CURRICULUM VITAE

Personal Details

Name: Illia Horenko

Born: August 23, 1977 in Kiev, Ukraine

Family status: Married to Julia Christiane Susanne Gerber;Two children: Nicolai (* Dec.10, 2012)and Elina (* Nov. 19, 2014)

Office Adress: Institute of Computational ScienceUniversita della Svizzera italianaVia Guiseppe Buffi 136904 Lugano, Switzerland

Email: illia.horenko@usi.ch

Education & Academic Degrees

2004 Dr. rer. nat in Mathematics from the Department of Mathematicsand Computer Science, Freie Universitat Berlin, Germany (supervisedby Prof. Dr. Christoph Schutte)

1999 M. Sc. in Applied Mathematics from the Department of ComputerScience, Taras Shevchenko National University of Kyiv, Ukraine

1998 B. Sc. in Applied Mathematics from the Department of ComputerScience, Taras Shevchenko National University of Kyiv, Ukraine

1998 B. Sc. in Chemical Physics from the Department of Chemistry, TarasShevchenko National University of Kyiv, Ukraine

Academic Career

2015 Promotion to Full Professor at the Faculty of Informatics and theInstitute of Computational Science initiated, Universita della SvizzeraItaliana, Switzerland (promotion process will be finished in 2016)

2010 Associate Professor at the Faculty of Informatics and the Institute ofComputational Science, Universita della Svizzera Italiana, Switzerland

2008-2010 Assistant Professor (W1 Juniorprofessor) at the Institute of Mathe-matics, Freie Universitat Berlin

2004-2008 Post-doctoral Research Fellow at the Institute of Mathematics,Freie Universitat Berlin (with Prof. Ch. Schutte)

2001-2004 Research Assistant at the Institute of Mathematics, Freie Univer-sitat Berlin, research group of Prof. Ch. Schutte

1999 - 2001 Research Assistant at the Konrad-Zuse-Zentrum for informationtechnology and supercomputing in Berlin (ZIB), Germany, researchgroup of Prof. P. Deuflhard

Distinctions and Fellowships (Since joining USI in 2010)

since 2014 DFG Mercator Professor-Fellowship at the Institute of Mathema-tics, Freie Universitat Berlin. Involved with 16% of working time in theDFG Collaborative Research Center (CRC) Scaling Cascades in Com-plex Systems. The official letter from DFG CRC was stating: ”Overthe past several years, Prof. Horenko has developed non-parametric,non-stationary, non-homogeneous data analysis techniques which, inour view, belong to the most advanced methodologies in this field.”(http://www.sfb1114.de/research/mercator-fellow)

2013 Offer of an IPAM Fellowship at the UCLA, California, ProgramMaterials for a Sustainable Energy Future

2004-2008 IPAM Fellowship at the UCLA California, Program Model and DataHierarchies for Simulating and Understanding Climate

Acquired competitive funding since joining USI in 2010 (1’711’906 EUR in five years)

2015-2018 DFG - CRC 1114Scaling Cascades in Complex SystemsProject Website: http://www.sfb1114.de/research/

mercator-fellow

Obtained funding: Mercator Professor-Fellowship of the DFG(60’000 EUR). Besides of this funding from DFG, I. Horenko wasprovided with a funding for one postdoc position for two years (112’000EUR) from the FU Berlin.

2015-2018 SNF/DFG - Project FOR 1898MS-GWaves: Multi-Scale Dynamics of Gravity WavesProject Website: https://ms-gwaves.iau.uni-frankfurt.de/index.php/de/

Obtained funding: 289’650 CHF (270’700 EUR)

2015- 2016 PASC (Platform for Advanced Scientific Computing)Mathematical modeling of credit and equityProject Website: http://www.pasc-ch.org/projects/projects/

multiscale-economical-data

Obtained funding: 319’696 CHF (300’000 EUR)

2014-2016 SNF - Project 152979AnaGraM: Adaptive numerical methods for time series analysis of time-dependent dynamical Graphs in the presence of Missing dataProject Website: http://p3.snf.ch/Project-152979Obtained funding: 120’960 CHF (113’000 EUR)

2012- 2015 SNF - Project HPC-Risk: Towards the HPC-inference of causalitynetworks from multiscale dataProject Website: http://www.pasc-ch.org/projects/projects/

multiscale-economical-data

Obtained funding: 221’650 CHF ( 207’000 EUR)

2010-2015 DFG Schwerpunktprogramm 1276 - MetstroemDiscrete-continuous hybrid models based on integral conservation lawsProject Website: http://metstroem.mi.fu-berlin.de/?page_id=11Obtained funding: 540’000 CHF (505’000 EUR)

2010-2013 SNF - Project 131845AnaGraph: Adaptive numerical methods for nonstationary time seriesanalysis of time-dependent graphs in context of dynamical systemsProject Website: http://p3.snf.ch/Project-131845Obtained funding: 155’744 CHF (145’000 EUR)

Research group

Post doc Dr. Lukas Pospisil is working on implementation and adaptation ofthe methods developed in the group for the High-Performance Compu-ting platforms of the Swiss Supercomputing Center in Lugano (CSCS)and their application to problems from economics, funded from thePASC-project Multi-Scale Economical Data.

Post doc Dr. Olga Kaiser is working on mathematical methods for non-stationary analysis and data-based prediction of rare events, funded fromSNF-Project Multi-Scale Dynamics of Gravity Waves.

PhD Dipl.-Inf. Dimitri Igdalov is working on data-based identification oftemporal changes in very large graphs, missing data problems and HPCimplementation of time series analysis methods library developed in thegroup, currently funded by USI Lugano.

PhD M.Sc. Anna Marchenko is working on mathematical modeling ofequity risks and non-parametric time series analysis methods based onmaximum entropy principles, funded from the SNF-project HPC-Risk.

Supervision Activities - since joining USI as an associate professor in 2010

PhilipMetzner

Postdoc between 2011 and 2014. - Philip was working on inverse pro-blems in fluid mechanics and their HPC-implementation, funded fromthe German DFG research initiative MetStroem.

