Curriculum Vitae of ALESSANDRO CHIUSO
Department of Information Engineering - University of Padova
Via Gradenigo, 6/b, 35131 Padova, Italy
Ph:+39-049-8277709 - Fax: +39-049-8277699 - email: chiuso@dei.unipd.it
Home page: http://automatica.dei.unipd.it/people/chiuso.html
Citation Metrics: links to Scholar Profile, Scopus Profile.
June 2016
PERSONAL DATA
Born on December 21st, 1972 in Venice (Italy)
EDUCATION
-Ph.D. Degree (Dottorato di ricerca) in System Engineering, University of Bologna. Thesis: “Geometric
Methods for Subspace Identification”, Advisor: Prof. G. Picci.
02/2000.
-Laurea Degree (joint B.S./M.S. equivalent) in Telecommunication Engineering (summa cum laude),
University of Padova (thesis: Image reconstruction from projections), Advisor: Prof. G. Cariolaro.
07/1996.
PRESENT POSITION
Associate professor, University of Padova
03/2006-
PREVIOUS/VISITING POSITIONS
-Research Consultant, University of California Los Angeles
-Researcher (Assistant professor), University of Padova
-Visiting Researcher, University of California Los Angeles, USA
-Research associate (“Assegnista di ricerca”), University of Padova
-EU-TMR Post-Doctoral fellow Royal Institute of Technology, Sweden
-Visiting Research Scholar, Washington University, USA
10/2011-09/2012
03/2001-02/2006
07/2001
08/2000-02/2001
03/2000-07/2000
09/1998-06/1999
INVITED SHORT TERM VISITS
Chinese Academy of Sciences (Invited for summer-fall 2016), University of Kyoto (11/2011), Linköping
University (5/2011 and 11/2011), University of California Los Angeles (08/2010), Royal Institute of
Technology (05/2007), University of Melbourne (04/2006), University of Texas at Austin (05/2005),
University of California Los Angeles, (05/2005-06/2005).
CITATION METRICS (as of June 8th, 2016)
-Scopus (Link to Scopus Profile). Total citations: 2016, h-index : 20.
-Scholar (Link to Scholar Profile). Total citations: 3904, h-index : 26.
1
HONORS AND AWARDS
-Invited Plenary Speaker at the 15th IFAC Symposium on System Identification (SYSID), Beijing,
China, October 2015 (Link).
-Best Oral Presentation: “For the relevance of the results and clarity of presentation” (SIDRA
annual meeting, Benevento, September 2012)
-Outstanding Reviewer (IEEE Transactions on Automatic Control, 2009)
-Outstanding Reviewer (Automatica, 2007, awarded to a pool of about 30 referees out of about 1300)
-IEEE Senior Member (2006)
-Ing. Aldo Gini Foundation Fellowship, Padova, Italy, 1999.
♦ FULL PROFESSOR QUALIFICATION
(09/G1 - Automatica - I Fascia)
-Abilitazione Scientifica Nazionale (Bando 2012) for the Full Professor position with the following
overall evaluation given by the Committee (only available in Italian):
“Il candidato presenta 20 pubblicazioni, la maggior parte in sedi altamente visibili nel settore concorsuale, di cui 9 pubblicate negli anni 2008-2012. Le pubblicazioni, che riguardano diverse tematiche
tra cui identificazione parametrica, subspace identification e stima distribuita, sono tutte coerenti con
le tematiche del settore concorsuale o con le tematiche interdisciplinari ad esso pertinenti. Nelle 16
pubblicazioni in collaborazione il contributo individuale, considerato paritetico, è valutato adeguato. La
qualità della produzione scientifica, caratterizzata da eccellente originalità, è svolta con eccellente
rigore metodologico ed è eccezionalmente innovativa. Relativamente ai Parametri di cui alle
lettere a)-g), il candidato possiede i seguenti titoli:
a) supera 3 mediane su 3.
b) responsabilità scientifica per progetti di ricerca internazionali e nazionali, ammessi al finanziamento
sulla base di bandi competitivi che prevedano la revisione tra pari.
d) partecipazione a comitati editoriali di riviste, collane editoriali, enciclopedie e trattati di riconosciuto
prestigio.
e) attribuzione di incarichi di insegnamento o di ricerca (fellowship) ufficiale presso atenei e istituti di
ricerca, esteri e internazionali, di alta qualificazione
Relativamente agli ulteriori criteri di valutazione, il candidato ha dimostrato:
Capacità di attrarre finanziamenti competitivi in qualità di responsabile di progetto e capacità di promuovere attività di trasferimento tecnologico.
=================
Giudizio complessivo
=================
Il candidato Chiuso Alessandro i cui indicatori dell’impatto della produzione scientifica complessiva
superano i requisiti richiesti per la prima fascia, viene valutato con giudizio di merito eccezionalmente positivo secondo i criteri e i parametri per la valutazione dei candidati, tenuto conto della
loro ponderazione. Al candidato Chiuso Alessandro viene attribuita l’abilitazione”
2
♦ RESEARCH FUNDING, GROUP, STUDENTS AND
COLLABORATIONS
RESEARCH FUNDING AS PRINCIPAL INVESTIGATOR (PI)
Total funding as PI ' 800Keuro
-National Coordinator (PI): “Progetto FIRB 2012” (RBFR12M3AC): Learning meets time: a new
computational approach for learning in dynamic systems. Total Funding 713Keuro.
-PI: “Progetto di Ateneo” (CPDA090135/09): Learning methods for estimation and identification of
large scale distributed dynamic systems, University of Padova (2010-2012), Total Funding 38Keuro.
- Coordinator - Ex 60% Project: Modeling, Estimation and Control. Automation group at the Department of Management and Engineering: 2007-2011, Total funding ' 34.7Keuro.
-PI: “Assegno di ricerca” (funds for a Post-Doc position) (CPDR058728): Sensor Networks for surveillance, University of Padova (2005). Total funding ' 14Keuro
RESEARCH FUNDING AS PRINCIPAL INVESTIGATOR (PI) - Submitted/In preparation
-National Coordinator (PI): “Progetto PRIN 2015”: Learning to Control (L2C): Data-driven methods
for constrained control of dynamical systems. Local Coordinators: Alberto Bemporad (IMT Lucca) and
Sergio Savaresi (PoliMI) (submitted)
-Coordinator (PI) “Progetto Ricerca SID 2016 (UniPD)”: Statistical learning methods for brain networks modeling” . Co-Investigators: Marco Zorzi, Alessandra Bertoldo, Leonardo Badia (in preparation).
RESEARCH FUNDING - PARTICIPANT
-Contract with the company MEMC s.p.a. (Merano), 2012.
-Contract with the company ELECTROLUX (Pordenone), 2012.
-HYCON2 Network of excellence (grant agreement #257462).
-European Project FeedNetBack [FP7/2007-2013] FP7-ICT-223866-FeedNetBack
-MIUR PRIN Projects: Algorithms and architectures for identification and control of industrial systems (1998), New techniques for identification and adaptive control of industrial systems (2000), New
techniques for identification and adaptive control of industrial systems (2002), New techniques for
identification and adaptive control of technological systems (2004), New techniques for Bayesian estimation, identification and distributed and adaptive control (2006), New Methods and Algorithms for
Identification and Adaptive Control of Technological Systems (2008).
-“Progetto di Ateneo”: MACONDO (Modelling, Analysis, and CONtrol of Deformable Objects) (2005).
-“Progetto di Ateneo”: Simulation of cognitive processes using generative neural networks (2003).
-U.S. Army Research Office DAAH0445 e DAAD19-99-1-0139 e NSF IIS-9876145
-Italian Space Agency project: Vision systems for autonomous navigation in space
-European Research Networks ERNSI (1998-2003) (European Research Network on System Identification).
-European Project RECSYS (2002-2005)
RESEARCH NETWORKS
-Team Leader of the UNIPD/Italy Team: European Research Network on System Identification
(ERNSI) (2013-) (home page).
