Vedran Novaković : CV

Born in Zagreb, Croatia; 1982.

Titular research associate¹ (znanstveni suradnik) since 2018-05-08.

Education

University of Zagreb, Faculty of Science, Department of Mathematics

• PhD (dr. sc.), thesis defended 2017-12-15 (abstract, summaries, repository (direct link), genealogy)
  Parallel Jacobi-type algorithms for the singular and the generalized singular value decomposition
• dipl. ing.^? (univ. mag. inf. et math.), thesis defended 2006-05-15 (abstract)
  Formal software specification and verification (in Croatian)

Employment

• "researcher" (técnico superior^?) (INTERTWinE, collaborating on OPRECOMP); Universitat Jaume I, Spain (March 2018 to September 2018).
• computational scientist (NLAFET); STFC Daresbury Laboratory (the Hartree Centre), United Kingdom (October 2016 to October 2017).
• computational scientist (SKA-SDP); STFC Daresbury Laboratory (the Scientific Computing Department), United Kingdom (February 2015 to September 2016).
• teaching assistant (asistent^?); University of Zagreb, Croatia (December 2007 to November 2014).
• "software developer" (stručni suradnik); University of Zagreb, Croatia (July 2006 to November 2007).

Engagement

Participated also in the following projects:

• Matrix Factorizations and Block Diagonalization Algorithms (MFBDA) by Croatian Science Foundation (HRZZ-IP-2014-09-3670) from May 2015 to August 2019; PI: Vjeran Hari
• NVIDIA Academic Partnership Program with University of Zagreb, granted in August 2010; PI: Zlatko Drmač, co-PI: Sanja Singer^†
• Numerical methods in geophysical models (037-1193086-2771) by Ministry of Science and Education (MZO) from October 2008 to December 2013; PIs: Mladen Rogina^† and Saša Singer^†

In the scientific & organizing committee of Parallel Numerics workshop (ParNum 2019), Dubrovnik, Croatia, 28–30 October 2019.

Publications

ORCID: 0000-0003-2964-9674

arXiv: novakovic_v_1
Google Scholar: Q2dA4DsAAAAJ
WoS author record: 1369587
ResearcherID: G-2568-2011
Scopus Author ID: 22035270200
Mathematical Reviews Author ID: 936387
zbMATH Author ID: novakovic.vedran
ACM profile: 99659486452
DBLP: 77/8398
CroRIS: 27922
BibTeX: venovako.bib

Preprints

• Vedran Novaković. Arithmetical enhancements of the Kogbetliantz method for the SVD of order two. arXiv:2407.13116 [math.NA] (GitHub: KogAcc and libpvn).