KatiaFiorucci

PhD student between 2012 and 2014. - Katia was working on ma-thematical modeling of credit and equity risk beyond homogeneity andstationarity assumptions: statistical factor models and high-performancedata mining.

Thomasvon Larcher

Postdoctoral Research Fellow whom I am co-supervising since 2011.He works predominantly at the Freie Universitat Berlin in the Geophysi-cal Fluid Dynamics group of Prof. Rupert Klein on the project

”Discrete-

continous hybrid models on the basis of the integral conservation laws“,a joint work by FU Berlin (R. Klein, M. Waidmann, T. von Larcher),USI Lugano (I. Horenko, P. Metzner, D. Igdalov), University of Stuttgart(A. Beck, C.-D. Munz), and University of Cologne (G. Gassner).

OlgaKaiser

PhD student between 2010 and 2015. Thesis title: Data-based analysisof extreme events: inference, numerics and applications.

LarsPutzig

PhD student between 2010 and 2014. Thesis title: Non-stationarydata-driven computational portfolio theory and algorithms.

Janade Wiljes

PhD student between 2011 and 2014 ( Freie Universitat Berlin, Depart-ment of Mathematics). Thesis title: Data-driven discrete spatio-temporalmodels: problems, methods and an arctic sea ice application. Main advi-sor: I. Horenko (co-advisor: R. Klein). For her thesis Jana was awardedwith a Tiburtius-Promotionspreis von Berliner Hochschulen inyear 2015 (first price, 4’000 EUR).

ChristianBlume

PhD student between 2009 and 2012 at Freie Universitat Berlin. Thesistitle: Statistical learning to model stratospheric variability. Co-advisor:I. Horenko (main advisor: Prof. Dr. Katja Matthes).

LucaVignola

Master student. Defended his Master thesis in 2013. Thesis title: Un-supervised topological mining of EEG Motor Movement/Imagery Data-base beyond usual assumptions.

ChristineKaiser

Visiting Bachelor student from Freie Universitat Berlin, Departmentof Mathematics. Did her semester abroad in my research group at USIin summer 2010.

LawrenceFarinola

Master student. Defended his Master thesis in 2014. Thesis title: In-vestigating the Biomechanics of the human lumbar spine using finiteelement methods.

AlenaKopanicakova

Master student. Defended her Master thesis in in 2015. Thesis title:Investigating Optimization Strategies for Quadratic Programming Com-ponents of a Data Analysis Framework.

Janade Wiljes

Diploma student. Defended her Diploma thesis in 2010 at Freie Uni-versitat Berlin, Department of Mathematics. Thesis title: Adopting aBayesian framework to multidimensional cluster modeling.

Most Recent Invited and Keynote Presentations (2011-2015)

2016 Invited speaker at the Oberwolfach Workshop”Multiscale Interactions

in Geophysical Fluids“ , Oberwolfach Institute, (Germany)

2016 Invited speaker at the Oberwolfach Workshop”Mathematical and Algo-

rithmic Aspects of Data Assimilation in the Geosciences“ , OberwolfachInstitute, (Germany)

2015 Invited speaker at the”DMV 2015“ ,Hamburg (Germany)

2015 Invited speaker at the Workshop”Causality in Turbulence“ at ETHZ

(organizer P. Koumoutsakos), Zurich (Switzerland)

2015 Invited speaker at the”MATHICSE Seminar“ at EPFL (organizer A.

Abdulle), Lausanne (Switzerland)

2015 Invited speaker at the”SFB1114 Seminar“ at the FU Berlin (organizer

R. Klein), Berlin (Germany)

2014 Invited speaker at the DBTA Workshop”Big Data: Stream Processing“,

Bern (Switzerland)

2014 Invited speaker at the CECAM conference”Long time dynamics from

short time simulations“ (organizer M. Parinello), Lugano (Switzerland)

2014 Invited lecture at the”CAOS Colloquium“ of the Courant Institute of

Mathematical sciences, (three seminars, organizer A. Majda), New YorkUniversity (USA)

2013 Keynote speaker at the 2013 NDNS+ Workshop”Stochastic Modeling

of Multiscale Systems“ (organizer J. Frank), Eindhoven (Netherlands)

2013 Invited speaker at the”Seminar of the Informatics Faculty“ at the Kon-

stanz University (invited by D. Saupe) , Konstanz (Germany)

2012 Invited speaker at the 2012 Oberwolfach Workshop”Mathematical and

Algorithmic Aspects of Atmosphere-Ocean Data Assimilation“ (orga-nizers A. Griewank, S. Reich, I. Roulstone, A. Stuart), Oberwolfach(Germany)

2012 Key-note presentation at the Swiss HPC Service Provider Communi-ty meeting

”Handling huge amount of data for HPC“, CSCS, Lugano

(Switzerland)

2012 Key-note lecturer (3 lectures) at the summer school of DFG SPP Met-Stroem in Berlin

”Methods of Data Analysis for Fluid Mechanics, Me-

teorology and Climate Research“, Berlin (Germany)

2012 Invited speaker at the”Seminar of Oeschger Center“ at the Bern Uni-

versity, (invited by O. Romppainen-Martius), Bern (Switzerland)

2012 Invited speaker at the CECAM conference”Machine Learning in Ato-

mistic Simulations“ (organizer M. Parinello), Lugano (Switzerland)

2011 Invited speaker at the seminar of the Meteorology Department of theLudwig Maximilian University, (invited by G. Craig), Munich (Germa-ny)

2011 Speaker at the”International Conference on Advances in Computational

Science“, (invited by R. Krause and M. Parinello), Lugano (Switzerland)

2011 Invited speaker at the”Seminar of the Meteorology Department“ at

Freie Universitat Berlin (invited by K. Matthes), (Germany)

2011 Invited lecturer (2 lectures) at the”International Workshop on Stati-

stical Inverse Modeling of Complex Nonlinear Systems“, (organizer A.Majda, Courant, NYU), Fudan University, Shanghai (China)

Organized workshops and conferences (since joining USI in 2010)