3
RESEARCH GROUP AND STUDENTS
- Mattia Zorzi: Researcher (RTDa) (2014- ) funded by the MIUR FIRB project “Learning meets
time” coordinated by Alessandro Chiuso.
- Ph.D. supervisor of Giulia Prando: PhD student (2014- ), now visiting PhD student @UCBerkeley
(Prof. M.I. Jordan)
- Ph.D. supervisor of Diego Romeres: Supervised as research collaborator in 2013, PhD student
(2014- ), now visiting PhD student @TU Darmstadt (Prof. J. Peters)
- Main Supervisor of Francesca Carli: Post Doctoral fellow (Assegnista di Ricerca) (2011-2012) (hired
within the “Progetto di Ateneo” titled “Learning methods for estimation and identification of large
scale distributed dynamic systems” coordinated by Alessandro Chiuso. Now Post Doctoral Associate
@University of Cambridge (Prof. R. Sepulchre)
- co-PhD advisor of Andrea Masiero (Ph.D. 2006)
ACTIVE INTERNATIONAL/NATIONAL RESEARCH COLLABORATIONS
-Prof. L. Ljung, Linköping University
-Prof. S. Soatto, UCLA
-Prof. S. Dey, Uppsala University (now visiting University of Padova)
-Prof. T. Chen, The Chinese University of Hong Kong
-Prof. A. Aravkin, Prof. J. Burke, University of Washington
-Prof. G. De Nicolao, Università di Pavia
-Prof. L. Rosasco, IIT and Università di Genova
-Prof. S. Bonettini, Università di Ferrara
-Prof. S. Formentin, Politecnico di Milano
-Dr. Riccardo Muradore, ESO Munich, now Università di Verona
♦ EDITORIAL WORK, ORGANIZATION OF SCIENTIFIC
EVENTS, COMMITTEES
EDITORIAL BOARDS
• Current
-Associate Editor: IEEE Transactions on Control System Technology (2013-)
-Associate Editor: European Journal of Control (2011-)
-Associate Editor: Automatica (2008-)
• Past
-Associate Editor: IEEE Transactions on Automatic Control (2010-2012)
-Editorial Board Member: IET Control Theory and Applications (2007-2013)
-IEEE CSS Conference Editorial Board (2004-2009)
• Invited but declined due to overlapping editorial commitments:
-Invited to joint the all-electronic archival journal IEEE Access (2015)
-Invited to joint the Editorial Board of SIAM Journal of Control and Optimization (2013)
4
ORGANIZATION OF SCIENTIFIC EVENTS
• Conferences, Workshops and Schools:
- Organizer (Chair, as Team Leader of the UNIPD/Italy Team) of the 25th Workshop of the
European Research Network on System Identification (link to web page), September 2016, Italy.
- Invited organizer of the course on “Stochastic Systems” for the Italian summer school for Ph.D.
Students in Automatic Control, 07/2012.
- Organizer (Chair) of the Annual SIDRA (Italian Association of Automatic Control) meeting,
2008.
- Organizer (with A. Ferrante and S. Pinzoni) of the workshop: Modeling, Estimation and Control:
A Symposium in honor of Giorgio Picci on the occasion of his 65-th birthday, 2007.
- Local Co-Organizer of the ERNSI (European Research Network System Identification) workshop, 2007.
• Invited Sessions:
- Invited organizer of the Tutorial Session on “New developments in system identification” - European Control Conference 2014.
- Organizer of the invited session on “Advances in Estimation and Control in Wireless Sensor
Networks” - IEEE CDC 2013 (with S. Dey and L. Schenato)
- Organizer of the invited session “Sparse methods for model structure determination and variable
selection”, IFAC SYSID, 2012.
- Organizer of the invited session “Distributed Estimation over Sensor Networks”, IEEE CDC,
2007.
- Organizer of the invited session “New Developments in Closed-Loop Identification”, IFAC SYSID,
2006.
- Organizer of the invited session “New results in closed-loop identification”, IFAC World Congress,
2005.
- Organizer of the minisymposium “Identification and Control in Computer Vision”, MTNS, 2004.
- Organizer of the invited sessions “New results in subspace identification” and “Subspace identification and applications” IFAC SYSID, 2003.
- Organizer of the Tutorial “Dynamical Systems methods in Computer Vision”, ECCV, 2002.
SCIENTIFIC COMMITTEES (IPCs/TPCs, Award Committees)
-Vice-Chair of the IFAC Technical Committee on Modeling, Identification and Signal Processing
(2014-).
-2017 IFAC World Congress: Technical Associate Editor for TC1.1.: Modeling, Identification, and
Signal Processing.
-2015 Annual SIDRA (Italian Association of Automatic Control) meeting, Member of the Awards
Committee
-ICRA 2015 Workshop on Sensorimotor Learning: Program Committee Member
-2013, 2015 Annual SIDRA (Italian Association of Automatic Control) meeting - Technical Program
Committee Member.
-2014 SIDRA Annual Meeting (Italian Association of Automatic Control), Invited to be member of the
Awards Committee, declined due to an overlapping Workshop in Sestri Levante
-2009, 2012, 2015 IFAC Symposium on System Identification, IPC Member.
-2014 IFAC World Congress: Technical Associate Editor for TC1.1.: Modeling, Identification, and
Signal Processing.
-2013 ROKS: Regularization, Optimization, Kernel Methods and Support Vector Machines: theory and
applications. Leuven, July 2013, IPC Member.
-2012 ERNSI Workshop (Maastricht) (Co-chair for the program together with P. Van den Hof and
5
H. Hjalmarsson)
-2005, 2007, 2008 and 2009 IEEE Conference on Decision and Control, IPC Member.
-2004, 2008 and 2010 Int. Symposium on Mathematical Theory of Networks and Systems, IPC Member.
-2010 International Conference on Informatics in Control, Automation and Robotics, IPC Member.
-2011 Third Int. Workshop on Wireless Sensors, Actor and Robot Networks (WiSARN-Fall 2011),
IPC Member.
-2006 Mediterranean Control Conference, IPC Member.
-IEEE CSS Technical Committee on System Identification and Adaptive Control (member).
MEMBER OF COMMITTEES FOR HIRING/PROMOTIONS
-External evaluator for a promotion to Professor, Chalmers University of Technology, Sweden, 2016
-Committee member (chair) for an RTDa position, University of Padova, 2016
DOCTORAL THESIS COMMITTEES
-Examiner : Licenciate thesis defense of Riccardo Risuleo, KTH, Sweden (2016)
-Ph.D. Committee Member : Politecnico di Milano (2015)
-Opponent: Ph.D. defense of Christian Lyzell, Linköping University, Sweden (2012)
-External Evaluator : Ph.D. defense of Dr. Chih-Hong Wang, University of Melbourne, Australia (2011)
-Ph.D. Committee Member : Politecnico di Torino (2009)
-Ph.D. Committee Member : Università di Padova (2006, 2007)
-Ph.D. Committee Member : Università di Pavia (2006)
REFEREE FOR GRANT APPLICATIONS
-Evaluator
-Evaluator
-Evaluator
-Evaluator
-Evaluator
for
for
for
for
for
FONDECYT-CHILE, 2014
a Consolidator ERC grant application, 2014
the Belgian Research Foundation - Flanders (FWO), 2014
an Advanced ERC grant application, 2013
the Romanian National Council for Development and Innovation, 2011-2012.
DUTIES WITHIN THE UNIVERSITY OF PADOVA
-Member of the “Comitato Ordinatore” of a Master in “Machine learning in clinical and surgical research and practice” (Approved).
-“Machine Learning working group” for setting up a new Machine Learning class (2015 -).
-Member of the “Commissione Risorse” as representative for Associate Professors (elected, 2015 -).
