- NumPI - Mailing list — Dipartimento di Matematica, UniPI

Annuncio di seminario
by Leonardo Robol 12 Dec '18

12 Dec '18

Buongiorno, allego l'annuncio del seminario per la settimana prossima, che sarà tenuto da Giulio Masetti (ISTI-CNR). Colgo l'occasione per ricordarvi l'appuntamento di domani, per il seminario di Stefano Massei, alle 10:00 in Aula Seminari Ovest, Informatica. -- Leonardo Robol. ==== Speaker: Giulio Masetti Affiliation: ISTI-CNR Time: Thursday, 13 December 2018, h. 10:00 Place: Aula Seminari, Dipartimento di Matematica Title: Computing dependability-oriented measures on Markov chains by means of matrix functions In the context of computing and communication system, dependability is defined as the ability to deliver service that can justifiably be trusted. Dependability is then an umbrella term that encompasses several attributes, such as: availability, reliability, safety, integrity, maintainability, and confidentiality. Design and analysis of models capable to express functional specifications and behavioural aspects is an important part of the system dependability justification process. In particular, Markov chains reflect important characteristics of computing systems such as discrete states and memoryless property, and then are often employed to model complex systems. Starting from the model Markovian process, other processes can be defined to express gains or losses according to which state the Markov process is in, and then all the mentioned dependability attributes can be evaluated as measures on these processes. More and more complex systems are designed, and then new techniques are required to tackle the evaluation of dependability-oriented measures on large models. In a world where systems comprise hundreds or thousands of interconnected components, justifying dependability is an ever increasing challenge. The main contribution presented in this talk is, focusing on Continuous Time Markov Chains, recasting dependability-oriented measures as the evaluation of a bilinear form where the matrix is indeed a matrix function of the infinitesimal generator matrix characterizing the Markovian stochastic process. In particular, chains with absorbing states, relevant to evaluate system reliability and connected attributes, represent one of the main challenges the modeling community has to deal with, and -- we think -- one of the cases where applying results from the matrix functions body of knowledge can have the greater impact.

1 1

Annuncio di seminario
by Leonardo Robol 29 Nov '18

29 Nov '18

Buongiorno, allego l'annuncio del seminario per la settimana prossima, che sarà tenuto da Stefano Massei (EPF Lausanne). L'appuntamento è giovedì 6 dicembre alle 10:00 presso il Dipartimento di Informatica, Sala Seminari Ovest. -- Leonardo Robol. ==== Speaker: Stefano Massei Affiliation: EPF Lausanne Time: Thursday, 6 December 2018, h. 10:00 Place: Sala Seminari Ovest, Dipartimento di Informatica. Title: A low-rank technique for computing the quasi stationary distribution of Galton-Watson processes Branching processes describe the dynamics of a population of individuals which reproduce and die independently, according to some specific probability distributions. More precisely, we assume that any individual has a unit lifetime, at the end of which it might give birth to one or more offsprings simultaneously. This is encoded into the probability generating function P(z):=\sum p_j z^j where p_j is the probability of generating j individuals. These kind of processes are known in the literature as Galton-Watson processes. We consider populations that are certain to become extinct, yet appear to be stationary over any reasonable time scale. More precisely, we are interested in characterizing the quasi-stationary distribution of the process, i.e., the asymptotic distribution of the population size, conditional on its survival. Yaglom proved that if m:=P'(1)<1 then the quasi stationary distribution exists and its probability generating function G(z):=\sum g_j z^j solves the Schroeder functional equation G(P(z))= mG(z)+1-m, z in [0,1]. (*) We study the link between the regularity of P(z) and that of G(z) and we propose a strategy for solving (*) in the case where P(z) and G(z) are analytic on a disc of radius r>1. We see that the discretization of (*) leads to a numerical method that is capable to find arbitrary accurate approximations of the coefficients of G(z). Moreover, we point out the (numerical) low-rank structure that appears in the discretized problem, and we show how to exploit it in the proposed procedure.

1 0

Annuncio di seminario
by Leonardo Robol 26 Nov '18

26 Nov '18

Buongiorno, allego l'annuncio del seminario di Marcus Carlsson, programmato per la prossima settimana. Il seminario si terrà in SNS, e sarà di martedì e non di giovedì. -- Leonardo Robol. ==== Speaker: Marcus Carlsson Affiliation: Lund University Time: Tuesday, 27 November 2018, h. 11:00 Place: Aula Bianchi scienze, SNS Title: Perturbation theory for the spectral decomposition of matrices Consider a Hermitian matrix and a small perturbation. We will consider how the eigenvalues and eigenvectors change with the perturbation. Typical theorems concerning this either consider a perturbation along a line or provide general estimates valid for all perturbations (such as Weyl’s inequalities). We first give an overview of the field and then provide new results which are global in the sense that the perturbation is not restricted to a line, but local in the sense that the estimates are valid modulo certain ordo terms. We apply this to the functional calculus of the matrix square root and matrix modulus. In the kernel-free case the corresponding result follows from the so called Daleskii-Krein theorem.