2016 Oberwolfach Seminar Different mathematical perspectives on descrip-tion of unresolved scales in multiscale systems (together with R. Klein,C. Hartmann and T. O’Kane), Oberwolfach (Germany)

2012 DFG Summer School Methods of Data Analysis for Fluid Mechanics,Meteorology and Climate Research, (co-organizer with Sebastian Reich),Berlin (Germany)

2012 Workshop Non-stationary Methods of Data Analysis and Applications,Lugano (Switzerland)

2011 Conference Advances in Computational Science,(co-organizer withRolf Krause and Michele Parinello), Lugano (Switzerland)

2011 Swiss Numerics Colloquium (co-organizer with Rolf Krause), Luga-no (Switzerland)

2010 Workshop Data Hierarchies for Climate Modeling at the Institute ofPure and Applied Mathematics (IPAM) in Los Angeles (co-organizertogether with Amy Braverman, Luis Kornblueh and Robert Pincus, LosAngeles (USA)

List of publicationsPublications with me being either the main author or the corresponding author are highlightedwith my name written in bold letters.

Submitted Journal Articles (currently under review)

[1] I. Horenko, S. Gerber, T. O’Kane, J. Risbey, and D. Monselesan. On inference and valida-tion of causality relations in climate teleconnections. under review in Nonlinear Processes inGeophysics, 2015.

[2] O. Kaiser and I. Horenko. Data-driven spatio-temporal modeling of threshold excesses withunresolved covariates. under review in Water Resources Research, 2015.

[3] S. Gerber and I. Horenko. On computation of relations between categorical data. underreview in Science Advances, 2015.

Published peer-reviewed Journal Articles

[4] Thomas von Larcher, Andrea Beck, Rupert Klein, Illia Horenko, Philipp Metzner, MatthiasWaidmann, Dimitri Igdalov, Gregor Gassner, and Claus-Dieter Munz. Towards a frameworkfor the stochastic modelling of subgrid scale fluxes for large eddy simulation. MeteorologischeZeitschrift, 24(3):313–342, 07 2015.

[5] J. Risbey, T. O’Kane, D. Monselesan, C. Franzke, and I. Horenko. Metastability of NorthernHemisphere teleconnection modes. J. Atmos. Sci., 72:35–54, 2015.

[6] Terence. O’Kane, James Risbey, Didier. Monselesan, Illia Horenko, and Christian Franzke.On the dynamics of persistent states and their secular trends in the waveguides of the southernhemisphere troposphere. Climate Dynamics, pages 1–31, 2015.

[7] O. Kaiser, D. Igdalov, and I. Horenko. Statistical regression analysis of threshold excesseswith systematically missing covariates. Multiscale Modeling & Simulation, 13(2):594–613,2015.

[8] S. Gerber and Horenko, I. Improving clustering by imposing network information. ScienceAdvances, 1(7), 2015.

[9] C. L. E. Franzke, T. J. O’Kane, D. P. Monselesan, J. S. Risbey, and Horenko, I. System-atic attribution of observed southern hemispheric circulation trends to external forcing andinternal variability. Nonlinear Processes in Geophysics Discussions, 2(2):675–707, 2015.

[10] O. Kaiser and I. Horenko. On inference of statistical regression models for extreme eventsbased on incomplete observation data. Communications in Applied Mathematics and Com-putational Science, 2014.

[11] S. Gerber and I. Horenko. On inference of causality for discrete state models in a multiscalecontext -. Proceedings of the National Academy of Sciences of the USA (PNAS) -as a directsubmission -, 111(41):14651–14656, 2014.

[12] J. de Wiljes, L. Putzig, and I. Horenko. Discrete non-homogenous and non-stationarylogistic and Markov regression models for spatio-temporal data with unresolved externalinfluences. Comm. in Appl. Math. and Comp. Sci. (CAMCoS), 8(2):101–146, 2014.

[13] T. O’Kane, J. Risbey, Ch. Franzke, I. Horenko, and D. Monselesan. Changes in the metastabil-ity of the midlatitude southern hemisphere circulation and the utility of nonstationary clusteranalysis and split-flow blocking indices as diagnostic tools. J. Atmos. Sci., 70:824–842, 2013.

1

[14] T. O’Kane, R. Matear, M. Chamberlain, J. Risbey, B. Sloyan, and I. Horenko. Decadalvariability in an OGCM Southern Ocean: Intrinsic modes, forced modes and metastablestates. Ocean Modelling, 69:1–21, 2013.

[15] T. O’Kane, R. Matear, M. Chamberlain, J. Risbey, I. Horenko, and B. Sloyan. Low fre-quency variability in a coupled ocean-sea ice general circulation model of the Southern Ocean.ANZIAM J., 54:200–216, 2013.

[16] J. de Wiljes and A. Majda and I. Horenko. An adaptive Markov chain Monte Carlo approachto time series clustering of processes with regime transition behavior. SIAM Mult. Mod. Sim.,11(2):415–441, 2013.

[17] P. Metzner, L. Putzig, and I. Horenko. Analysis of persistent non-stationary time seriesand applications. Comm. in Appl. Math. and Comp. Sci. (CAMCoS), 7(2):175–229, 2012.

[18] D. Giannakis, A. Majda, and I. Horenko. Information theory, model error, and predictiveskill of stochastic models for complex nonlinear systems. Physica D: Nonlinear Phenomena,241(20):1735–1752, 2012.

[19] Ch. Blume, K. Matthes, and I. Horenko. Supervised learning approaches to classify strato-spheric warming events. Journal of the Atmospheric Sciences, 69:1824–1840, 2012.

[20] I. Horenko. On analysis of nonstationary categorical data time series: dynamical dimensionreduction, model selection and applications to computational sociology. SIAM Multi. Mod.Sim., 9(4):1700–1726, 2011.

[21] I. Horenko. Nonstationarity in multifactor models of discrete jump processes, memory andapplication to cloud modeling. J. Atmos. Sci., 68(7):1493–1506, 2011.

[22] I. Horenko and Ch. Schutte. On metastable conformational analysis of nonequilibriumbiomolecular time series. SIAM Mult. Mod. Sim., 8(2):701–716, 2010.