-Responsible for “Course Catalogue” of the Doctoral Program in Information Engineering, Dept. of
Information Engineering, (2015 -)
-Responsible for the Erasmus exchange program with the Dept. of Electrical Engineering, Royal Institute of Technology (KTH), Sweden (2013 -).
-Member of the “Collegio dei Docenti” for the Ph.D. program in Information Engineering , Dept. of
Information Engineering, 2007, 2008, 2009, 2014, 2015, 2016
-Member of the “Commissione Assegni e Progetti - Area 11” (2011). Resigned as participant in one of
the projects submitted for possible funding.
-Member of the “Commissione Scientifica”, Dept. of Management and Engineering, 2010
-Member of the “Commissione Biblioteca”, Dept. of Information Engineering, 2004-2006
-Member of the “Commissione Risorse” , Dept. of Management and Engineering and Dept. of Information Engineering (2010-2011)
-Member of the “Esami di Stato” Committee for the habilitation to the profession of Engineer : 2009
(member) and 2011 (adjunct member).
6
♦ INVITED TALKS/PAPERS
INVITED PLENARY ADDRESSES g
-Invited Plenary Speaker at the 15th IFAC Symposium on System Identification (SYSID), Beijing,
China, October 2015 (Link).
-Invited Tutorial Talk: Bayesian Techniques in System Identification, ERNSI Workshop, Cambridge,
09/2010.
-Invited Talk: Recent advances in subspace identification, ERNSI Workshop, Linköping (Sweden),
09/2006.
SELECTED (INVITED) LECTURES
-Invited Speaker at the workshop “Optimization techniques for Inverse Problems III”, Modena, 09/2016
-“Learning Representations for Visual Recognition and their Deep Approximations”, Dept. of Electrical Engineering, Royal Institute of Technology, 06/2016.
-Invited Speaker at the Sino-Italian Workshop on Applied Statistics. Dept. of Statistical Sciences,
University of Padova, February 2016.
-Invited Speaker at the session: “Structure in multivariate time series and high dimensional time series.”, International Conference on Computational and Financial Econometrics (CFE), London 2015
(Organizer: Manfred Deistler).
-Invited Speaker at the Workshop: “Out of the Box: Robustness in High Dimension”, NIPS 2014
-Invited Speaker at the Workshop: “Optimization and dynamical processes in statistical learning and
inverse problems”, Sestri Levante, September 8-12, 2014.
-Invited Tutorial Lecturer: Tutorial Session on “Identification and model (in)validation”, IEEE Conference on Decision and Control, Los Angeles, 12/2014
-Invited Lecturer (4 hours): DISC Ph.D. Summer school on System Identification, Zandvoort, The
Netherlands 06/2014
-Invited Talk: Designing and tuning priors for Bayesian system identification: a classical perspective,
DISI, Università di Genova, 2/2014.
-Invited Talk: LQG control over finite capacity channels: the role of data losses, delays and SNR limitations, Linköping University, 11/2013.
-Invited speaker for the workshop: “Machine Learning for System Identification”, workshop in conjunction to the International Conference on Machine Learning 2013 (organizers: Francesco Dinuzzo,
Abdeslam Boularias, and Lennart Ljung), 06/2013.
-Invited Talk: A Bayesian approach to sparse dynamic network identification, Optimization techniques
for Inverse Problems II, Modena, 09/2012
-Invited Talk: A Bayesian approach to sparse dynamic network identification, Kyoto University, 09/2011
-Invited Talk: Bayesian methods for System Identification and Variable Selection, Linköping University, 05/2011.
-Invited Talk: Gaussian processes for identification of sparse, Large-scale linear systems Dept. of Computer Science, University of California Los Angeles, 08/2010.
-Invited Tutorial Presentation: Some Identification Techniques in Computer Vision, Invited Tutorial
Presentation IFAC ROCOND Symposium, Haifa, Israel, 06/2009.
-Invited Tutorial: Some Identification Techniques in Computer Vision, IEEE CDC, 12/2008.
-Invited Talk: An overview of recent results in subspace identification, Dept. of Mathematics, Royal
Institute of Technology, 05/2007.
-Invited Talk: Recent advances in subspace identification, Dept. of Electrical Engineering, University
of Melbourne, 04/2006.
-Invited Talk: Recent advances in subspace identification Dept. of Chemical Engineering, University of
Texas at Austin, 05/2005.
7
-Tutorial on Subspace Identification (3 hours), Dept. of Chemical Engineering, University of Texas at
Austin, 05/2005
-Invited Talk: Stochastic Realization with Inputs and Subspace Identification, Dept. of Mathematics,
Royal Institute of Technology, 03/2000.
INVITED PAPERS
-Invited to write the chapter on “System Identification Techniques: Convexification, Regularization,
Relaxation” for the Springer Encyclopedia of Systems and Control (2014)
♦ TEACHING
STUDENTS SUPERVISION
- Supervisor of several Laurea Thesis (“Triennale” and “Magistrale/Specialistica” in Mechatronic Engineering and Automation Engineering)
COURSES AS MAIN INSTRUCTOR
-Systems Theory, 2015-2016
-Signals and Systems, 2005-2016
-System Identification and Data Analysis, 2009-2010, 2011-2015
-Subspace System Identification (PhD Course), 2008-2009
-Monte Carlo Methods (PhD Course), 2007-2008
-Subspace System Identification (PhD Course), 2006-2007
-Data Analysis (Basic Probability and Statistics), 2003-2005
-System Identification and Data Analysis, 2001-2004
TEACHING ASSISTANTSHIP
-System Theory, 2003-2004
-System Identification and Data Analysis, 1997-2004
-Calculus II, 1999-2000
-Probability and Stochastic Processes (Washington University St. Louis), 1998-1999
-Electrical Communications, 1997-1998
-Automatic Control, 1997-1998
SHORT COURSES/PHD SCHOOLS/INDUSTRIAL
-Ducth Institute for Systems and Control (DISC) Ph.D. Summer school on System Identification (invited), Zandvoort, The Netherlands June 16-19, 2014
-Identification Techniques, one day course for the company Electrolux, October 2012
-Stochastic Systems, Ph.D. School in Bertinoro, July 2012
-Identification Techniques: (with Prof. G. Picci) for the consulting company S.A.T.E. s.r.l. (most
participants from Ferrari S.p.A.), Venezia, Fall 2001
-Modeling and Control of Mechanical Systems, (with Prof. G. Picci and R. Frezza) for the company
Salvagnini S.p.A. Vicenza, Spring 2001
-Subspace Methods and Linear Algebra in Subspace Identification: QR, SVD, QSVD, Ph.D. School in
Bertinoro, July 1999
8
♦ MAIN SCIENTIFIC ACHIEVEMENTS
(a) Bayesian methods in system identification
A. Chiuso and co-workers (G. Pillonetto and G. De Nicolao), developing on seminal work by G.
Pillonetto and G. De Nicolao, have recently made fundamental contributions to system identification
introducing tools from statistical machine learning (Gaussian processes, kernel methods) and sparsity
enhancing methods.
In particular:
(i) The new Bayesian methods have lead to algorithms for system identification which are very robust
w.r.t. model complexity selection, outperforming classical order estimation techniques for parametric
identification methods. The robustness is inherited by the regularization properties of the Bayesian
approach and allows identification to be performed also for small data set by providing an “optimal”
tradeoff between information content of the data and model complexity.
(ii) His work, combining “sparse Bayesian methods” with the Bayesian approach to system identification has made it possible to tackle simultaneous identification and variable selection for systems
involving large number of variables; this has important implications in modeling sparsely interconnected (networked) dynamical systems.
(iii) He has shown that “sparsity based” methods in a Bayesian framework possess very nice properties
in classical (Fisher) terms, e.g. in terms of their mean squared error properties; the robustness of the
marginal likelihood for hyperparameter tuning has been theoretically studied.
