Upis na master i doktorske studije, školska 2011/2012 godina

detaljnije

Predavanje prof. Zorana Obradovića na Matematičkom fakultetu, 2. septembar 2011.

Prof. Zoran Obradović sa Univerziteta Temple, Filadelfija, održaće predavanje na Matematičkom fakultetu 2. septembra 2011. u 12 sati, sala 718/IV.

Naslov predavanja je "Learning from Inconsistent and Unreliable Annotators by a Gaussian Mixture Model and Bayesian Information Criterion".

Pozivamo sve zainteresovane kolege i studente da prisustvuju ovom predavanju. Sadržaj predavanja i podaci o autoru mogu se preuzeti OVDE.

detaljnije

Seminar Applications of NVIDIA Tesla Hardware Platform and CUDA Parallel Computing Environment in Complex Computations

Drage kolege,

Pozivamo Vas da prisustvujete seminaru "Applications of NVIDIA Tesla Hardware Platform and CUDA Parallel Computing Environment in Complex Computations", koji će se održati u utorak 6. septembra, sa početkom u 16 sati, u sali 706 Matematičkog fakulteta Univerziteta u Beogradu.

CUDA je paralelizovana računarska platforma kompanije NVIDIA koja omogućava izuzetno velika ubrzanja korišćenjem GPJ (grafičke procesorske jedinice) za složena izračunavanja. Dođite da saznate najnovije vesti iz oblasti izračunavanja pomoću GPJ i najuspešnijih hibridnih rešenja.

Tokom seminara imaćete priliku da razgovarate sa stručnjacima iz Rusije i Srbije i dobijete odgovore na vaša pitanja. Od stručnjaka iz akademskih i industrijskih sredina ćete čuti kako su uz pomoć paralelne arhitekture CUDA iskoristili NVIDIA GPJ za pravljenje računski zahtevnih aplikacija u različitim naučnim oblastima.

Agenda:
1) Pozdravni govor
2) GPJ računarstvo sa platformom NVIDIA Tesla
3) Uvod u razvojno okruženje CUDA (koncept programiranja)
4) GPJ računarstvo i studije slučajeva iz različitih naučnih oblasti
5) GPJ računarstvo i CUDA u obrazovanju u Srbiji

Izlaganja će biti na engleskom jeziku.

Biće nam drago da učestvujete na inovativnom seminaru koji se organizuje u okviru saradnje NVIDIA, Matematičkog fakulteta i Tehnološkog partnerstva.

Detaljnije informacije će biti blagovremeno raspoložive na veb lokaciji:
http://poincare.matf.bg.ac.rs/~smalkov/files/razno/cuda.day.2011.09.06

detaljnije

Uspeh studenata Matematičkog fakulteta na medjunarodnom takmičenju 18th IMC 2011 u Blagoevgradu

Na međunarodnom studentskom takmičenju iz matematike 18th IMC 2011, u Blagoevgradu, Bugarska, odražanom od 28. jula do 3. avgusta (http://www.imc-math.org/) studenti Matematičkog fakulteta u Beogradu postigli su vrlo lep uspeh:

  Marija Jelić (srebro)
  Marijana Smailagić (srebro)
  Petar Melentijević (pohvala)
  Nikola Mutavdžić (pohvala)

Čestitamo našim studentima, kao i kolegama koji su predvodili našu ekipu i učestvovali u pripremama za takmičenje.

detaljnije

Intenzivni kurs 'Optimizacija' u okviru tekućeg Tempus projekta, 29.8-23.9. 2011.

U okviru tekućeg Tempus projekta "SEE Doctoral Studies in Mathematical Sciences" na Matematičkom fakultetu u Beogradu biće održan intenzivni kurs "Optimizacija", u periodu 29.8-23.9, koji je namenjen studentima doktorskih studija. Kurs će pohađati 27 studenata sa fakulteta-partnera zapadnog Balkana i EU. Predavači su profesori sa Univerziteta u Gracu i Bugarske Akademije nauka.

Više detalja o kursu može se naći na strani

http://www.matf.bg.ac.rs/tempus/

detaljnije

Seminar mehanike, 20. jul 2011.

Naredni sastanak seminara mehanike održaće se u sredu, 20 jula, u 16h u sali 307F Matematičkog instituta SANU.

Predavači: Prof. Peter Bradshaw (Professor of Experimental Aerodynamics, Department of Aeronautics, Imperial College, London University, England) i dr Srba Jović (NASA, Kalifornija)

Naslov predavanja: A REVIEW OF TURBULENT FLOW

Sadržaj:

The enormous range of length scales makes it impossible to solve the exact Navier-Stokes* *(N-S) equations directly ("Direct Numerical Simulation" or DNS) for high-Reynolds-number turbulent flows.

The most gentle simplification is to solve N-S for large scales and to approximate ("model") the small eddies (Large Eddy Simulation, LES). Near a solid surface, unfortunately, there are NO large eddies! The approximate small-eddy model has to be used for the full range of length scales. This is the big obstacle to using LES in "wall-bounded" flows.

Today's engineers use Reynolds-Averaged N-S (RANS) equations. Usually, the complete range of eddy sizes is represented by just one length scale (plus the viscous length scale needed near solid surfaces). Nearly all the terms in the equations are modeled, i.e. replaced by dimensionally-correct and physically plausible combinations of the length and velocity scales and their spatial gradients. Each modeled term contains an empirical parameter, usually a constant.

The most obvious way to produce a RANS model is to derive Reynolds-averaged transport equations for all the mean-velocity components and all the Reynolds stresses, and model the unknown terms. This has been done quite successfully, but stress-transport models have the reputation of suffering numerical problems. This reputation dates from 20-30 years ago at least.

