Prijeđite na glavni sadržaj

Aditivni modeli - Novi algoritmi učenja za vrlo velike skupove podataka

Vrijeme
11.1.2019. 10:30
Lokacija
IRB, dvorana III.

Abstract

The seminar has presentation at two levels but before entering the main topics it touches some hot events around the recently very popular area of AI first. Next, it presents some commonalities shared by the various well known machine learning tools today which belong to the additive models. This will lead us to discussing similarities and differences between NN and SVMs. However, this can’t be done without pointing at the different norms used in designing the two models. Why composite norms of the two parts (loss and regularizer) are needed will be explained on a very simple example.

The second seminar’s part introduces the novel and powerful On Line Learning Algorithm (OLLA) for SVMs (and seven more machine learning models) aimed at really huge datasets. Theoretical derivation of the models will be shown together with the experimental results obtained on 23 different datsets.

Short CV

Vojislav Kecman is the tenured full professor with a Computer Science Department at the Virginia Commonwealth University (VCU) in Richmond, VA, USA where he directs the Learning Algorithms and Applications Laboratory (LAAL).

He was Fulbright Scholar at the M.I.T., Cambridge, MA, USA, Konrad Zuse Professor at FH Heilbronn, DFG Visiting Scientist at TH Darmstadt, Research Fellow at the Drexel University, Philadelphia, PA, DAAD Professor at the FHTW in Berlin and Soest as well as at the Stuttgart University. He also was the full faculty at The University of Auckland and Zagreb University.

Prof. Kecman authored several books in the areas of machine learning and in the fields of system dynamics modeling and control; notably ‘Learning and Soft Computing, Support Vector Machines, Neural Networks, and Fuzzy Logic Models’ published by The MIT Press, Cambridge, MA (2,210 citations) and ‘Kernel Based Algorithms for Mining Huge Data Sets, Supervised, Semi-supervised, and Unsupervised Learning’, published by Springer-Verlag, Berlin, Heidelberg (222 citations). The former book has been/is a recommended university textbook at about 60 universities all around the world and at about 20 ones in USA. He also co-authored a book on Optimal Control Systems published by CRC, Francis & Taylor, Pub. Comp. in 2009.

Prof. Kecman’s current research interests include: learning from experimental data (machine learning, data mining) by support vector machines and neural networks, as well as modeling human knowledge by fuzzy logic systems. Theory, practice, philosophy and versatility of these soft computing tools is used in broad fields of different (nonlinear) regression and pattern recognition (classification, decision making) tasks in – e-commerce, bioinformatics, vision systems, computer graphics, data and signal processing, numerical mathematics, credit assignment problems, (financial) time series analysis, image compression, expert and decision systems and computer intelligence.

Ova stranica koristi kolačiće. Neki od tih kolačića nužni su za ispravno funkcioniranje stranice, dok se drugi koriste za praćenje korištenja stranice radi poboljšanja korisničkog iskustva. Za više informacija pogledajte naše uvjete korištenja.

Prilagodi postavke
  • Kolačići koji su nužni za ispravno funkcioniranje stranice. Moguće ih je onemogućiti u postavkama preglednika.