(iv) New Bayesian priors are now being studied to favour “low complexity” (i.e. small McMillan degree) systems, extending ideas recently proposed in machine learning and signal processing using so
called “log-det heuristics”. Preliminary results show substantial improvement in terms of performance
(quality of the estimators and robustness) w.r.t. classical methods.
(v) A new class of “Sparse plus Low Rank” dynamical models, which extends their “static” counterparts, has been recently proposed. For this new class of models, also identification algorithms based
on regularisation have been developed. Interest in this work has been shown by the Econometrics
community: in fact Alessandro Chiuso has been invited to give a talk on the subject at the last Computational and Financial Econometrics (CFE 2015) conference in London (Dec. 2015).
(vi) Exploiting the structure of the marginal likelihood optimisation problem, new fast first order
methods have been developed for its minimisation.
Alessandro Chiuso has been invited to give a Plenary Talk at the recent SYSID on his research in this
area. His plenary lecture has discussed fundamental statistical and probabilistic issues which are behind
regularisation methods, chiefly compound estimation and exchangeability; in doing so, the historical
developments of the subject has been discussed, pointing out also weaknesses of early contributions,
explaining why they had not had significant impact. The material of his plenary lecture has been
recently published as a survey journal paper.
(b) Visual Recognition, Learning Representations and Deep Approximations
The recent success of Deep Neural Network models (Convolutional Neural Networks, CNN) for visual
recognition tasks has contributed to revitalise an area (neural networks) which had been subject of
intense research in the ’80s and ’90s. Our recent work, starting from the basic observation that a
(statistical) representation built for solving visual recognition tasks should posses certain invariance
properties (contrast, translation, rotations and other, possibly more complicated, group actions) has
attempted to show that Convolutional Neural Networks provide a natural approximation framework
encoding such invariances. As an outcome of our analysis, certain “tricks” commonly utilised in visual
descriptors and CNN such as clamping and normalisation of the coefficients, can be explained and
motivated.
9
(c) Contributions to subspace identification
While subspace algorithms have been extensively studied during the 80’s and until the late 90’s, two
important questions were left open for a long time: (i) developing consistent subspace identification of
systems operating in closed loop and (ii) deriving manageable expressions for the asymptotic covariance
of subspace estimates. In this respect A. Chiuso has given major contributions to the solution of these
problems:
(i) A key idea has been the introduction of the predictor-based identification approach, based on
the observation that both the one-step Kalman predictor and the system state space are the same,
irrespective of closed-loop operation. By a stochastic realization approach the state space can be
constructed for systems operating in closed loop. This has led to new efficient algorithms for closed
loop subspace identification, solving a question which had been open for long time.
(ii) A major step forward have been the new expressions for the asymptotic covariances of subspace
methods. These expressions have led to a comparative analysis of most available subspace algorithms,
including open loop and newly developed closed loop methods.
(d) System-theoretic methods in computer vision: real time structure from motion (SFM)
The SFM problem has been one of the main research areas of interest for the computer vision community. In real-time it can be formulated as a nonlinear filtering problem in the Euclidean manifold
SE(3). Traditionally efforts had been concentrated on the geometric reconstruction while less attention was payed to correctly handling uncertainty and “noise” in the inference problem. As a result
a robust algorithm which could work in unstructured environments has long been lacking. The main
contributions to the solution of this problem can be described as follows:
(i) Initially, a first principle analysis of the structure from motion (SFM) problem has led to characterization of well known illusions and ambiguities (such as rubbery-motion and bass-relief ambiguity)
in terms of local minima of a suitable cost function.
(ii) Further, an original analysis (including convergence proofs) of the relevant nonlinear filtering techniques on Lie groups such as SE(n) and SO(n) has been done and then applied to the problem of
motion and structure determination from monocular vision.
(iii) This research has led to the first real time reconstruction system for simultaneous motion and
structure estimation; the system, featured in the July 11 2000 issue of the journal EE Times, consists
in a body of hardware, software and algorithms that allows estimating the three-dimensional motion
and point-wise structure of a moving scene with a single camera in real-time, with no prior information
about the shape and motion of the scene, except for its rigidity. This system represents the first case
of a causal algorithm implemented in real-time, and has been made available for academic use via the
Internet.
(iv) More recently (ICRA 2015), combining vision with inertial sensors and, most importantly, with
tools from robust statistics for handling outliers, it has been possible to demonstrate a system working
in real-time, in a completely unstructured (and non-rigid) environment; see video here.
(e) Modeling and synthesis of complex dynamic images and visual phenomena
A. Chiuso and co-workers have proposed a new approach based on dynamical system modeling and
identification for recognition and synthesis of complex visual phenomena such as textures and gaits.
Among many non-trivial issues to be solved was, above all, the very large dimension of the order of
tens of thousands of the signals involved. These techniques have been proven effective for synthesis
of synthetic scenes and provide high potential for video compression, model based recognition and
classification tasks (see Section on Practical Fallout/Technology Transfer fro more details).
(f) Contribution to estimation and control in networked system
In this area the main contributions are as follows:
(i) An analysis of Kalman-type algorithms for distributed estimation in networked systems has been
performed, analyzing the effect of the network topology. This has evidenced the role of certain synthetic
10
network properties such as the essential spectral radius or the Frobenius norm of the system transition
matrix, depending on the specific topology.
(ii) A systematic comparison of different (suboptimal) strategies for data fusion in the presence of
lossy networks. It has been shown that it is not possible to establish a clear-cut superiority of one
algorithm when estimation accuracy and computational complexity have to be traded. Also the relative
performance may depend on the specific experimental conditions. New performance bounds have been
provided.
(iii) A new synchronization algorithm for a class of interconnected systems has been proposed and
analyzed.
(iv) Recent research activity is targeted to control problems in the presence of non-ideal communication
channels accounting for signal to noise ratio limitations, packet dropouts and delays. Needless to
say, these are the ingredients which need to be traded when jointly designing control and realistic
communication protocols where data rate, decoding delay, and error probability are tightly linked.
While these aspects have been separately (and thoroughly) studied in the literature, a framework
accounting for all these aspects is still lacking; the problem formulation and some promising results
have been recently submitted for publication.
(g) Hybrid Systems
Mainly motivated by modeling problems in the context of Computer Vision, A. Chiuso has also done
research on the analysis of the observability and identifiability properties of switching linear state space
models; in particular new rank conditions that the structural parameters of the model must satisfy in
order for filtering and smoothing algorithms to operate correctly have been found. Also identifiability
of the model parameters has been studied by characterizing the set of models that produce the same
output measurements, giving also conditions under which the true model can be identified.
(h) Applications to Adaptive Optics
Traditional methods for calibrating Adaptive Optics (AO) systems are based on static models and offline identification experiments. This approach may be criticized for two reasons: (i) while static models
are accurate enough for existing systems, it is expected that next generation AO systems will exhibit
non-trivial dynamic behavior, thus requiring dynamic models and (ii) both the deformable mirror and
the sensors exhibit linear characteristics only around the operating conditions. To overcome these
problems, the new closed loop subspace identification techniques discussed above have been successfully applied to modeling state-of-the-art adaptive optics systems (in collaboration with the European
Southern Observatory, Munich). With this new approach dynamical models can be constructed; most
importantly this can be performed on-line and, therefore, around the targeted operating conditions.
Identification of these systems is highly nontrivial as it involves tens (or even hundreds) of inputs and
outputs.
♦ PRACTICAL FALLOUT/TECHNOLOGY TRANSFER
The research activity of Alessandro Chiuso has also been targeted and motivated by several applications,
in particular:
(a) The work in vision has been widely followed worldwide. In particular the paper on Dynamic Textures
has received more that 700 citations. This also thanks to the fact that Dynamic Texture Models are very
simple, easy to understand, and have been widely followed in fields ranging from Computer Graphics
(where they are routinely used in the development of realistic backdrops for games) to Surveillance
(where they are used as null-hypothesis models for outdoor video intrusion detection), to the analysis
of vehicular traffic and human gaits. Before Dynamic Textures, outdoor surveillance video was plagued
11
by false alarms whenever vegetation would trigger motion. Several major manufacturers have adopted
(and attempted to patent) Dynamic Texture or closely related models both in the analysis of Textures
as well as of human motion.