In journals

• Vedran Novaković. Accurate complex Jacobi rotations. J. Comput. Appl. Math. 450 (2024); art. no. 116003 (6 p.). doi:10.1016/j.cam.2024.116003 (preprint, GitHub: AccJac and libpvn).
  short communication
• Vedran Novaković. Vectorization of a thread-parallel Jacobi singular value decomposition method. SIAM J. Sci. Comput. 45 (2023), 3; C73–C100. doi:10.1137/22M1478847 (preprint, GitHub: VecJac).
  Erratum: Sect. 5.3.1, r_U=||U^*U-I||_F² and r_V=||V^*V-I||_F² (the squares have been omitted in the paper).
• Vedran Novaković, Sanja Singer. A Kogbetliantz-type algorithm for the hyperbolic SVD. Numer. Algorithms 90 (2022), 2; 523–561. doi:10.1007/s11075-021-01197-4 (preprint, GitHub: JKogb).
  SharedIt full text (prepend https://): rdcu.be/czAWt
• Vedran Novaković, Sanja Singer. Implicit Hari–Zimmermann algorithm for the generalized SVD on the GPUs. Int. J. High Perform. Comput. Appl. 35 (2021), 2; 170–205. doi:10.1177/1094342020972772 (preprint, GitHub: GPUHZGSVD).
• Vedran Novaković. Batched Computation of the Singular Value Decompositions of Order Two by the AVX-512 Vectorization. Parallel Process. Lett. 30 (2020), 4; art. no. 2050015 (23 p.). doi:10.1142/S0129626420500152 (preprint, GitHub: VecKog).
• Sanja Singer, Edoardo Di Napoli, Vedran Novaković, Gayatri Čaklović. The LAPW Method with Eigendecomposition Based on the Hari–Zimmermann Generalized Hyperbolic SVD. SIAM J. Sci. Comput. 42 (2020), 5; C265–C293. doi:10.1137/19M1277813 (preprint, GitHub: FLAPWxHZ and MPHZ).
• Goran Flegar, Florian Scheidegger, Vedran Novaković, Giovanni Mariani, Andrés E. Tomás, A. Cristiano I. Malossi, Enrique S. Quintana-Ortí. FloatX: A C++ Library for Customized Floating-Point Arithmetic. ACM Trans. Math. Software 45 (2019), 4; art. no. 40 (23 p.). doi:10.1145/3368086
• Vedran Novaković, Sanja Singer, Saša Singer. Blocking and parallelization of the Hari–Zimmermann variant of the Falk–Langemeyer algorithm for the generalized SVD. Parallel Comput. 49 (2015) 136–152. doi:10.1016/j.parco.2015.06.004 (accepted manuscript, GitHub: hz-gsvd).
• Peter Benner, Vedran Novaković, Antonio Plaza, Enrique S. Quintana-Ortí, Alfredo Remón. Fast and Reliable Noise Estimation for Hyperspectral Subspace Identification. IEEE Geosci. Remote Sens. Lett. (GRSL) 12 (2015), 6; 1199–1203. doi:10.1109/LGRS.2014.2388133
• Vedran Novaković. A hierarchically blocked Jacobi SVD algorithm for single and multiple graphics processing units. SIAM J. Sci. Comput. 37 (2015), 1; C1–C30. doi:10.1137/140952429 (preprint, GitHub: GPUJACHx).
• Sanja Singer, Saša Singer, Vedran Novaković, Davor Davidović, Krešimir Bokulić, Aleksandar Ušćumlić. Three-Level Parallel J-Jacobi Algorithms for Hermitian Matrices. Appl. Math. Comput. 218 (2012), 9; 5704–5725. doi:10.1016/j.amc.2011.11.067 (preprint).
• Sanja Singer, Saša Singer, Vedran Novaković, Aleksandar Ušćumlić, Vedran Dunjko. Novel Modifications of Parallel Jacobi Algorithms. Numer. Algorithms 59 (2012), 1; 1–27. doi:10.1007/s11075-011-9473-6 (preprint).
• Vedran Novaković, Sanja Singer. A GPU-based hyperbolic SVD algorithm. BIT 51 (2011), 4; 1009–1030. doi:10.1007/s10543-011-0333-5 (preprint).
• Vedran Novaković, Sanja Singer, Saša Singer. Estimates for the Spectral Condition Number of Cardinal B-Spline Collocation Matrices. Math. Commun. 15 (2010), 2; 503–519. (www)

In proceedings

• ISC High Performance 2018 International Workshop on "Approximate and Transprecision Computing on Emerging Technologies" (ATCET 2018); 28 June 2018, Frankfurt, Germany.
  Hartwig Anzt, Goran Flegar, Vedran Novaković, Enrique S. Quintana-Ortí, Andrés E. Tomás. Residual Replacement in Mixed-Precision Iterative Refinement for Sparse Linear Systems.
  Lecture Notes in Computer Science 11203 (2018), High Performance Computing; 554–561.
  doi:10.1007/978-3-030-02465-9_39
• ISC High Performance 2016 International Workshop on OpenPOWER for HPC (IWOPH'16); 23 June 2016, Frankfurt, Germany.
  Mike Ashworth, Jianping Meng, Vedran Novaković, Sergi Siso. Early Application Performance at the Hartree Centre with the OpenPOWER Architecture.
  Lecture Notes in Computer Science 9945 (2016), High Performance Computing; 173–187.
  doi:10.1007/978-3-319-46079-6_13
• 9^th International Conference on Telecommunications (ConTEL 2007); 13–15 June 2007, Zagreb, Croatia.
  Matko Botinčan, Vedran Novaković. Model-based testing of the Conference Protocol with Spec Explorer.
  doi:10.1109/CONTEL.2007.381861 (131–138)

In popular science venues

• Marko Doko, Vedran Novaković. Izračunljivost i apstraktni strojevi. Math.e 9 (2006); 30–48. (www)