1 1

Avviso di seminario (e reminder)
by Leonardo Robol 22 Nov '18

22 Nov '18

Buongiorno, allego l'annuncio del secondo seminario nella serie NumPI, e con l'occasione vi ricordo l'appuntamento di domani alle 10:00 per quello di Giovanni Barbarino (Sala Seminari Est, Dipartimento di Informatica). ==== Speaker: Jie Meng Affiliation: Pusan National University Time: Thursday, 22 November 2018, h. 10:00 Place: Sala Seminari, Dipartimento di Matematica, Università di Pisa Title: On the positive definite solution of the nonlinear matrix equation X^p=A+M^T(X#B)M The nonlinear matrix equation X^p = A + M^T (X # B)M is considered, where p ≥ 1 is a positive integer, M is an n × n nonsingular matrix, A is a positive semidefinite matrix and B is a positive definite matrix. We call C # D the geometric mean of positive definite matrices C and D. Some properties of the positive definite solution of the nonlinear matrix equation is investigated, including the existence and uniqueness of the solution, a lower and an upper bound. Iteration method for finding the numerical solution is proposed. Perturbation bounds with respect to small perturbations on the coefficient matrices are obtained. -- Leonardo Robol

1 1

Fwd: [dynlist] Dyn.Sys - C. Wormell @ Centro De Giorgi 14, 00h - 14th Nov.
by Federico Poloni 13 Nov '18

13 Nov '18

Mi hanno segnalato questo seminario domani al Centro De Giorgi --- non è uno della nostra serie, ma è comunque un argomento che può essere interessante per alcuni. Cordialmente, >> Begin forwarded message: >> >> *From: *Paolo Giulietti <paologiulietti.math(a)gmail.com >> <mailto:paologiulietti.math@gmail.com>> >> *Subject: **[dynlist] Dyn.Sys - C. Wormell @ Centro De Giorgi 14,00h - >> 14th Nov.* >> *Date: *November 12, 2018 at 5:35:20 PM GMT+1 >> *To: *dynlist-pisa(a)googlegroups.com <mailto:dynlist-pisa@googlegroups.com> >> >> Dear Dynamical Friends, >> >> this week, on Wednesday 14th of November at 14:00h at the Centro de >> Giorgi, ground floor, C. Wormell from the University of Sydney will >> talk about "Chebyshev Galerkin methods for transfer operators in >> uniformly-expanding dynamics" (abstract below). >> >> Let me Remind you that at 15:00h at Palazzo della Carovana there will >> be the monthly science colloquium of the SNS. >> >> Moreover, if you are using the GoogleCalendar, check it out to see >> latest additions to the January/February schedule! >> >> Cheers! >> P&J >> >> ----------------------------------------------------------------------------------------------------------------------------------- >> >> Abstract: Full-branched expanding maps, a class of simple chaotic >> systems, are >> commonly used as models for chaotic dynamics, as well as being of >> number-theoretical interest. However, nonspecialised numerical methods >> to calculate long-time statistical quantities such as invariant measures >> have a poor trade-off between computational effort and accuracy. >> >> In this talk I will present a rigorous method to calculate statistics of >> these maps by discretising the transfer operator in a Chebyshev >> polynomial basis. This is a highly efficient discretisation: I will show >> that, for analytic maps, numerical estimates obtained with this method >> converge exponentially quickly in the order of the discretisation. >> Finally, I will discuss some implementations of these algorithms for >> exploration and rigorous proof: in particular, in only a fraction of a >> second on a personal computer one can produce estimates of most >> statistical properties accurate to 14 decimal places. >> >> >> -- >> ------------------------------------------------------------------------- >> "NO! Try not. Do, or do not. There is no try." >> >> "Much to learn, you still have." >> >> \\Yoda\\ <smb://Yoda//> >> >> -- >> Hai ricevuto questo messaggio perché sei iscritto al gruppo >> "dynlist-pisa" di Google Gruppi. >> Per annullare l'iscrizione a questo gruppo e non ricevere più le sue >> email, invia un'email a dynlist-pisa+unsubscribe(a)googlegroups.com >> <mailto:dynlist-pisa+unsubscribe@googlegroups.com>. >> Per postare in questo gruppo, invia un'email a >> dynlist-pisa(a)googlegroups.com <mailto:dynlist-pisa@googlegroups.com>. >> Visita questo gruppo all'indirizzo >> https://groups.google.com/group/dynlist-pisa. >> Per altre opzioni visita https://groups.google.com/d/optout. > -- --federico poloni Dipartimento di Informatica, Università di Pisa http://www.di.unipi.it/~fpoloni/ tel:+39-050-2213143