Also the work Structure from Motion and its integration with other sensors (e.g. Inertial) have been
demonstrated in real-time realistic (non-controlled) scenarios; see video here.
(b) The work on subspace identification has been, by and large, motivated by the lack of reliable tools for
identification of “high dimensional” linear systems. One of the main drawbacks of the early literature
was that early subspace methods could not be applied in closed loop, thus preventing their application
in many practical scenarios where control loops could not be opened.
Alessandro Chiuso has developed state-of-the-art closed loop subspace algorithms, which provide the
key enabling methodology for state-of-the-art Advanced Process Control of multivariable systems (see
e.g. Aspen Technologies inc. ).
(c) Identification and prediction methodologies have been developed for, and applied to, several scenarios
such as modelling of laser cutters (in collaboration with Salvagnini S.p.A. where a master student
is now performing his thesis under the supervision of Alessandro Chiuso), adaptive optics systems
(in collaboration with ESO, Munich), energy market modelling, detection of defects in gas-bottles
(in collaboration with the company VideoSystems s.r.l. where a master student has performed his
thesis under the supervision of Alessandro Chiuso). More recently, in collaboration with IIT Genova,
identification and learning methods are being developed and tested in the context of robotics, and in
particular applied to the iCub robot developed by IIT.
(d) Lately, a collaboration has been started with Neuroscientists, Cognitive Psychologists and Biomedical
Engineers for estimation of brain networks models from experimental (fMRI, EEG) data. These models,
which are at the ground of the controllability analysis of such networks, has potentially tremendous
impact in the context of diagnosis and therapeutic treatment of patients (e.g. who suffered cerebral
strokes). This collaboration is leading to a joint research proposal (in preparation).
12
♦ PUBLICATIONS
International Books
[B.1]. A. Chiuso, A. Ferrante and S. Pinzoni (Eds.) (2007), “Modeling Estimation and Control, Festschrift
in Honor of Giorgio Picci on the Occasion of his 65-th Birthday”. Springer Lect. Notes in Control and
Information Sciences.
[B.2]. A. Chiuso, L. Fortuna, M. Frasca, A. Rizzo, L. Schenato, S. Zampieri (Eds.) (2009). “Modeling
Estimation and Control of Networked Complex Systems”. Springer complexity: understanding complex
systems.
International Journals
[J.1]. A. Chiuso, R. Brockett and S. Soatto (2000), “Optimal Structure From Motion: Local Ambiguities
and Global Estimates”. IJCV, International Journal on Computer Vision, 39 (3), pp. 195-228.
[J.2]. A. Chiuso, and G. Picci (2001). “Some Algorithmic aspects of Subspace Identification with Inputs”,
Int. Journal Applied Mathematics and Computer Sciences, Vol. 11, No.1, pp. 55-75.
[J.3]. A. Chiuso, P. Favaro, H. Jin and S. Soatto (2002). “Structure from Motion Causally Integrated over
Time”. IEEE Trans. on Pattern Analysis and Machine Intelligence 24(4),pp. 509-522.
[J.4]. G. Doretto, A. Chiuso, Y.N. Wu, S. Soatto (2003), “Dynamic Textures”. Int. Journal of Computer
Vision, 51(2), 2003, pp. 91-109.
[J.5]. A. Chiuso and G. Picci (2004), “Asymptotic Variance of Subspace Estimates”. Journal of Econometrics, 118(1-2), pp. 257–291.
[J.6]. A. Chiuso, G. Picci (2004), “On the Ill-conditioning of subspace identification with inputs”. Automatica, 40(4), pp. 575–589, Regular paper.
[J.7]. A. Chiuso, G. Picci (2004), “Numerical conditioning and asymptotic variance of subspace estimates”.
Automatica, 40(4), pp. 677-683, Brief Paper.
[J.8]. A. Chiuso, G. Picci (2004), “Subspace identification by data orthogonalization and model decoupling”.
Automatica, 40(10), pp. 1689–1703, Regular paper.
[J.9]. A. Chiuso, G. Picci (2004), “Asymptotic Variance of Subspace methods by data orthogonalization
and model decoupling: a comparative study.”. Automatica, 40(10), pp. 1705–1717 , Regular paper.
[J.10]. A. Chiuso, G. Picci (2005), “Consistency Analysis of some Closed-Loop Subspace Identification Methods”. Automatica, Special Issue on System Identification, March 2005 (41(3)), pp. 377–391, Special
issue (regular) paper.
[J.11]. A. Chiuso (2006), “Asymptotic Variance of Closed-Loop Subspace Identification Methods”. IEEE
Trans. on Automatic Control, 51(8), pp. 1299-1314 (regular paper).
[J.12]. A. Chiuso (2007), “The role of Vector AutoRegressive modeling in Predictor Based Subspace Identification”. Automatica 43(6), pp. 1034-1048 (regular paper).
[J.13]. A. Chiuso (2007) “On the relation between CCA and predictor-based subspace identification” IEEE
Transactions on Automatic Control, 52(10), pp. 1795-1812 (regular paper).
[J.14]. A. Bissacco, A. Chiuso and S. Soatto (2007), “Classification and Recognition of Dynamical Models:
The Role of Phase, Independent Components, Kernels and Optimal Transport”. IEEE Trans. on
Pattern Analysis and Machine Intelligence, 29(11), pp. 1958-1972 (regular paper).
13
[J.15]. R. Carli, A. Chiuso, L. Schenato and S. Zampieri (2008), “Distributed Kalman filtering using consensus strategies” IEEE Journal on Selected Areas in Communications 26(4), pp. 622-634 (regular
paper).
[J.16]. A. Chiuso, G. Picci, S. Soatto (2008), “Wide-sense Estimation on the Special Orthogonal Group”
Communications in Information and Systems, 8(3), pp. 185-200.
[J.17]. G. Pillonetto, A. Chiuso (2009), “Fast computation of smoothing splines subject to equality constraints” Automatica, 45(12), pp. 2842-2849 (brief paper).
[J.18]. A. Chiuso, R. Muradore, E. Marchetti (2010), “Dynamic Calibration of Adaptive Optics Systems: A
System Identification Approach” IEEE Transactions on Control Systems Technology, 18(3), pp. 705
–713, (brief paper).
[J.19]. A. Chiuso (2010), “On the asymptotic properties of closed loop CCA-type Subspace Algorithms:
equivalence results and choice of the future horizon” IEEE Transactions on Automatic Control, 55(3),
pp. 634 -649 (regular paper).
[J.20]. G. Pillonetto, A. Chiuso, G. De Nicolao (2011), “Prediction error identification of linear systems: a
Gaussian regression approach” Automatica, 47(2), pp. 291-305, (regular paper).
[J.21]. A. Chiuso, L. Schenato (2011), “Information fusion strategies and performance bounds in packet-drop
networks” Automatica, 47(7), pp. 1304-1316, (regular paper).
[J.22]. R. Carli, A. Chiuso, L. Schenato and S. Zampieri (2011), “Distributed synchronization of noisy
non-identical double integrators” IEEE Transactions on Automatic Control, 56(5), pp. 1146-1152,
(technical note).
[J.23]. A. Chiuso, F. Fagnani, L. Schenato, S. Zampieri (2011), “Gossip algorithms for simultaneous distributed estimation and classification in sensor networks”, IEEE Journal of Selected Topics in Signal
Processing, 5(4), pp. 691 -706, (regular paper).
[J.24]. G. Pillonetto, M. H. Quang, A. Chiuso (2011), “A new kernel-based approach for nonlinear system
identification”, IEEE Trans. on Automatic Control, 56(12) pp. 2825 – 2840 (regular paper).