[23] I. Horenko. On identification of nonstationary factor models and its application to atmo-spherical data analysis. J. of Atmos. Sci., 67(5):1559–1574, 2010.

[24] I. Horenko. On clustering of non-stationary meteorological time series. Dynam. of Atm.and Oceans, 49:164–187, 2010.

[25] I. Horenko. Finite element approach to clustering of multidimensional time series. SIAMJ. of Sci. Comp., 32(1):62–83, 2010.

[26] L. Putzig, D. Becherer, and I. Horenko. Optimal allocation of futures portfolio utilizingnumerical market phase-detection. SIAM J. Financial Mathematics, 1(1):752–779, 2010.

[27] I. Horenko. On robust estimation of low-frequency variability trends in discrete Markoviansequences of atmospheric circulation patterns. J. of Atmos. Sci., 66(11):1941–1954, 2009.

[28] Ch. Franzke, I. Horenko, A. Majda, and R. Klein. Systematic metastable atmospheric regimeidentification in an AGCM. J. Atmos. Sci., 66(7):1997–2012, 2009.

[29] I. Horenko and R. Klein, S. Dolaptchiev, and Ch. Schutte. Automated generation of reducedstochastic weather models I: simultaneous dimension and model reduction for time seriesanalysis. SIAM Mult. Mod. Sim., 6(4):1125–1145, 2008.

[30] I. Horenko and Ch. Schutte. Likelihood-based estimation of multidimensional Langevinmodels and its application to biomolecular dynamics. SIAM Mult. Mod. Sim., 7(2):731–773,2008.

[31] I. Horenko, S. Dolaptchiev, A. Eliseev, I. Mokhov, and R. Klein. Metastable decompositionof high-dimensional meteorological data with gaps. J. of Atmos. Sci., 65(11):3479–3496, 2008.

2

[32] I. Horenko, E. Dittmer, F. Lankas, J. Maddocks, P. Metzner, and Ch. Schutte. Macroscopicdynamics of complex metastable systems: theory, algorithms and application to B-DNA.SIAM J. Appl. Dyn. Syst., 7(2):532–560, 2008.

[33] I. Horenko. On simultaneous data-based dimension reduction an hidden phase identification.J. of Atmos. Sci., 65(6):1941–1954, 2008.

[34] I. Horenko, C. Hartmann, Ch. Schutte, and F. Noe. Data-based parameter estimation ofgeneralized multidimensional Langevin processes. Phys. Rev. E, 76(01):016706, 2007.

[35] F. Noe, I. Horenko, Ch. Schutte, and J. C. Smith. Hierarchical analysis of conforma-tional dynamics in biomolecules: Transition networks of metastable states. J. Chem. Phys.,126(15):155102, 2007.

[36] Ph. Metzner, I. Horenko, and Ch. Schutte. Generator estimation of Markov jump processesbased on incomplete observations nonequidistant in time. Phys. Rev. E, 76(06):066702, 2007.

[37] A. Fischer, S. Waldhausen, I. Horenko, E. Meerbach, and Ch. Schutte. Identification ofbiomolecular conformations from incomplete torsion angle observations by Hidden MarkovModels. J. Comp. Chem., 28(15):2453–2464, 2007.

[38] I. Horenko, E. Dittmer, A. Fischer, and Ch. Schutte. Automated model reduction forcomplex systems exhibiting metastability. SIAM Mult. Mod. Sim., 5(3):802–.827, 2006.

[39] I. Horenko, S. Lorenz, Ch. Schutte, and W. Huisinga. Adaptive approach for non-linearsensitivity analysis of reaction kinetics. J. Comp. Chem., 26(9):941–948, 2005.

[40] I. Horenko, E. Dittmer, and Ch. Schutte. Reduced stochastic models for complex molecularsystems. SIAM Comp. Vis. Sci., 9(2):89–102, 2005.

[41] I. Horenko, M. Weiser, B. Schmidt, and Ch. Schutte. Fully adaptive propagation of thequantum-classical liouville equation. J. Chem. Phys., 2004.

[42] I. Horenko and M. Weiser. Adaptive integration of molecular dynamics. J. Comp. Chem.,2003.

[43] I. Horenko, Ch. Salzmann, B. Schmidt, and Ch. Schutte. Quantum-classical liouville ap-proach to molecular dynamics: Surface hopping gaussian phase-space packets. J. Chem.Phys., 117(24):11075–11088, 2002.

[44] I. Horenko, B. Schmidt, and Ch. Schutte. Multidimensional classical liouville dynamicswith quantum initial conditions. J. Chem. Phys., 2002.

[45] I. Horenko, B. Schmidt, and Ch. Schutte. A theoretical model for molecules interactingwith intense laser pulses: The floquet-based quantum-classical liouville equation. J. Chem.Phys., 115(13):5733–5743, 2001.

Book Chapters and Proceedings

[46] P. Metzner, M. Waidmann, D. Igdalov, T. von Larcher, I. Horenko, R. Klein, A. Beck,G. Gassner, and C.D. Munz. A stochastic closure approach for les with application to tur-bulent channel flow. In Jochen Frhlich, Hans Kuerten, Bernard J. Geurts, and VincenzoArmenio, editors, Direct and Large-Eddy Simulation IX, volume 20 of ERCOFTAC Series,pages 49–55. Springer International Publishing, 2015.

3

[47] Th.von Larcher, R. Klein, I. Horenko, P. Metzner, M. Waidmann, D. Igdalov, A.D. Beck, G.J.Gassner, and C.D. Munz. Towards a stochastic closure approach for large eddy simulation.In Jurgen Fuhrmann, Mario Ohlberger, and Christian Rohde, editors, Finite Volumes forComplex Applications VII-Elliptic, Parabolic and Hyperbolic Problems, volume 78 of SpringerProceedings in Mathematics and Statistics, pages 883–890. Springer International Publishing,2014.