1 0

Fwd: 2 open PhD positions, TU Eindhoven with industry in Milano
by Gianna M. Del Corso 06 Nov '18

06 Nov '18

Michiel Hochstenbach mi chiede di diffondere questa call per 2 posizioni di PhD a Eindhoven/Milano. Gianna > Inizio messaggio inoltrato: > > Da: "Hochstenbach, M.E." <M.E.Hochstenbach(a)tue.nl> > Oggetto: 2 open PhD positions, TU Eindhoven with industry in Milano > Data: 6 novembre 2018 15:10:43 CET > A: "delcorso(a)di.unipi.it" <delcorso(a)di.unipi.it> > > Dear Gianna, > > I hope you are fine! > > This email is to let you know that we have 2 open 4-year PhD positions at TU Eindhoven per March 1, 2019 in the "BIGMATH" project; I will be the main academic advisor. > These are 2 Marie Curie EID (European Industrial Doctorate) positions, together with industry in Milano (AcomeA and SDG). > The PhD students will spend ca 28 months in Eindhoven and 20 months in Milano. > > Your friend Alessandra Micheletti is also involved in this project. > > All information is available at http://www.win.tue.nl/~hochsten/bigmath.html. > The project descriptions can be found there; the topics might include numerical linear algebra, big data, large-scale optimization, modeling, probability and statistics, model reduction. > The deadline is December 5. > > Could you please forward this info to potentially interested students? > > Thank you very much in advance! All the very best and kind regards, > > Michiel Hochstenbach > > PS: There are also 2 positions at the University of Milano (please see http://itn-bigmath.unimi.it/); for those positions the restriction is a presence of <= 12 months in Italy over the last 3 years (this is not a restriction for the Eindhoven positions). Moreover, there are 2 positions in Lisbon and 1 in Novi Sad. >

1 0

Avviso di seminario
by Leonardo Robol 01 Nov '18

01 Nov '18

Speaker: Giovanni Barbarino Affiliation: Scuola Normale Superiore, Pisa Time: Thursday, 15 November 2018, h. 10:00 Place: Sala Seminari Est, Dipartimento di Informatica, Università di Pisa Title: Higher order spectral symbols and eigenvalues approximation for Toeplitz matrices When modeling natural phenomena with linear partial differential equations, the discretized system of equations is generally represented by a sequence of matrices, often coupled with a function called spectral symbol that enclosures asymptotic spectral information of the sequence. In fact, when sampling the symbol, we obtain an approximation of the spectrum of the matrices, with an error having magnitude inversely proportional with respect to the matrix size. Under particular hypothesis, the sequence presents an higher-order asymptotic formula for the eigenvalues, that let us precisely compute the eigenvalues of large matrices with a technique of interpolation-extrapolation at a low computational cost and that is also parallelizable. This algorithm seems also to work on different kind of matrices, but with irregularities on the extreme eigenvalues, for which a more accurate analysis is needed. -- Leonardo Robol

1 0

Seminari di Analisi Numerica
by Leonardo Robol 09 Oct '18

09 Oct '18

Carissimi, come negli scorsi anni ci piacerebbe organizzare una serie di seminari di analisi numerica, con lo scopo di conoscere le attività all'interno del gruppo e dare spazio a studenti e visitatori. Per cercare di massimizzare la possibilità di seguirli, chiederei a tutti gli interessati di compilare il Doodle [1], segnalando gli slot in cui siete liberi da lezioni e/o che preferite. Poi cercheremo di tenerne conto nel programmare i seminari. Ultimo, ma non meno importante, siamo in cerca di volontari per tenere qualche seminario! Per proporvi, o per proporre visitatori e/o studenti scrivete una e-mail a me e Federico, eventualmente con la disponibilità di date. Può essere una buona occasione per provare una presentazione prima di una conferenza o di una tesi, per esempio. Aspettiamo con fiducia adesioni numerose! [1]: https://doodle.com/poll/eqfkkrztkzywp2qp Ciao, -- Leonardo e Federico.

