new copula based method for measuring dependence between random variables. Our approach extends the Maximum Mean Discrepancy to the copula of the joint distribution. We prove that this approach has several advantageous properties. Similarly to Shannon mutual information, the proposed dependence measure is invariant to any strictly increasing transformation of the marginal variables. This is important in many applications, for example in feature selection. The estimator is consistent, robust to outliers, and uses rank statistics only. We derive upper bounds on the convergence rate and propose independence tests too. We illustrate the theoretical contributions through a series of experiments in feature selection and low-dimensional embedding of distributions.

Structure Discovery in Nonparametric Regression through Compositional Kernel Search

Despite its importance, choosing the structural form of the kernel in nonparametric regression remains a black art. We define a space of kernel structures which are built compositionally by adding and multiplying a small number of base kernels. We present a method for searching over this space of structures which mirrors the scientific discovery process. The learned structures can often decompose functions into interpretable components and enable long-range extrapolation on time-series datasets. Our structure search method outperforms many widely used kernels and kernel combination methods on a variety of prediction tasks.