Seminar on Computer Science and Applied Mathematics

PROGRAM

UTORAK, 11.10.2016. u 14:15, Sala 301f, MI SANU, Kneza Mihaila 36
Dr Dušan Savić, Fakultet organizacionih nauka
RAZVOJ SOFTVERA ZASNOVAN NA MODELU SLUČAJEVA KORIŠĆENJA I MDD PRISTUPU
U predavanju će biti predstavljen Silab-MDD pristup koji povezuje Larmanuovu metodu razvoja softvera sa MDD pristupom.Predlaže se nova metoda za preciznu i detaljnu specifikaciju slučajeva korišćenja. Specifikacija slučajeva korišćenja je podržana sopstvenim domenski-specifičnim jezikom UCDSL koji je razvijen u okviru JetBrains MPS alata. UCDSL se može koristiti samostalno ili kao dodatak za druga razvojna okruženja. Pristup obuhvata i sopstvene domeski specifične jezike pomoću kojih se mogu opisati dijagrami tokova podataka i rečnik podataka. Izvršena je evaluacija predloženog pristupa kako bi se utvrdila mogućnost primene razvijenih metoda i jezika.


UTORAK, 18.10.2016. u 14:15, Sala 301f, MI SANU, Kneza Mihaila 36
Marko Petkovic, Prirodno-matematicki fakultet u Nišu
ITERATIVE METHODS FOR COMPUTING GENERALIZED INVERSES
We consider the class of iterative methods for computing generalized inverses $A^{(2)}_{\mathcal{R}(G),\mathcal{N}(G)}$, given by $X_{k+1}=X_k p(AX_k)$ where $p$ is polynomial. These methods generalize well-known hyper-power methods of order $r$, obtained by taking $p(x)=1+x+\ldots+x^{r-1}$. We examine the convergence properties and convergence order of these methods, depending on the polynomial $p(x)$. Moreover, we give the methods with highest possible convergence order, taking a certain number of matrix multiplications per iteration. These results can be efficiently used to improve several methods for finding 1-norm minimal solution of the linear system and have various applications, including image deblurring and compressed sensing.


UTORAK, 25.10.2016. u 14:15, Sala 301f, MI SANU, Kneza Mihaila 36
Bogdana Stanojević, Matematički institut SANU
PARAMETRIC COMPUTATION OF A FUZZY SET SOLUTION TO A CLASS OF FUZZY LINEAR FRACTIONAL OPTIMIZATION PROBLEMS
The class of fuzzy linear fractional optimization problems with fuzzy coefficients in the objective function is considered. We propose a parametric method for computing the membership values of the extreme points in the fuzzy set solution to such problems. We use a parametric analysis of the efficiency of the feasible basic solutions to a certain bi-objective linear fractional programming problem through the optimality test in a related linear programming problem.