1 0

corso dottorato @SNS "Metodi di Krylov e applicazioni"
by Federico Poloni 25 Sep '18

25 Sep '18

Buongiorno, segnalo, per gli interessati, questo corso di dottorato che si terrà alla Scuola Normale Superiore. Le lezioni (per circa 60 ore) si terranno a partire dal 5 novembre col seguente orario: Lunedi ore 11-13 Aula Bianchi (Scienze) Mercoledi ore 11-13 Aula Tonelli presso il Palazzo della Carovana in Piazza dei Cavalieri. Indicativamente le lezioni termineranno a fine maggio. Ulteriori dettagli nella scheda allegata. Cordialmente, -- --federico poloni Dipartimento di Informatica, Università di Pisa http://www.di.unipi.it/~fpoloni/ tel:+39-050-2213143

1 0

Fwd: Master in High Performance Computing, Trieste
by Federico Poloni 06 Jul '18

06 Jul '18

Ricevo e pubblicizzo questo annuncio di master. Un saluto a tutti. =============================================================================== Buongiorno, avrete piacere di sapere e di condividere che le iscrizioni <https://www.mhpc.it/how-apply> per la 5a edizione del Master in High Performance Computing <https://www.mhpc.it/> (MHPC) sono state estese *_fino al 24 luglio_*. Al momento abbiamo 5 borse sponsorizzate da beanTech <www.beantech.it>, Plus Solutions <https://www.plusolutions.it/>, OGS <www.ogs.trieste.it/en> e INFN. <www.ts.infn.it/en> Si tratta di un percorso professionalizzante post-laurea (verranno considerati solo candidati in possesso di una laurea magistrale o dottorandi/con dottorato di ricerca) e si svolge a Trieste in un ambiente di ricerca internazionale e stimolante, costituito da centri di eccellenza come SISSA <https://www.sissa.it/> e ICTP <https://www.ictp.it/>. *Obiettivo del master è preparare in 12 mesi specialisti per l'accademia e l'industria nel settore dell'HPC/Big Data, in cui le offerte di lavoro sono in crescita costante. *** Nelle prime 3 edizioni del master sono stati formati 33 specialisti, il 100% di loro ha trovato un'occupazione nel proprio settore di competenza, sia in ambito accademico che industriale, ed un terzo di loro ha già una posizione a tempo indeterminato. Le lezioni sono tenute in lingua inglese da docenti SISSA e ICTP e da esperti di fama internazionale, provenienti da realtà come Intel e Nvidia. Sperando di aver suscitato il vostro interesse verso la nostra iniziativa, vi preghiamo di far girare questa comunicazione grazie ai vostri canali. Vi invitiamo a seguirci al Campus Party di Milano <italia.campus-party.org> il 20 Luglio alle ore 17 (sezione Coding Stage). Rimaniamo a vostra disposizione <mailto:mhpc@sissa.it> per ulteriori informazioni. Cordialmente, lo staff di MHPC Twitter <https://twitter.com/mhpc_sissa_ictp> Facebook <https://www.facebook.com/mhpc.sissa.ictp/> LinkedIn <https://www.linkedin.com/company/mhpc---sissa---ictp/> =============================================================================== Dear Sir/Madame, you will be interested to know and to share that the enrollment <https://www.mhpc.it/how-apply> for the 5th edition of the Master in High Performance Computing <https://www.mhpc.it/> (MHPC) has been extended to a new *_deadline: July 24th_*. We currently offer 5 full scholarships sponsored by beanTech <www.beantech.it>, Plus Solutions <https://www.plusolutions.it/>, OGS <www.ogs.trieste.it/en> and INFN. <www.ts.infn.it/en> This master is a post-graduation school that is organized in Trieste by the synergistic cooperation of two excellence research centers: SISSA <https://www.sissa.it/> and ICTP <https://www.ictp.it/>. *Our aim is to prepare, within 12 months of master, HPC/Big Data specialists, which is a fast growing job in either academy and industry settings. *** During the first three editions of this master, 33 people have been educated, 100% of them are working in their area of expertise and one third of them have already a permanent job. The course is kept in english by SISSA/ICTP lecturers and world-renowed experts from both academy and industry, such as Intel and Nvidia. Hoping that we interested you with this short mail, we would be grateful if you could forward this communication to your networks. You can meet us at the Milano Campus Party <italia.campus-party.org> July 20th 5pm (Coding Stage section). For any further information, please write to us <mailto:mhpc@sissa.it>. Yours faithfully, MHPC staff Twitter <https://twitter.com/mhpc_sissa_ictp> Facebook <https://www.facebook.com/mhpc.sissa.ictp/> LinkedIn <https://www.linkedin.com/company/mhpc---sissa---ictp/>

