Petak, 5. mart 2004. u 14h, sala 2 MI SANU :

Radni sastanak (zajednicki sastanak sa Odeljenjem za mehaniku)
ANALIZA REZULTATA NA MATEMATICKIM PROJEKTIMA PRI MNTR

Rezime: Clanovi Komisije za matematiku i mehaniku pri Ministarstvu za nauku, tehnologije u razvoj ce izloziti zakljucke Komisije kao i pojedinacne stavove o rezultatima na projektima i vrednovanju naucnih publikacija. Pozivaju se se naucni saradnici, ucesnici projekata da prisustvuju ovom sastanku i iznesu svoje misljenje o vrednovanju naucnih rezultata kao i ostalim pitanjima vezanim za naucni rad u okviru projekata MNTR.

Petak, 12. mart 2004. u 14h, sala 2 MI SANU:

Prof. Stevan Pilipovic (Departman za matematiku, PMF Novi Sad)
O FUNKCIJAMA I UOPSTENIM FUNKCIJAMA

Rezime: Posle kratkog istorijskog pregleda pojmova koji se odnose na funkcije i funkcijske prostore, bice izlozene osnovne ideje uopstenih funkcija, hiperfunkcija i daljih uopstenja tih pojmova. Jednostavnim primerima su motivisana izucavanja vrlo sofisticiranih pojmova. Moguci pravac razvoja u analizi ce biti nagovesten.

Petak, 19. mart 2004. u 14h, sala 2 MI SANU:

Prof. Gradimir Milovanovic, Elektronski fakultet, Nis
GAUSSIAN QUADRATURE FORMULAE OF NON-STANDARD TYPE

Abstract. We consider quadrature formulae of the form
(1) \int f\dd\mu=\sum_{k=1}^n w_k \,(\op f)(x_k),
where m is a positive measure supported on a bounded subset of the real line and the linear operator A is an isomorphism on the space of all algebraic polynomials P. Taking
\[(\op f)(x_k):=\frac{1}{2h}\int_{x_k-h_k}^{x_k+h_k} f\,{\dd}x, \]
with suitable hk, we prove the uniqueness of the corresponding Gaussian interval quadrature formula (1) in the case of Jacobi measure on [-1,1]. Several other cases are investigated.

Petak, 26. mart 2004. u 14h (MF):

Prof. Leo Liberti, DEI, Politecnico di Milano, Italy
DETERMINISTIC GLOBAL OPTIMIZATION OF CONTINUOUS NONCONVEX PROGRAMMING PROBLEMS

Abstract: Many real-life problems are nonconvex, and therefore present many complicating features, including multiple local minima. One of the most widely used deterministic techni-ques for the global solution of continuous nonconvex problems is spatial Branch-and-Bound. We describe a spatial B&B algorithm where the convex relaxation at each node is built with the help of symbolic computation, and discuss a family of tightening cuts that significantly accelerate the convergence of the algorithm on a wide class of bilinear nonconvex problems.

OBAVESTENJE