[J.25]. A. Chiuso and G. Pillonetto (2012), “A Bayesian approach to sparse dynamic network identification”,
Automatica 48(2), pp. 1553-1565. (regular paper)
[J.26]. A. Aravkin, J. Burke, A. Chiuso and G. Pillonetto (2014) , “Convex vs non-convex estimators for
regression and sparse estimation: the mean squared error properties of ARD and GLasso”, Journal of
Machine Learning Research, 15(Jan):217–252, 2014.
[J.27]. T. Chen, M.S. Andersen, L. Ljung, A. Chiuso and G. Pillonetto (2014) “System identification via
sparse multiple kernel-based regularization using sequential convex optimization techniques”, IEEE
Transactions on Automatic Control, 59(11) pp. 2933 - 2945 (regular paper).
[J.28]. A. Chiuso (2014), “System Identification Techniques: Convexification, Regularization, Relaxation”,
Springer Encyclopedia of Systems and Control. Editors J. Baillieul and T. Samad.
[J.29]. S. Dey, A. Chiuso, L. Schenato (2014). Remote estimation with noisy measurements subject to packet
loss and quantization noise. IEEE Transactions on Control of Network Systems, 1(3), pp. 204 – 217
(Regular paper).
[J.30]. A. Chiuso, N. Laurenti, L. Schenato, A. Zanella (2014). LQG-like control over communication channels for scalar systems: the role of data losses, delays and SNR limitations. Automatica, 50(12), pp.
3155–3163 (Brief paper)
14
[J.31]. S. Bonettini, A. Chiuso, M. Prato (2015). A scaled gradient projection methods for Bayesian Learning
in Dynamical Systems. SIAM J. Scientific Computing, 37(3), A1297–A1318.
[J.32]. G. Pillonetto, A. Chiuso (2015). Tuning complexity in regularized kernel-based regression and linear
system identification: the robustness of the marginal likelihood estimator. Automatica, 58(8), pp.
106–117 (Regular paper)
[J.33]. T. Chen, T. Ardeshiri, F.P. Carli, A. Chiuso, L. Ljung, G. Pillonetto (2016). Maximum entropy
properties of discrete-time first-order stable spline kernel. Automatica, 66(4), pp. 34–38 (Technical
Communique).
[J.34]. G. Pillonetto, T. Chen, A. Chiuso, G. De Nicolao, L, Ljung (2016). Regularized linear system identification using atomic, nuclear and kernel-based norms: the role of the stability constraint. Automatica
69(7), pp. 137–149 (Regular paper)
[J.35]. A. Chiuso (2016). Regularization and Bayesian Learning in Dynamical Systems: Past, Present and
Future. Annual Reviews in Control, in press (available online).
Book Chapters
[BC.1]. A. Chiuso and G. Picci (1998) “Visual tracking of points as estimation on the unit sphere”, in The
Confluence of Vision and Control, D. Kriegman, W. Hager, S. Morse (Eds.), Springer LNCIS 1998,
pp. 91-104.
[BC.2]. A. Chiuso, H. Jin, P. Favaro and S. Soatto (2000). “MFm” : 3-D Motion and Structure from 2-D
Motion Causally Integrated Over Time: Implementation. In Computer Vision -ECCV 2000, D. Vernon
ed., Lect. notes in Computer Science 1843, pp. 734-750.
[BC.3]. A. Chiuso, G. Picci (2002), “Geometry of Oblique Splitting, Minimality and Hankel Operators”. in
Directions in Mathematical Systems Theory and Optimization, A. Rantzer and C. Byrnes eds. Springer
Lect. Notes in Control and information Sciences, 286, pp. 85-124 (2002).
[BC.4]. R. Vidal, A. Chiuso, S. Soatto and S. Sastry (2003) “Observability Linear Hybrid Systems” In Hybrid
Systems: Computation and Control, Lecture Notes in Computer Science 2623, pp. 526-539.
[BC.5]. Saisan, P., A. Bissacco, A. Chiuso and S. Soatto (2004). “Modeling and synthesis of facial motion
driven by speech”. In: Computer Vision -ECCV 2004, T. Pajdla and J. Matas eds., Lect. notes in
Computer Science 3023, pp. 453-467.
[BC.6]. A. Masiero, A. Chiuso (2006) “Non linear temporal textures synthesis: a Monte Carlo approach” in
Computer Vision -ECCV 2006, Part II, A. Leonardis, H. Bischof and A. Prinz (eds.), Lect. Notes in
Computer Science 3952, pp. 283-294.
International Conference Proceedings
[C.1]. A. Chiuso and G. Picci (1998) “A wide-sense estimation theory on the unit sphere”. In Proc. of
IEEE 37th Conference on Decision and Control, Tampa, Florida, December 1998, pp. 3743-9754.
[C.2]. A. Chiuso, and G. Picci, (1999), “Subspace Identification by orthogonal decomposition”, Proc. 14th
IFAC World Congress , Pechino, Cina, July, 1999, Volume H, pp. 241-246.
[C.3]. H. Kawauchi, A. Chiuso, T. Katayama, G.Picci. (1999), “Comparison of Two Subspace Identification
Methods for Combined Deterministic - Stochastic Systems”. In Proc. of The 31st ISCIE International
Symposium on Stochastic Systems Theory and its Applications, Yokohama, Japan, 1999, pp. 7-12.
15
[C.4]. A. Chiuso and S. Soatto (2000), “3-D Motion and Structure Causally Integrated Over Time: Analysis”, Tutorial lecture presented at IEEE Intl. Conf. on Robotics and Automation, San Francisco, April
2000. Preliminary version registered as ESSRL Technical Report 99-001
[C.5]. A. Chiuso and G. Picci (2000), “Error Analysis of Certain Subspace Methods”. In Proc. of IFAC
International Symposium on System Identification, Santa Barbara, June 2000, pp. 85-90.
[C.6]. A. Chiuso and G. Picci (2000), “Probing Inputs for Subspace Identification”. In Proc. of IEEE International Conference on Decision and Control CDC 2000, Sydney, Australia. (Invited paper number
INV0201.)
[C.7]. A. Chiuso and S. Soatto (2000), “Monte Carlo filtering on Lie Groups”, In Proc. of IEEE International
Conference on Decision and Control CDC 2000, Sydney, Australia. (Regular paper number REG1407.)
[C.8]. A. Chiuso and G. Picci (2001), “Asymptotic Variance of Subspace Estimates”. Proc. of IEEE
International Conference on Decision and Control , Orlando, Florida,USA, December 2001.
[C.9]. A. Bissacco, A. Chiuso, Y. Ma and S. Soatto (2001) “Recognition of human gaits.” In Proc. of the
IEEE Intl. Conf. on Comp. Vision and Patt. Recog., Hawaii, Dec. 2001.
[C.10]. R. Vidal, A. Chiuso and S. Soatto (2002) “Observability and Identifiability of Jump Linear Systems”
In Proc. of the IEEE Conf. on Decision and Control, Las Vegas, USA, 2002.
[C.11]. G. Gennari, A. Chiuso, F. Cuzzolin, R. Frezza (2002) “Integrating dynamic and probabilistic shape
information for data association and tracking” In Proc. of the IEEE Conf. on Decision and Control,
Las Vegas, USA, 2002.
[C.12]. A. Chiuso, G. Picci (2003) “Asymptotic Variance of Subspace Methods by data Orthogonalization
and Model Decoupling”, in Proc. of the IFAC Int. Symposium on System Identification (SYSID),
Rotterdam, 2003.
[C.13]. A. Chiuso, G. Picci (2003) “Subspace Identification of Random Processes with Feedback”, in Proc.
of the IFAC Int. Symposium on System Identification (SYSID), Rotterdam, 2003.