1 0

Pisa-Hokkaido-Tor Vergata Summer School in Mathematics, 27 Agosto-8 Settembre
by Federico Poloni 28 Jun '18

28 Jun '18

Buongiorno, segnalo questa scuola per i possibili interessati. Si tratta di una scuola dedicata a studenti di master/dottorato di matematica, con diversi mini-corsi di 4-8 ore su vari argomenti (tra cui alcuni più numerici/applicati). ``` PISA-HOKKAIDO-ROMA2 Summer School on Mathematics and Its Applications 2018 Centro De Giorgi, Pisa, 27 Agosto - 8 Settembre 2018. Nell'ambito delle attività congiunte dei Dipartimenti di Matematica della Università di Pisa, di Roma Tor Vergata e di Hokkaido University, dal 27 Agosto all'8 Settembre 2018, al Centro De Giorgi, a Pisa, si terrà la prossima scuola estiva. I corsi offerti riguardano diverse aree della matematica pura e applicata: Paola Boito and Federico Poloni: "Matrix functions and network analysis Claudio Bonanno: "Infinite Ergodic Theory" Piero D'Ancona: "Dispersive estimates and application" Martina Lanini: "Introduction to Schubert varieties" Ugo Locatelli: " The KAM theorem, following the original approach designed by Kolmogorov" Jun Masamune: " Global Analysis - the heat kernel, old and new" Maurizio Parton: "Introduction to neural networks for mathematicians" Simona Settepanella " Discrete Morse Theory" A questo indirizzo si trovano tutte le indicazioni e le informazioni a riguardo: http://www.crm.sns.it/event/432/ In particolare: 1) La scuola è gratuita. 2) Per partecipare alla scuola occorre presentare domanda, seguendo le istruzioni sul sito. 3) L'università può erogare crediti per la frequenza ed il superamento degli esami della scuola. Per l'erogazione dei crediti è necessaria una iscrizione ulteriore e più formale, le cui modalità saranno comunicate agli interessati. 4) All'indirizzo http://www.crm.sns.it/event/432/speakers.html?page=1#title si trovano gli abstract dei corsi. ``` -- --federico poloni Dipartimento di Informatica, Università di Pisa http://www.di.unipi.it/~fpoloni/ tel:+39-050-2213143

1 0

Seminario Veneziani
by Beatrice Meini 20 Jun '18

20 Jun '18

Buongiorno, vi inoltro questo avviso di seminario al dipartimento di Matematica, che potrebbe essere di interesse per il gruppo. Beatrice ---- Giovedi 28 giugno, ore 11:30, Aula Magna, Dipartimento di Matematica Prof. Alessandro Veneziani, Emory Center for Mathematics and Computing in Medicine, Atlanta Title:Old and new challenges of incompressible computational fluid dynamics in cardiovascular problems Abstract: With the progressive use of computational fluid dynamics for supporting medical doctors decision in clinical treatment of cardiovascular diseases, new and old challenges arise in terms of efficiency and reliability. In this talk, we will consider some of them, particularly related to (1) the efficient stabilization and solution of incompressible steady flows (as a surrogate of time average solutions over a heartbeat); (2) the reliable treatment of defective boundary data problems. Applications will be illustrated to coronary as well as aortic diseases.

1 0

seminario Iliescu
by Federico Poloni 14 Jun '18

14 Jun '18

Salve, ricevo e inoltro questo avviso di seminario che potrebbe essere di interesse per gli iscritti a questa lista. === Avviso di seminario: Martedi 19 giugno 2018, ore 1500 presso il dipartimento di Matematica Aula Magna Traian Iliescu (Virginia Tech) Title: "Physically-Constrained Data-Driven Correction for Reduced Order Modeling of Fluid Flows" Abstract: We propose a data-driven correction reduced order model (DDC-ROM) framework for the numerical simulation of fluid flows, which can be formally written as DDC-ROM = Galerkin-ROM + Correction. The new DDC-ROM is constructed by using ROM spatial filtering and data-driven ROM closure modeling (for the Correction term). Furthermore, we propose a physically-constrained DDC-ROM (CDDC-ROM), which aims at improving the physical accuracy of the DDC-ROM. The new physical constraints require that the CDDC-ROM operators satisfy the same type of physical laws (i.e., the Correction term's linear component should be dissipative and the Correction term's nonlinear component should conserve energy) as those satisfied by the fluid flow equations. To implement these physical constraints, in the data-driven modeling step, we replace the unconstrained least squares problem with a constrained least squares problem. We perform a numerical investigation of the new CDDC-ROM and standard DDC-ROM for a 2D channel flow past a circular cylinder at Reynolds numbers $Re=100, Re=500$, and $Re=1000$. To this end, we consider a reproductive regime as well as a predictive (i.e., cross-validation) regime in which we use as little as $50\%$ of the original training data. The numerical investigation clearly shows that the new CDDC-ROM is significantly more accurate than the DDC-ROM in both regimes. -- --federico poloni Dipartimento di Informatica, Università di Pisa http://www.di.unipi.it/~fpoloni/ tel:+39-050-2213143