[48] Terence J. O’Kane, James S. Risbey, Christian Franzke, Illia Horenko, and Didier P. Mon-selesan. The metastability of the mid-latitude southern hemisphere circulation. In ScottMcCue, Tim Moroney, Dann Mallet, and Judith Bunder, editors, Proceedings of the 16thBiennial Computational Techniques and Applications Conference, CTAC-2012, volume 54 ofANZIAM J., pages C233–C249, May 2013.

[49] E. Meerbach, Ch. Schutte, I. Horenko, and B. Schmidt. Metastable conformational structureand dynamics: Peptides between gas phase and aqueous solution. In O. Kuhn and L. Woste,editors, Analysis and Control of Ultrafast Photoinduced Reactions, volume 87 of Series inChemical Physics, pages 796–806. Springer, 2007.

[50] A. Weisse, I. Horenko, and W. Huisinga. Adaptive approach for modelling variability inpharmacokinetics. In M. Berthold, R. Glen, and I. Fischer, editors, Lecture Notes in ComputerScience, volume 4216 of Computational Life Sciences, pages 194–204. Springer, 2006.

[51] I. Horenko and J. Schmidt-Ehrenberg and Ch. Schutte. Set-oriented dimension reduction:Localizing principal component analysis via hidden Markov models. In M. Berthold, R. Glen,and I. Fischer, editors, Lecture Notes in Computer Science, volume 4216 of ComputationalLife Sciences, pages 74–85. Springer, 2006.

[52] E. Meerbach, E. Dittmer, I. Horenko, and Ch. Schutte. Multiscale modelling in moleculardynamics: Biomolecular conformations as metastable states. In M. Ferrario, G. Ciccotti, andK. Binder, editors, Computer Simulations in Condensed Matter: Systems: From Materials toChemical Biology. Volume I, volume 703 of Lecture Notes in Physics, pages 475–497. Springer,2006.

Patents

[53] I. Horenko and Ch. Schutte. Method for operation of the signal device. WorldIntellectual Property Organisation, Pub. No. WO/2008/113705, 2008. (Available atwww.wipo.int/pctdb/en/wo.jsp?WO=2008113705).

4

Publication Impact

Bibliometrics1:

• 52 peer-reviewed journal publications

• 1 patent

• h-index: 21

• Total number of citations: 1107

• Number of citations since joining USI in 2010: 787

Number of citations for my ten most cited papers

cited Article Reference

219 Hierarchical analysis of conformational dynamics inbiomolecules:Transition networks of metastable states

[35]

105 Quantum-Classical Liouville Approach to Molecular Dynamics:Surface Hopping Gaussian Phase-Space Packets

[43]

46 Automated model reduction for complex systems exhibitingmetastability

[38]

45 Finite element approach to clustering of multidimensional timeseries

[25]

41 Fully adaptive propagation of the quantum-classical Liouvilleequation

[41]

40 On the identification of nonstationary factor models and theirapplication

[23]

39 Systematic metastable atmospheric regime identification in anAGCM

[28]

37 On clustering of non-stationary meteorological time series [24]35 Likelihood-based estimation of multidimensional Langevin models

and its Application to Biomolecular Dynamics[30]

31 Data-based Parameter Estimation of Generalized Multidimen-sional Langevin Processes

[34]

1Google Scholar as of 08/2015. scholar.google.com

5

Impact factor of journals I published in 2:

IF Journal name Articles published9.67 Proceedings of the National Academy of Sciences of USA (PNAS) [11]4.67 Climate Dynamics [6]3.59 Journal of Computational Chemistry [42, 39, 37]3.55 Water Resources Research [2]3.03 Journal of the Atmospheric Sciences (J. of Atmos. Sci.) [33, 31, 28, 27, 23]

[21, 19, 13, 5]3.02 The Journal of Chemical Physics [45, 43, 44, 41, 35]2.93 Ocean Modeling & Simulation [14]2.29 Physical Review E [36, 34]1.80 Multiscale Modeling & Simulation [7]2.40 Communications in Applied Mathematics and Comp. Science [17, 12, 10]2.01 SIAM Journal of Multiscale Modeling and Simulation [38, 29, 30, 22, 20, 16]1.90 SIAM Journal of Scientific Computing [25]1.89 Physica D Nonlinear Phenomena [18]1.69 Nonlinear Processes in Geophysics [9][1]1.58 Dynamics of Atmosphere and Oceans (DYNAT) [24]1.44 SIAM Journal on Applied Dynamical Systems [32]1.19 Meteorologische Zeitschrift [4]1.03 The Journal of the Australian Mathematical Society (ANZIAM) [15]1.25 SIAM Journal on Financial Mathematics [26]

2As of 2014. Source: http://www.journal-database.com

6

Statements on Research, Teaching and ServiceSince becoming an Associate Professor at USI (2010-2015)

ILLIA HORENKOUniversità della Svizzera italianaSeptember 2, 2015

Teaching

Coming to USI in 2010 as the youngest tenured faculty member (joined USI INF Dept. as anassociate professor at the age of 32), I was one of the authors of a new master program in“Applied Mathematics and Computational Science“ (AMCS). Furthermore, I was the academicdirector of this program between 2011 and the beginning of 2014.To satisfy my mandatory teaching assignment in Lugano, I created a pipeline of three novelconsecutive courses: “Deterministic Methods“ (first semester of AMCS, 6 ECTS); “StochasticMethods“ (second semester of AMCS, 6 ECTS) and “Comp. Data Analysis“ (third semesterof AMCS, 6 ECTS). Together, these three courses build one of the pillars of the mastercuriculum and combine elements from the following disciplines:

(i) Pure and applied mathematics (scalar/vector/matrix/function analysis, algebra, realand stochastic calculus, numerical analysis, probability theory);

(ii) Informatics (databases, parallel computing, randomized algorithms, information re-trieval and machine learning);

(iii) Application areas (classical and statistical physics, biophysics, bioinformatics, cli-mate/atmosphere/ocean science, economics and sociology, engineering).