[C.14]. S. Soatto, A. Chiuso (2003) “Snippets of System Identification in Computer Vision”, in Proc. of
the IFAC Int. Symposium on System Identification (SYSID), Semi-Plenary lecture given by S. Soatto,
Rotterdam, 2003.
[C.15]. A. Chiuso (2004), “Asymptotic Variance of a Certain Closed-Loop Subspace Identification Method”
Proc. of the 43rd IEEE Conf. on Decision and Control, 2004.
[C.16]. A. Chiuso, G. Picci (2004), “Consistency Analysis of Certain Closed-Loop Subspace Identification
Methods” Proc. of the 43rd IEEE Conf. on Decision and Control, 2004.
[C.17]. G. Gennari, A. Chiuso, F. Cuzzolin, R. Frezza (2004), “Integration of shape constraints in data
association filters” Proc. of the 43rd IEEE Conf. on Decision and Control, 2004.
[C.18]. A. Chiuso, G. Picci (2005), “Prediction Error vs. Subspace methods in open and closed loop identification”. To appear in the Proc. of the 16th IFAC World Congress.
[C.19]. A. Chiuso (2005), “On the relation between CCA and predictor based subspace identification”. IEEE
Conf. on Dec. and Control 2005
[C.20]. A. Chiuso, A. Ferrante, G. Picci (2005) “Reciprocal realization and modeling of textured images”
IEEE Conf. on Dec. and Control 2005
[C.21]. R. Frezza, A. Chiuso (2005) “Learning and exploiting invariants for multi-target tracking and data
association” IEEE Conf. on Dec. and Control 2005.
16
[C.22]. A. Chiuso (2006) “Asymptotic Equivalence of Certain Closed-Loop Subspace Identification Methods”
IFAC SYSID 2006, Newcastle, Australia (March 2006).
[C.23]. A. Chiuso, G. Picci (2006) “Estimating the Asymptotic Variance of Closed-Loop Subspace Estimators” IFAC SYSID 2006, Newcastle, Australia (March 2006).
[C.24]. A. Chiuso, A Ferrante and G. Picci (2006) “Realization of Reciprocal processes and applications to
computer vision” Presented at the Int. Symp. Mathematical Theory of Network and Systems, July
2006.
[C.25]. A. Chiuso (2006) “The role of Vector AutoRegressive Modeling in Subspace Identification”, Proc. of
the IEEE Conf. on Dec. and Control, San Diego, Dec. 2006.
[C.26]. A. Chiuso (2007), “Some insights on the choice of the future horizon in CCA-type subspace algorithm”
Proc. of the American Control Conference, ACC’07
[C.27]. A. Chiuso, R. Muradore and E. Fedrigo (2007), “Adaptive Optics Systems: a challenge for closed-loop
subspace identification” Proc. of the American Control Conference, ACC’07
[C.28]. R. Vidal, S. Soatto, A. Chiuso (2007) “Applications of Hybrid System Identification in Computer
Vision” ECC 2007 (Invited Minitutorial Paper), Kos, Greece.
[C.29]. R. Carli, A. Chiuso, L. Schenato and S. Zampieri (2007) “Consensus algorithm design for distributed
estimation” 3 rd International Workshop on Networked Control Systems: Tolerant to Faults, Nancy,
France, Jun. 2007
[C.30]. R. Carli, A. Chiuso, L. Schenato and S. Zampieri (2007), “Distributed Kalman filtering using consensus strategies” Proc. of IEEE Conf. on Decision and Control, New Orleans, USA, Dec. 2007
[C.31]. A. Chiuso (2008) “A note on estimation using quantized data” 17th IFAC World Congress, Seoul
(Korea), July 2008.
[C.32]. R. Carli, A. Chiuso, L. Schenato and S. Zampieri (2008), “A PI Consensus Controller for Networked
Clocks Synchronization” 17th IFAC World Congress, Seoul (Korea), July 2008.
[C.33]. E. Toffoli, G. Baldan, G. Albertin, L. Schenato, A. Chiuso, A. Beghi (2008), “Thermodynamic
Identification of Buildings using Wireless Sensor Networks” 17th IFAC World Congress, Seoul (Korea),
July 2008.
[C.34]. A. Agnoli, A. Chiuso, P. D’Errico, A. Pegoraro, L. Schenato (2008), “Sensor fusion and estimation strategies for data traffic reduction in rooted wireless sensor networks” International Symp. on
Communication, Control and Signal Processing (ISCCSP 2008), Malta, March 2008.
[C.35]. A. Chiuso, R. Muradore, E. Marchetti (2008), “Dynamic Calibration of Adaptive Optics Systems: A
System Identification Approach” 2008 IEEE Conf. on Dec. and Control.
[C.36]. A. Chiuso, L. Schenato (2008), “Information fusion strategies from distributed filters in packet-drop
networks” 2008 IEEE Conf. on Dec. and Control.
[C.37]. G. Pillonetto, A. Chiuso, G. De Nicolao (2008), “Predictor Estimation via Gaussian regression” 2008
IEEE Conf. on Dec. and Control.
[C.38]. A. Chiuso, G. Pillonetto, G. De Nicolao (2008), “Subspace Identification using predictor estimation
via Gaussian regression” 2008 IEEE Conf. on Dec. and Control.
[C.39]. A. Chiuso, G. Picci (2008), “Identification techniques in Computer Vision” 2008 IEEE Conf. on Dec.
and Control (Invited Tutorial Paper)
17
[C.40]. G. Pillonetto, A. Chiuso (2009), “Gaussian Processess for Wiener-Hammerstein System Identification” IFAC SYSID
[C.41]. M. H. Quang, G. Pillonetto, A. Chiuso (2009), “Nonlinear system identification via Gaussian regression and mixtures of kernels” IFAC SYSID 2009
[C.42]. A. Chiuso, L. Schenato (2009), “Performance bounds for information fusion strategies in packet-drop
networks” ECC 2009
[C.43]. G. Pillonetto, A. Chiuso (2009), “A Bayesian learning approach to linear system identification with
missing data” CDC 2009.
[C.44]. G. Pillonetto, A. Chiuso, G. De Nicolao (2010), “Regularized estimation of sums of exponentials in
spaces generated by stable spline kernels” ACC 2010.
[C.45]. A. Chiuso, G. Pillonetto (2010), “Nonparametric sparse estimators for identification of large scale
linear systems”, IEEE CDC 2010.
[C.46]. A. Chiuso, R. Muradore, E. Aller-Carpentier (2010), “Sparse Calibration of an Extreme Adaptive
Optics System”, IEEE CDC 2010.
[C.47]. A. Chiuso, G. Pillonetto (2010), “Learning sparse dynamic linear systems using stable spline kernels
and exponential hyperpriors”, NIPS 2010.
[C.48]. A. Chiuso, F. Fagnani, L. Schenato, S. Zampieri (2010). “Simultaneous distributed estimation and
classification in sensor networks”. IFAC Workshop on Distributed Estimation and Control in Networked
Systems (NecSys’10).
[C.49]. S. Soatto, A. Chiuso (2011), “Controlled Recognition Bounds for Scaling and Occlusion Channels”,
Data Compression Conference, 2011
[C.50]. A. Chiuso, F. Fagnani, L. Schenato, S. Zampieri (2011), “Gossip algorithms for distributed ranking”,
ACC 2011.
[C.51]. A. Aravkin, J. Burke, A. Chiuso and G. Pillonetto (2011), “Convex vs nonconvex approaches for
sparse estimation: GLasso, Multiple Kernel Learning and Hyperparameter GLasso”, IEEE CDC 2011.
[C.52]. F.P. Carli, A. Chiuso, G. Pillonetto (2012), “Efficient algorithms for large scale linear system identification using stable spline estimators”, IFAC SYSID 2012.
[C.53]. A. Aravkin, J. Burke, A. Chiuso and G. Pillonetto (2012), “On the estimation of hyperparameters for
Empirical Bayes estimators: Maximum Marginal Likelihood vs Minimum MSE”, IFAC SYSID 2012.