1 0

mincorso su "Functions of matrices"
by Federico Poloni 26 May '18

26 May '18

Buongiorno! Segnalo questo mini-corso di 12 ore che si terrà in SNS nelle prossime due settimane. "Functions of matrices: theory, numerical methods and applications" Prof. Michele Benzi (Emory University, Atlanta, USA) La prima lezione sarà lunedì in aula Fermi (Scuola Normale) alle 11:00. In caso ci fosse qualcuno extra-SNS interessato a partecipare vi prego di spedirmi un'e-mail così tengo traccia dei numeri di partecipanti esterni. Arrivederci, -- --federico poloni Dipartimento di Informatica, Università di Pisa http://www.di.unipi.it/~fpoloni/ tel:+39-050-2213143

1 0

Seminario G. Barbarino - poll
by Federico Poloni 21 Apr '18

21 Apr '18

Buongiorno! Vorremmo far tenere un seminario a Giovanni Barbarino nelle prossime settimane, sul tema "Higher Order Spectral Symbols and Eigenvalues Approximation for Toeplitz Matrices", e volevo chiedervi scegliere tra le varie possibilità per l'orario del seminario. Se siete interessati a seguire il seminario, segnate le vostre disponibilità di massima su https://doodle.com/poll/vv96x5zbth5cbss4 e poi sceglierò l'opzione più gettonata. Grazie e a risentirci per la data definitiva. Cordialmente, -- --federico poloni Dipartimento di Informatica, Università di Pisa http://www.di.unipi.it/~fpoloni/ tel:+39-050-2213143

1 2

Avviso di seminario
by Dario A. Bini 30 Mar '18

30 Mar '18

Buongiorno, vi invio gli estremi del seminario di Raf Vandebril. Dario --- Giorno: 6 Aprile, ore 10:00 Luogo: Aula Magna, DIpartimento di Matematica Autore: Raf Vandebril, KU Leuven Titolo: On classical, extended, and rational Krylov and the associated QR algorithms Abstract: In this lecture we will discuss three main classes of Krylov and QR type methods: the classical, the extended, and the rational versions. We will initially focus on the classical QR method and reintepret all ingredients such as: associated Krylov spaces, associated Hessenberg matrices, implicit theorems, bulge chasing, eigenvalue swapping, and subspace iteration. Another look at these essentials will allow us to straightforwardly generalize all of this to the extended and rational case. More precisely, we will start by deducing the structure of the recurrence pencil associated to all Krylov methods. It will be shown that the rational Krylov method produces the most general pencil, namely two Hessenberg matrices, which implicitly store the poles of the rational Krylov method. We will revisit the implicit Q-Theorem and link it to its dual theorem, the implicit HK-Theorem, allowing us to provide a pole chasing algorithm for two Hessenberg matrices. This generalizes the classical and extended QR algorithms. Correctness will follow directly from the previous analysis of the classical QR algorithm. The rational QR algorithm allows for additional freedom, which is the introduction of poles. Even without a good pole-strategy we will see that the iteration count reduces with almost 10 procent.

1 0

Fwd: opportunità nella geofisica computazionale - INGV Pisa
by Dario A. Bini 25 Feb '18