The ratio of a Reynolds stress to the rate of strain in the plane of that stress defines a so-called "eddy viscosity". It is different for different stresses, and varies greatly in space within the flow domain. It is a physically-meaningful quantity only if the scales of the turbulence are closely related to the scales of the mean flow. Most current engineering prediction methods are based on a scalar eddy viscosity (same for all stresses, a.k.a. isotropic") defined as   ( **) The dimensionless empirical parameter is (almost) always chosen to reproduce the logarithmic "law of the wall".

The review will end with a discussion of the capabilities of today's prediction methods.

detaljnije

Seminar mehanike, 4. jul 2011.

Naredni sastanak Seminara mehanike održaće se u ponedeljak, 4. jula, u 13h na Matematičkom institutu SANU, sala 307 F.

Predavač: Prof. Subhash C. Sinha, Alumni Professor and Director of Nonlinear Systems Research Laboratory, Department of Mechanical Engineering, Auburn University, AL 36849, USA

Naslov predavanja: ANALYSIS AND CONTROL OF PARAMETRICALLY EXCITED SYSTEMS: A NEW APPROACH

Abstrakt: A general framework for the analysis and control of parametrically excited linear/nonlinear dynamical systems is presented. This class of problems appears in the modeling of rotorcraft blades in forward flight, asymmetric rotor-bearing systems, automotive components such as connecting rods, universal joints, asymmetric satellites, fluids under gravity modulations, etc. These dynamical systems are represented by a set of differential equations which contain time-periodic coefficients. First, an efficient computational scheme for the solution of linear problem is discussed through an application of Chebyshev polynomials. The idea is further developed to obtain the Lyapunov-Floquet (L-F) transformation associated with a linearized or a quasilinear time-periodic dynamical system. An application of L-F transformation yields equivalent systems whose linear parts are time-invariant. Therefore, the controls for all time-periodic linear systems can be designed using the standard time-invariant methods such as pole placement or optimal control theory. A symbolic control technique for Floquet multiplier placement is also suggested for linear time-periodic systems. In the case of nonlinear systems, a periodic orbit in the original coordinates has a fixed point representation after the L-F transformation. The local stability and bifurcation analyses are studied via time-dependent 'center manifold reduction' and 'normal form theory'. Results for fold, flip and secondary Hopf bifurcations are discussed. Bifurcation control and feedback linearization techniques for nonlinear time-periodic systems are also developed and applied to some typical problems. Further, the order reduction problem associated with free and forced parametrically excited large-scale nonlinear systems is also addressed using an invariant manifold approach. A methodology for reduced order controller design is also suggested. The practical significance of these approaches is demonstrated through simulations and experimental investigations of a number of mechanical systems. These include vibration control of a multi-bladed rotor, shafts supported by magnetic bearings, bifurcation analyses of a flexible slider crank mechanism and an autoparametric mechanical system, among others.

detaljnije

Seminar Geometrija, vizuelizacija i obrazovanje sa primenama, 30. jun 2011.

Sastanak Seminara Geometrija, vizuelizacija i obrazovanje sa primenama održaće se u četvrtak, 30. juna 2011, od 17:15 u sali 301f Matematičkog instituta, Knez Mihajlova 36.

Predavač: Ivan Dimitrijević

Naslov predavanja: Ajnštajnova jednačina za gravitaciono polje i njena f(R) modifikacija, 2. deo

detaljnije

CGTA seminar, 28. jun 2011.

Naredni sastanak CGTA seminara održaće se u utorak, 28. juna u 18h , u projektor sali na Matematičkom fakultetu

Kreda Lutalica: John Milnor kao dobitnik Abelove nagrade za 2011 godinu.

detaljnije

Seminar za primenjenu matematiku, 28. jun 2011.

Naredni sastanak Seminara za primenjenu matematiku održaće se u utorak, 28. juna 2011. u 14:15 h na Matematičkom institutu SANU, sala 301f, 3. sprat.

Predavači: Zoran Obradović, Director, Center for Data Analytics and Biomedical Informatics, Temple University, Philadelphia, USA

Naslov predavanja: SPATIO-TEMPORAL ANALYSIS OF MULTI-SOURCE DATA IN REMOTE SENSING

ABSTRACT: In this talk we will first present an overview of the results obtained over the last five years on our data mining based study aimed at understanding spatio-temporal distribution of aerosol by taking advantage of collected measurements from multiple ground and satellite-based sensors. In more details we will describe our recent method for information fusion and learning from data of varying quality based on Continuous Conditional Random Fields. Finally, we will discuss our effective method for learning from spatio-temporal data with a large fraction of missing values. This method that will be published at IJCAI 2011 outperformed alternatives when imputing 80% of nonrandom missing values in aerosol observations. Using such enlarged data we have developed a predictor of aerosol optical depth that was much more accurate than predictors based on alternative imputation methods when tested rigorously over entire continental US in year 2005.

SPEAKER INFORMATION:

Zoran Obradović, professor of Computer and Information Sciences and the director of the Center for Data Analytics and Biomedical Informatics at Temple University in Philadelphia is an internationally recognized leader in data mining and bioinformatics. He has published about 230 articles addressing data mining challenges in health informatics, climate and ecological management, the social sciences, and other domains. Obradovic was the program chair at six, track chair at seven and program committee member at about 40 international conferences on data mining. He currently serves as an editorial board member at seven journals and is the executive editor at the journal on Statistical Analysis and Data Mining which is the official publication of the American Statistical Association (ASA).

detaljnije

Vesti i dešavanja

Aktivnosti na seminarima

sve vesti