[C.54]. A. Aravkin, J. Burke, A. Chiuso and G. Pillonetto (2012), “On the MSE Properties of Empirical
Bayes Methods for Sparse Estimation”, IFAC SYSID 2012.
[C.55]. F. Carli, T. Chen, A. Chiuso, L. Ljung, G. Pillonetto. “On the estimation of hyperparameters for
Bayesian system identification with exponential kernels”, IEEE CDC 2012.
[C.56]. T. Chen, M. Andersen, L. Ljung, A. Chiuso, F. Carli, G. Pillonetto. “Sparse multiple kernels for
impulse response estimation with majorization minimization algorithms”, IEEE CDC 2012.
[C.57]. G. Georgiadis, A. Ravichandran, S. Soatto and A. Chiuso. “Encoding Scene Structures for Video
Compression” SPIE, 2012
[C.58]. V. Karasev, A. Chiuso and S. Soatto. “Controlled Recognition Bounds for Visual Learning and
Exploration”, NIPS 2012
18
[C.59]. A. Chiuso, N. Laurenti, L. Schenato, A. Zanella. “LQG cheap control subject to packet loss and SNR
limitations”, ECC 2013
[C.60]. V. Karasev, A. Chiuso and S. Soatto. “Controlled Recognition Bounds for Visual Learning and
Exploration”, IEEE Information Theory Workshop 2013
[C.61]. G. Georgiadis, A. Chiuso and S. Soatto. “Texture Compression” Data Compression Conference, 2013
[C.62]. A. Chiuso, T. Chen, L. Ljung, G. Pillonetto. “Regularization strategies for nonparametric system
identification”, CDC 2013.
[C.63]. T. Chen, A. Chiuso, L. Ljung, G. Pillonetto. “Rank-1 kernels for regularized system identification”,
CDC 2013.
[C.64]. A. Chiuso, N. Laurenti, L. Schenato, A. Zanella. “LQG control over SNR-limited lossy channels with
delay”, CDC 2013.
[C.65]. S. Dey, A. Chiuso, L. Schenato, “Remote estimation subject to packet loss and quantization noise”,
CDC 2013.
[C.66]. A. Chiuso, G. Pillonetto. “Bayesian and nonparametric methods for system identification and model
selection.” Proc. of ECC 2014.
[C.67]. G. Pillonetto, A. Chiuso. “Tuning complexity in kernel-based linear system identification: the robustness of the marginal likelihood estimator”. Proc. of ECC 2014.
[C.68]. T. Chen, M. Andersen, A. Chiuso, G. Pillonetto, L. Ljung. “Anomaly detection in homogenous
populations: a sparse multiple kernel-based regularization method”. IEEE CDC 2014
[C.69]. A. Chiuso, T. Chen, L. Ljung, G. Pillonetto. “On the design of Multiple Kernels for nonparametric
linear system identification” IEEE CDC 2014
[C.70]. G. Prando, A. Chiuso, G. Pillonetto. “Bayesian and regularization approaches to multivariable linear
system identification: the role of rank penalties ” IEEE CDC 2014, Invited Tutorial Paper
[C.71]. D. Romeres, A. Chiuso, G. Pillonetto. “Identification of stable models via nonparametric prediction
error methods”. Proc. of the European Control Conference, 2015
[C.72]. S. Dey, A. Chiuso, L. Schenato. “Linear Encoder-Decoder-Controller Design over Channels with
Packet Loss and Quantization Noise” . European Control Conference ECC15, 2015
[C.73]. K. Tsotsos, A. Chiuso, S. Soatto. “Robust Inference for Visual-Inertial Sensor Fusion” . ICRA 2015
[C.74]. G. Georgiadis, A. Chiuso, S. Soatto. “Texture Representations for Image and Video Synthesis”. Proc.
of CVPR 2015.
[C.75]. G. Prando, A. Chiuso, G. Pillonetto. “The role of rank penalties in linear system identification”.
Proc. of SYSID 2015.
[C.76]. A. Chiuso, “Regularization and Bayesian Learning in Dynamical Systems: Past, Present and Future”.
Plenary Lecture, SYSID 2015.
[C.77]. T. Chen, G. Pillonetto, A. Chiuso, L. Ljung “Spectral analysis of the DC kernel for regularized system
identification ”. IEEE CDC 2015
[C.78]. M. Zorzi, A. Chiuso. “A Bayesian Approach to Sparse plus Low rank Network Identification ”. IEEE
CDC 2015
19
[C.79]. G. Prando, A. Chiuso. “Model reduction for linear Bayesian System Identification ”. IEEE CDC
2015
[C.80]. S. Soatto, A. Chiuso. “Modeling Visual Representations: Defining Properties and Deep Approximations”. International Conference on Learning Representation (ICLR), 2016
[C.81]. D. Romeres, G. Prando, G. Pillonetto, A. Chiuso (2016). “On-line Bayesian System Identification”.
ECC 2016, to appear.
[C.82]. G. Prando, D. Romeres, G. Pillonetto, A. Chiuso (2016) “Classical vs. Bayesian methods for linear
system identification: point estimators and confidence sets ”. ECC 2016, to appear.
[C.83]. G. Rallo, S. Formentin, A. Chiuso, S.M. Savaresi (2016). “Virtual Reference Feedback Tuning with
Bayesian regularization”. ECC 2016, to appear.
Technical Reports
[R.1]. A. Chiuso, H. Jin, P. Favaro and S. Soatto (1999), “Application of Extended Kalman Filtering to the
Reconstruction of 3-D Shape from Visual Motion”. ESSRL Technical Report 99-001.
[R.2]. A. Chiuso (2000) “A Matlab Toolbox for Subspace Identification”, Università di Padova.
[R.3]. A. Chiuso, G. Picci (2003),“Canonical Correlation Analysis of Linear Stochastic Systems with Inputs”,
Mittag-Leffler Technical Report No. 45, 2002/2003, Spring ISSN 1103-467X, ISN IML-R-45-02/03–
SE+spring.
[R.4]. A. Chiuso (2005), “Asymptotic Variance of Closed-Loop Subspace Identification Methods”. Università di Padova -Complete version (contains all proofs which have been omitted due to space limitation
of the paper appeared in the IEEE Transactions on Automatic Control (July 2006)).
[R.5]. F. Parise, L. Dal Col, A. Chiuso, N. Laurenti, L. Schenato, A. Zanella (2013). “Impact of a realistic
transmission channel on the performance of control systems”, Università di Padova.
Submitted Papers
[S.1]. G. Prando, A. Chiuso, G. Pillonetto (2015). “Maximum Entropy Vector Kernels for MIMO system
identification”. Revised version submitted to Automatica.
[S.2]. M. Zorzi, A. Chiuso (2015). “Sparse plus Low rank Network Identification: A Nonparamteric Approach”. Revised version submitted to Automatica.
[S.3]. S. Dey, A. Chiuso, L. Schenato (2016). “Feedback Control over lossy SNR-limited channels: linear
encoder-decoder-controller design”. Submitted to IEEE Transactions on Automatic Control (Technical
Note).
[S.4]. G. Prando, D. Romeres, A. Chiuso. On-line Identification of Time-Varying Systems: a Bayesian
approach. IEEE CDC 2016 - submitted
[S.5]. D. Romeres, M. Zorzi, A. Chiuso. Online semi-parametric learning for inverse dynamics modeling.
IEEE CDC 2016 - submitted
[S.6]. T. Chen, G. Pillonetto, A. Chiuso, L. Ljung. DC kernel - a stable generalized first order spline kernel.
IEEE CDC 2016 - submitted
Book in preparation
[BP.1]. T. Chen, A. Chiuso, G. De Nicolao, L. Ljung, G. Pillonetto. Regularized approaches to System
Identification. In preparation
20