25 Feb '18

Mi viene chiesto di diffondere questa informazione. Provvedete anche voi. Dario --- -------- Forwarded Message -------- Subject: opportunità nella geofisica computazionale - INGV Pisa Date: Sat, 24 Feb 2018 10:35:55 -0700 From: Mattia De' Michieli Vitturi <mattia.demichielivitturi(a)ingv.it> To: stefano.galatolo(a)unipi.it, giovanni.federico.gronchi(a)unipi.it, dario.bini(a)unipi.it CC: Tomaso Esposti Ongaro <tomaso.espostiongaro(a)ingv.it> Buongiorno, Vi disturbo perché avremonei prossimi mesi diverse opportunità per assegni di ricerca pre/post-dottorali annuali su tematiche di geofisica computazionale all'INGV di Pisa. Prima di procedere con i bandi, vorremmo esplorare possibili candidature. In particolare, per: 1) data assimilation in atmospheric dispersal models 2) High Performance Computing of multiphase volcanic flows 3) Modeling/simulation of flow and mechanics of porous media 4) Development of automated procedures for the integration of monitoring data into numerical models of volcanic plumes Il profilo 1) sarebbe più matematico applicato, il 2) più computazionale, il 3) più fisico/numerico, il 4) informatico. Se avete in mente persone potenzialmente interessate potete indirizzarle a me o al collega Tomaso Esposti Ongaro per maggiori informazioni: mattia.demichielivitturi(a)ingv.it <mailto:mattia.demichielivitturi@ingv.it> tomaso.espostiongaro(a)ingv.it <mailto:tomaso.espostiongaro@ingv.it> Tutte le attività sarebbero presso l'INGV a Pisa, ma nel caso 2) la borsa è finanziata da OGS (Istituto Naz di Oceanografia e Geofisica Sperimentale) di Trieste. Grazie, buona giornata e buon lavoro Mattia e Tomaso -- Mattia de' Michieli Vitturi, PhD Istituto Nazionale di Geofisica e Vulcanologia Sezione di Pisa Via della Faggiola 32, I-56126 Pisa, Italia ph.: +39-050-8311923 fax.: +39-050-8311942 email: mattia.demichielivitturi(a)ingv.it <mailto:mattia.demichielivitturi@ingv.it> skype: demichie

1 0

Numerical analysis seminars - Durastante
by Federico Poloni 05 Feb '18

05 Feb '18

Buongiorno, inoltro questo avviso di seminario. > il giorno 14 Febbraio alle ore 11 in sala seminari est presso il dipartimento di informatica > il Dott. Fabio Durastante assegnista presso il Dipartimento di Informatica dell'Universita` di Pisa > terra` un seminario dal titolo "Preconditioning Techniques for Some Large Localized or Structured Linear Systems with Application to Discrete Fractional Calculus". Cordialmente, -- --federico poloni Dipartimento di Informatica, Università di Pisa http://www.di.unipi.it/~fpoloni/ tel:+39-050-2213143

1 0

Posizione PhD a Strathclyde
by Federico Poloni 20 Dec '17

20 Dec '17

Ricevo e inoltro in caso ci siano persone interessate. > PhD Studentship on novel preconditioned iterative solvers for radial basis function linear systems > > Applications are invited for a PhD studentship on novel preconditioners for radial basis function linear systems, under the supervision of Dr Jennifer Pestana and Dr Alison Ramage, at the Department of Mathematics and Statistics at the University of Strathclyde, Glasgow, Scotland. The project will suit students with interests in numerical analysis and linear algebra. > > The studentship covers UK/EU tuition fees and comes with an annual tax-free stipend at the standard UK rate. International students who can fund the difference between UK/EU and International fee rates are also encouraged to apply. > > The project will begin 1 March, 2018, or earlier by arrangement. > > Further details are available at > > http://personal.strath.ac.uk/jennifer.pestana/RBF_Project.pdf > > Informal enquiries to Jennifer Pestana (jennifer.pestana(a)strath.ac.uk) are welcome. Buone feste a tutti, -federico -- --federico poloni Dipartimento di Informatica, Università di Pisa http://www.di.unipi.it/~fpoloni/ tel:+39-050-2213143

1 0

due giorni a Padova
by Dario Bini 27 Nov '17

27 Nov '17

Carissimi, vi ricordo che l'8 e il 9 febbraio 2018 si terra' la "Due Giorni di Algebra Lineare Numerica e Applicazioni" www.math.unipd.it/~pd2ggaln Se interessati vi invito a comunicare all'indirizzo pd2ggaln(a)math.unipd.it la vostra manifestazione di interesse possibilmente specificando se si desidera tenere una comunicazione. Ciao, Dario ---

1 0

assegno di ricerca UNIPI
by Federico Poloni 21 Nov '17

21 Nov '17

Luca Gemignani mi comunica che è stato pubblicato il seguente bando per un assegno di ricerca sul progetto "Modelli e algoritmi innovativi per problemi strutturati e sparsi di grandi dimensioni" (scadenza domande 20 dicembre; necessario avere il dottorato). Info su https://www.unipi.it/ateneo/bandi/assegni/pra2017/inf/20dic2017/index.htm Un saluto a tutti, -- --federico poloni Dipartimento di Informatica, Università di Pisa http://www.di.unipi.it/~fpoloni/ tel:+39-050-2213143