The first course in this pipeline - Deterministic Methods - is dedicated to the introductionof the basic concepts necessary for understanding and high performance implementation ofdeterministic computational algorithms in the area of scientific computing. Concepts, meth-ods and algorithms are always first motivated by application examples from different areas(e.g., from robotics/mechanics, data compression, insurance/banking). The recurrent themeof the course is built upon the concept, that those respective computational problems andalgorithms can be described from the common optimizational perspective; also the necessaryphysics background is introduced (e.g., Eulerian and Lagrangian formulations of classicalmechanics). Relation is established to the concepts from partial differential equations that areintroduced in the other courses of AMCS Master in the first semester.The second course in the pipeline Stochastic Methods is addressing the problem that manyof the real-life applications (e.g., in banking/insurance, mechanics, medicine, etc.) can onlybe approached, modelled and computed as stochastic (or random) processes. The aim of this

1

course is to introduce the most essential mathematical concepts and computational methodsfrom the area of stochastic and random processes. Besides of gaining the theoretical andpractical background in the areas of stochastic calculus, random processes and uncertaintyquantification, the participants are gain practical skills by doing supervised short researchprojects from real-life applications. The recurrent theme of the course is in establishing a jointstochastic/statistic perspective and optimizational formulation for various computationalmethods and algorithms from computational science, machine learning and informatics.The last course in the pipeline - Computational Data Analysis - aims to give a wide practicaloverview of data mining, time series analysis and inverse problems from the viewpoints ofapplied mathematics, statistics and high performance computing strategies. The topics consid-ered in the course include: clustering methods; concepts from information theory; generalizedlinear models for statistical regression of discrete processes; statistical methods of inferencebeyond the homogeneity/stationarity assumptions (local kernel methods, FEM-based meth-ods). Parametric vs. non-parametric methods. Applications of considered methods andtools is illustrated on realistic data sets (e.g., problems from fluid dynamics, climate/weatherresearch, biophysics, sociology, finance/insurance). Here - as in two previous courses of thepipeline - it is demonstrated that all of the considered methods can be formulated and solvedas optimization problems, herewith undermining the main perspective of the whole pipelineand giving a joint coherent picture of very different numerical methods and computationaltools.

List of lectured courses:

Four times during the years 2011-2014 these courses have been ranked by the students as thebest master courses at the USI faculty of informatics.

Spring 2016 Stochastic Methods - (Master course, 6 ECTS)Fall 2015 Probability & Statistics - (Bachelor course, 6 ECTS)Fall 2015 Deterministic Methods - (Master course, 6 ECTS)Spring 2015 Stochastic Methods - (Master course, 6 ECTS)Fall 2014 Computational Data Analysis - (Master course, 6 ECTS)Fall 2014 Deterministic Methods - (Master course, 6 ECTS)Spring 2014 Stochastic Methods - (Master course, 6 ECTS)Fall 2013 Computational Data Analysis - (Master course, 6 ECTS)Fall 2013 Deterministic Methods - (Master course, 6 ECTS)Spring 2013 Stochastic Methods - (Master course, 6 ECTS)Fall 2012 Computational Data Analysis - (Master course, 6 ECTS)Fall 2012 Deterministic Methods - (Master course, 6 ECTS)Spring 2012 Stochastic Methods - (Master course, 6 ECTS)Fall 2011 Computational Data Analysis - (Master course, 6 ECTS)Fall 2011 Advanced Computational methods I - (Master course, 6 ECTS)Spring 2011 Advanced Computational methods II - (Master course, 6 ECTS)Fall 2010 Computational Data Analysis - (Master course, 6 ECTS)Fall 2010 Advanced Computational methods I - (Master course, 6 ECTS)

2

Supervising

During my time at USI I have supervised

• 4 Post docs: Dr. Philip Metzner, Dr. Thomas von Larcher, Dr. Susanne Gerber and Dr.Olga Kaiser (the latter two ones are currently working in my research group).

• 7 PhD students: Lars Putzig, Jana de Wiljes, Olga Kaiser, Christian Blume (as a co-supervisor), Dimitri Igdalov, Anna Marchenko and Katia Fiorucci. The first 4 studentsare meanwhile successfully graduated and D. Igdalov as well as A. Marchenko areconfident of receiving their doctorate an the near future. K. Fiorucci, however, decidedagains an academic career for personal reasons and abandoned her PhD training.

• 3 Master students: Luca Vignola, Alena Kopanicakova and Lawrence Farinola. All ofthem received successfully their master’s degree.

• One Diploma student: Jana de Wiljes. She graduated at the Freie Universität under my"long-distance-supervision" since she has her family in Berlin and did not move togetherwith me to Lugano.

• One Bachelor student: Kristine Kaiser did her semester abroad in my research group atUSI in summer 2010

3

Service Statement

During my time as an associate professor, I served as the academic director of the master pro-gram “Applied Mathematics and Computational Science“ as well as a member of dissertationand examinations committees in the faculty on the regular basis. Besides of that I investeda lot of effort into building collaboration networks, both within the faculty and within theuniversity.

Work in committees

On the regular basis and whenever asked by the faculty, I was participating in numerouscommittees appointed by the faculty, including master thesis defense committees, PhDprospectus, proposal and PhD defense committees as well as in the promotion committeesfor various faculty colleagues.

Reviewing

I serve frequently as a reviewer for the journals of the Society of Industrial and AppliedMathematics (SIAM MMS, SISC, SIADS), American Meteorological Society (e.g., in journallike JAS, JCli, Climate Dynamics etc.) and as a member of review panel boards for largeprograms (e.g., the program "Mathematics of Planet Earth" of the Science Foundation ofNetherlands NWO). Between 2010 and 2014 I have co-chaired several international workshopsand conferences (e.g., IPAM/UCLA conferences).

Providing open source software and tools

My research group and I keep maintaining and updating the open-source FEM-BV-toolboxfor nonstationary time series analysis, summarizing all of the methods developed in mygroup as a documented collection of Matlab tools and application examples. Currently thisFEM-BV-Toolbox is used by various research groups , e.g.: in Australia (CSIRO in Hobart,Tasmania), United States (the groups of D. Giannakis and A. Majda at the Courant Institute ofNYU, New York), Germany (e.g., the group of Valerio Lucarini in Hamburg, group of RupertKlein in Berlin).