Conferences

• 10^th International Workshop on Parallel Matrix Algorithms and Applications (PMAA18); 27–29 June 2018, Zürich, Switzerland.
  Vedran Novaković. A GPU variant of the implicit Hari-Zimmermann algorithm for the generalized SVD. (talk)
  Sanja Singer, Edoardo di Napoli, Vedran Novaković, Gayatri Čaklović. Parallel solution of the generalized eigenvalue problem given in a factored form.
• 8^th International Workshop on Parallel Matrix Algorithms and Applications (PMAA14); 2–4 July 2014, Lugano, Ticino (Switzerland).
  Vedran Novaković. A blocked Jacobi SVD algorithm for single and multiple GPU(s). (talk)
  Sanja Singer, Vedran Novaković, Saša Singer. Parallelization of the Falk–Langemeyer method.
• 10^th International Workshop on Accurate Solution of Eigenvalue Problems (IWASEP 10); 2–5 June 2014, Dubrovnik, Croatia.
  Vedran Novaković. A hierarchically blocked Jacobi SVD algorithm for single and multiple GPU(s). (poster)
• Workshop on Numerical Linear Algebra and Optimization (www); 8–10 August 2013, Vancouver, British Columbia (Canada).
  Vedran Novaković. A Parallel Jacobi-type SVD Algorithm for the GPU Clusters. (poster)
• 9^th International Workshop on Accurate Solution of Eigenvalue Problems (IWASEP 9); 4–7 June 2012, Napa Valley, California (USA).
  Sanja Singer, Vedran Novaković. Kogbetliantz-like Method for the Hyperbolic SVD.
  Vedran Novaković, Sanja Singer. Towards Multi-GPU Jacobi (H)SVD.
• International Conference on Scientific Computing (SC2011); 10–14 October 2011, S. Margherita di Pula, Sardinia (Italy).
  Sanja Singer, Vedran Novaković. Kogbetliantz-like method for hyperbolic SVD.
• 7^th Conference on Applied Mathematics and Scientific Computing (ApplMath11); 13–17 June 2011, Trogir, Croatia.
  Vedran Novaković, Saša Stanko, Sanja Singer. Parallel Hybrid CPU-GPU Full-Block Jacobi Algorithm for Hyperbolic SVD.
• 8^th International Workshop on Accurate Solution of Eigenvalue Problems (IWASEP 8); 28 June – 01 July 2010, Berlin, Germany.
  Vedran Novaković, Sanja Singer. A GPU-based hyperbolic SVD algorithm. (poster)
• BIT50 Trends in Numerical Computing; 17–20 June 2010, Lund, Sweden.
  Vedran Novaković, Sanja Singer. A GPU-based hyperbolic SVD algorithm.
• SIAM Conference on Applied Linear Algebra (LA09); 26–29 October 2009, Seaside, California (USA).
  Sanja Singer, Saša Singer, Vedran Novaković. Orthosymmetric Block Reflectors and QR Factorizations.
• 6^th Conference on Applied Mathematics and Scientific Computing (ApplMath09); 14–18 September 2009, Zadar, Croatia.
  Vedran Novaković, Sanja Singer, Saša Singer. Cardinal Splines, Töplitz Matrices and Circulants.
  Vjeran Hari, Saša Singer, Vedran Novaković, Sanja Singer. Parallelization and pivoting in one-sided Jacobi algorithms.
• 23^rd Biennial Conference on Numerical Analysis (naconf); 23–26 June 2009, Glasgow, Scotland (UK).
  Vjeran Hari, Saša Singer, Vedran Novaković, Sanja Singer. New Advances in One-Sided Jacobi Algorithms.

Teaching

Assisted with teaching of the following courses (in Croatian) at University of Zagreb:

Faculty of Mechanical Engineering and Naval Architecture

• Mathematics I
  fall semester; 2007/08.
• Mathematics II
  spring semester; 2007/08.
• Mathematics 3
  fall semester; 2008/09, 2009/10, 2010/11, 2011/12, 2012/13.
• Mathematics III A
  fall semester; 2008/09, 2009/10, 2010/11, 2011/12, 2012/13.
• Mathematics III B
  fall semester; 2008/09, 2009/10, 2010/11, 2011/12, 2012/13.
• Mathematics IV
  spring semester; 2008/09, 2009/10, 2010/11.
• Mathematics IV A
  spring semester; 2008/09, 2009/10, 2010/11.

Faculty of Science, Department of Mathematics

• Introduction to parallel computing²
  fall semester; 2008/09, 2009/10, 2010/11, 2011/12, 2012/13, 2014/15.
• Applications of parallel computers²
  spring semester; 2008/09, 2009/10, 2010/11, 2011/12, 2012/13.
• Computer programming 1 (C, www)
  fall semester; 2008/09, 2009/10.
• Computer programming 2 (C, www)
  spring semester; 2008/09, 2009/10.
• Parallel Algorithms 1 & 2³
  fall & spring semester; 2007/08.
• Computer Science Lab 4 (C++)³
  fall semester; 2006/07.