1 0

Seminar
by Dario Bini 02 Nov '17

02 Nov '17

Avviso di seminario: Speaker: Samandar Babaev (Bukhara State University, Uzbekistan) Date: Wednesday, November 8 Time: 11:00 Place: Sala Seminari, Dipartimento di Matematica Title: Optimal Interpolation Formulas in W_2^{(m,m-1)} space Abstract: In this talk, using S.L.Sobolev’s method, optimal interpolation formulas are constructed in /W_{2} ^/(/m,m//−/1)(0/,/1) space. Explicit formulas for coefficients of optimal interpolation formulas are obtained. Some numerical results are presented.

1 0

Avviso di Seminario
by Dario Bini 26 Oct '17

26 Oct '17

Avviso di seminario: Speaker: Federico Poloni (Università di Pisa) Date: Tuesday 2017-10-31 Time: 11:00 Place: Sala Seminari Ovest, Dipartimento di Informatica Title: Counting Fiedler pencils using diagrams Fiedler pencils are a family of matrix pencils which generalizes the well-known companion matrix: they are*linearizations*, i.e., they provide a method to construct, given a matrix polynomial, a linear eigenvalue problem with the same eigenvalues and multiplicities. Fiedler pencils are constructed as products of special block matrices that act nontrivially only on two contiguous blocks. They have a rich structure that give rise to many combinatorial properties. In this talk, we introduce a notation that associates to each pencil a diagram that depicts its action on the blocks. Using this notation, we can obtain visual proofs of several statements in the theory, and we can solve several counting properties (such as "how many distinct Fiedler pencils with repetitions of a given dimension exist"). Among them, in particular, we are interested in counting Fiedler pencils associated to symmetric and palindromic matrix polynomials which preserve the same structure. This talk is based on a joint work with Gianna Del Corso.

1 0

Avviso di seminario
by Dario A. Bini 13 Oct '17

13 Oct '17

Invio il seguente avviso di seminario Dario --- Gorno: Martedi 17 ottobre Ora: 11:00 Aula: Seminari, Dipartimento di Matematica Speaker: Leonardo Robol, ISTI --- CNR Titolo: Backward error analysis for polynomial rootfinders. Abstract: The most common methods to compute roots of polynomials is to rephrase the problem in linear algebra terms: one constructs a companion matrix (or pencil) that has the roots of the polynomial as eigenvalues. However, even when the eigenvalue method used is backward stable, it's non-trivial to map the backward error back on the polynomial. This problem has been studied extensively in the past and it has been shown that relying on the QR method does not provide a normwise backward-stable rootfinder. For this reason, one needs to rely on QZ + appropriate scaling. However, companion matrices are unitary plus rank 1, and we claim that relying on fast QR method that exploits this structure can lead to a backward-stable method. To this end, one needs to ensure that the errors on the unitary and rank 1 parts satisfy different bounds. We show how to perform this analysis of "mixed perturbation backward-error", as well as a method that satisfies these constrains. Time permitting, we discuss some more details that one should consider. We show, for example, that normwise backward stable operations (i.e., unitary rotations in our case) do not necessarily imply backward stable operations on the unitary and low-rank factors, unless some particular identities involving sines and cosines are preserved. This is joint work with Jared Aurentz, Thomas Mach, Raf Vandebril, and David Watkins.

1 0

Optimization & Numerical Analysis seminars: Iannazzo
by Federico Poloni 26 Jul '17

26 Jul '17

Speaker: Bruno Iannazzo Affiliation: Università di Perugia, Italy Time: Thursday 2017-08-27, h. 11:00 Place: Sala Seminari Est, Dipartimento di Informatica, Università di Pisa Title: The Riemannian Barzilai-Borwein algorithm and the matrix geometric mean computation Abstract: We consider the problem of computing the geometric mean of positive definite matrices, the so-called Karcher mean. For this purpose, we adapt the Barzilai-Borwein method from nonlinear optimization to Riemannian manifold optimization. Incomporating a nonmonotone line-search strategy yields global convergence under standard assumptions, fulfilled in the problem of the matrix geometric mean. By a set of numerical tests, we show that the Riemannian Barzilai-Borwein method notably outperforms existing first-order optimization methods for computing the matrix geometric mean. === Everyone is welcome. Sorry for the short notice! -- --federico poloni Dipartimento di Informatica, Università di Pisa http://www.di.unipi.it/~fpoloni/ tel:+39-050-2213143

1 1

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

NumPI