Establishing internal collaboration networks

Inside the faculty, productive collaborations have been initiated with the following researchgroups:

• The group of Michele Parrinello (investigating new possibilities to infer rates of chemi-cal reactions from non-equilibrium metadynamics data);

• The group of Mauro Pezze (investigating potential of advanced time series analysismethods in prediction of workload and performance in cloud computing systems);

• The group of Olaf Schenk (FEM and efficient HPC solvers for large structured linearand quadratic programming problems);

4

On the university level, I have initiated a collaboration with Prof. Patrick Gagliardini fromthe USI faculty of Economics. This collaboration resulted in two joint research projects - with541.346 CHF of the total research funding. One of these projects is funded by SNF and isdealing with computational data analysis of credit and eqity risks in finance (with 319.696CHF of the total funding). The second project deals with causality inference problems infinance and is funded with 221.650 CHF by the Platform for Advanced Scientific Computing(PASC). This collaboration also involves the Swiss Supercomputing Center CSCS in Lugano -Will Sawyer from CSCS is the co-PI of both projects. Collaboration with CSCS was establishedalso in other joint projects: e.g., in terms of the ETHZ-financed internship, CSCS is currentlyinvestigating the potential of deploying the lossy compression methods developed in my USIresearch group for the operational compression of very large climate simulation data setsproduced in CSCS.More details on the results of these collaborations can be found below in the section Applica-tion part of the research.

Establishing international collaboration networks

Besides my ambition of contributing to a wide branched university-internal network, I havealso established an multidisciplinary network of international scientists. Below is a list ofworld leading experts in their fields, that visited USI Lugano within the last 5 years on myinitiative.

07/2015 Prof. Dr. Andrew J. Majda, Department of Mathematics and Climate,Atmosphere, Ocean Science (CAOS), Courant Institute of MathematicalSciences, New York University

09/2013 Dr. Terence O’Kane is Senior Research Scientist at the Centre for Aus-tralian Weather & Climate Research (CSIRO) Marine & AtmosphericResearch, Hobart, Australia.

08/2013 Prof. Dr. Andrew J. Majda, Department of Mathematics and Climate,Atmosphere, Ocean Science (CAOS), Courant Institute of MathematicalSciences, New York University

08/2012 Prof. Dr. Bruce Turkington, University of Massachusetts, Departmentof Mathematics and Statistics, Amherst Center, MA, United States

04/2012 Prof. Dr. Dr. h.c. Edda Klipp, Humboldt-Universität zu Berlin, Theo-retical Biophysics, Germany

04/2012 Prof. Dr. Andreas Herrman, Humboldt-Universität zu Berlin, Molecu-lar Biophysics

03/2012 Prof. Dr.-Ing. Rupert Klein, Freie Universität Berlin, Departmentof Mathematics and Computer Science, Geophysical Fluid DynamicsGroup, Germany

5

08/2011 Ph. D Hella Lichtenberg, Universität Bonn, Institute for Cellular andMolecular Botanics (IZMB), Head of the research group molecular Bioen-ergetics.

08/2011 Prof. Dr. Carla Falugi, University of Genoa (UNIGE), Head of theLaboratory of Experimental Embryology in the Department for theStudy of Territory and its Resources (DIPTERIS).

09/2010 Prof. Dr. Andrew J. Majda, Department of Mathematics and Climate,Atmosphere, Ocean Science (CAOS), Courant Institute of MathematicalSciences, New York University

Organisation of workshops and conferences

Within the last 5 years I organized and co-organized the following international events:

2015 Oberwolfach Seminar Different mathematical perspectives on descriptionof unresolved scales in multiscale systems (together with R. Klein, C. Hart-mann and T. O’Kane), Oberwolfach (Germany)

2012 DFG Summer School Methods of Data Analysis for Fluid Mechanics, Mete-orology and Climate Research, (co-organizer with Sebastian Reich), Berlin(Germany)

2012 Workshop Non-stationary Methods of Data Analysis and Applications,Lugano (Switzerland)

2011 Conference Advances in Computational Science,(co-organizer with RolfKrause and Michele Parinello), Lugano (Switzerland)

2011 Swiss Numerics Colloquium (co-organizer with Rolf Krause), Lugano(Switzerland)

2010 Workshop Data Hierarchies for Climate Modeling at the Institute of Pureand Applied Mathematics (IPAM) in Los Angeles (co-organizer togetherwith Amy Braverman, Luis Kornblueh and Robert Pincus, Los Angeles(USA)

6

Research statement (since 2010)

My main research motivation lies in the design of HPC-scalable methods for handling realisticmulti-scale data beyond the standard statistics assumptions. The methods I am developingshould be applicable to understanding data coming from very different areas like geosciences,biophysics, bioinformatics, sociology, economics, engineering. My second main area ofinterest is the implementation of theoretical concepts and results from modern pure and ap-plied mathematics in context of statistics and data analysis, especially considering the aspectof high-performance computing (HPC) on modern and emerging/future supercomputerarchitectures.

Mathematical part of the research

Combining the concepts and ideas from pure and applied mathematics (such as FiniteElement Method (FEM) from numerics of partial differential equations, regularization ininfinite-dimensional spaces from the theory of ill-posed problems, stochastic calculus andtheory of stochastic processes, information criteria from information theory, embeddingtheorems from the theory of dynamical systems), I and my research group have created afamily of non-stationary, non-homogenous and non-parametric time series analysis methods.This FEM-BV family of time series analysis techniques, which is reviewed concisely in θi ([Ph.Metzner, L. Putzig, I. Horenko Analysis of persistent non-stationary time series and applications,Communications in Applied Mathematics and Computational Science, 7(2), 175-229, 2012]),allows for systematic time-dependent model identification when assumptions of temporalstationarity or spatial homogeneity of some underlying statistics are not justifiable.The main idea is based on a regularized variational minimization of a scalar-valued functionaldescribing the distance g(xt, θ(t)) of the model output (being characterized by the time-dependent set of parameters θ(t)) from the given observation xt at time t:

∫ T

0g(xt, θ(t))dt→ min

θ(t

Regularization ansatz: g(x, θ(t)) = ∑Ki=1 γi(t)g(xt, θi)

FEM-BV-approach:

L(Θ, Γ(t)) =∫ T

0

K

∑i=1

γi(t)g(xt, θi)→ minΓ(t),Θ

K

∑i=1

γi(t) = 1, γi(t) ≥ 0, ∀t ∈ [0, T], i = 1, ..., K

||γi(t)||BV ≤ C, (Bounded Variation constraints)

The HPC numerical scheme of this approach is based on the adaptive

Finite Elemente Method [1, 2, 3, 4, 5, 6]

In contrast to standard BV-, L1- andL2-regularized statistical inverseproblem formulations (that can beall reformulated as convex regu-larized linear regression problemswith known regressors θi and canbe followed back to ideas originallyformulated by Grace Wahba in 80’sand 90’s and based on Tykhonovand Phillips regularization ideasfrom 40’s), the above variationalproblem is non-convex since regres-sors θi are not known a priori andneed to be inferred simultaneouslywith the inference of γi(·).

The nice algebraic structure of the above problem allows to deploy efficient numerical al-gorithms, based on iterative application of linear and/or quadratic programming problems(optimizing γi(·). for fixed θi) followed by stationary/convex inference of θifor a fixed γi(·).

7

This procedure (repeated until the change in L is less then some small threshold) results inmonotonic minimization of L. Moreover, as demonstrated in [6] standard clustering methods(like K-Means) and Bayesian learning methods (like Gaussian Mixture Models and Hid-den Markov Models) are special cases of this algorithmic realization of the more generalvariational FEM-BV-problem shown above, if some additional assumptions/constraints areadded. E.g., HMMs are obtained additionally assuming that γi(t). is an output of homoge-nous Markov chain and setting C = T.

Finite Element Methods are employed in the numerical representation of indicator functionsγi(·). for the space-time domains of applicability of different models from a common modelclass (characterized by the choice of the particular form of the error function g(xt, θ(t)) . Asshown above, these indicators are regularized using a Bounded Variation criterion, hence theacronym “FEM-BV”. The choice of the model class depends on the type of data considered.The current implemented versions include:

(i) Vector Auto-Regressiv models with eXternal influences (VARX) with finite memorydepth (for regression analysis of continuous data);

(ii) “K-means” models for geometric clustering of continuous data;

(iii) Empirical Orthogonal Function decompositions with and without Takens-embeddingfor model reduction of continuous data in high-dimensional vector spaces;

(iv) Markov/Bernoulli and generalized linear models for discrete (categorical) data;

(v) Generalized Extreme-Value distributions (GEV) for regression analysis with emphasison extreme event characteristics;

(vi) Bayesian causality inference in multiscale systems (for causality analysis of discrete/Booleandata).

The number of different spatio-temporal regimes, the model parameters to be chosen withinthese regimes, such as memory depth and number of EOFs, and the indicator functions γi(·)signaling activation of the respective models are all determined simultaneously in a globaloptimization procedure. This yields a judicious compromise between low residuals in repro-ducing the data of a training set on the one hand, and the demand for the smallest-possibleoverall number of free parameters of the complete model on the other. The optimization isbased on a new non-parametric modified Akaike Information Criterion (mAIC) and may thusbe interpreted as a constructive implementation of “Occam’s Razor” for data analysis problem.

8

Figure 1: Ideogram of developed methods and their relation to existing approaches

9

Informatics part of the research

To analyze the large sets of data by solving the constrained variational FEM-BV-problemsthat emerge in the mathematical part of research, three types of numerical approaches hasbeen deployed and investigated so far wrt. their parallelizability and HPC-performance:

(i) Adaptive Finite Element Method;

(ii) Adaptive Markov Chain Monte Carlo Methods (in collaboration with Andrew Majda,Courant Institute of Mathematical Sciences New York);

(iii) Sequential Minimization Methods (based on domain decomposition ideas for γi(·)).

In close collaboration with experts from Swiss Supercomputing Center in Lugano (e.g., withWilliam Sawyer from CSCS/ETHZ), an open-source toolbox for FEM-BV time series analysiswas developed, implemented and tested on the Monte Rosa Cray XE6 supercomputer forparallel analysis of 40960 different time series. As demonstrated in the Figure below, method-ology demonstrates almost perfect parallelizability and is thereby applicable to analysis ofvery large sets of data.

10

Application part of the research (already published work):

a) Applications to Geoscience

b) Applications to Biophysics and Bioinformatics

11

c) Applications to Social Science and Economics

d) Application to Neuroscience (published in Science Advances)

12

Application part of research (yet unpublished work):

e) Applications to Engineering

f) Application to Computational Genomics (submitted to Science Advances)

13

g) Applications to Cloud Computing in Informatics (collaboration with Mauro Pezze, USILugano)

h) Applications to Ab Initio Estimation of Biological/Chemical Reaction Kinetics Rates(collaboration with Michele Parrinello, USI/ETHZ)

14

References

1. I. Horenko. Finite element approach to clustering of multidimensional time series. SIAM J.of Sci. Comp., 32(1):62–83, 2010.

2. I. Horenko. On clustering of non-stationary meteorological time series. Dynam. of Atm. andOceans, 49:164–187, 2010.

3. I. Horenko. On identification of nonstationary factor models and its application to atmo-spherical data analysis. J. of Atmos. Sci., 67(5):1559–1574, 2010.

4. I. Horenko. On analysis of nonstationary categorical data time series: dynamical dimensionreduction, model selection and applications to computational sociology. SIAM Multi. Mod.Sim., 9(4):1700–1726, 2011.

5. I. Horenko. Nonstationarity in multifactor models of discrete jump processes, memoryand application to cloud modeling. J. Atmos. Sci., 68(7):1493–1506, 2011.

6. P. Metzner, L. Putzig, and I. Horenko. Analysis of persistent non-stationary time seriesand applications. Comm. in Appl. Math. and Comp. Sci. (CAMCoS), 7(2):175–229, 2012.

15