My cur­rent research is about lar­ge-sca­le distri­bu­ted systems, cloud com­pu­ting, lar­ge-sca­le gra­ph ana­ly­sis, data mining and P2P systems. Students of the Laurea Triennale inte­re­sted in doing their the­sis with me should have alrea­dy com­ple­ted the cour­ses on Operating Systems, Computer Networks, Algorithms and Data Structures, Programming 1. Students of Laurea Magistrale should have com­ple­ted cour­ses on Distributed Systems and/or Big Data, depen­ding on the con­tent of the­sis.

How to ask for a thesis

The pro­cess of pai­ring a stu­dent to her/his super­vi­sor is real­ly a ran­dom one, in Trento as well as in Italian uni­ver­si­ties. Students ask for a the­sis, pro­fes­sors pro­po­se some ideas or refu­se clai­ming that they are over­com­mit­ted / they have too many stu­den­ts / they have no pro­jec­ts at the moment / etc.

In my case, some­ti­mes I am real­ly obli­ged to say no; the­re are periods in which I recei­ve four-five requests per week, and clear­ly I can­not be a good super­vi­sor for all of them. In order to under­stand if a stu­dent is the right per­son for a the­sis, I ask you to send me a mail spe­ci­fy­ing the fol­lo­wing infor­ma­tion:

  • When you want to start
  • When you want to finish (ideal­ly)
  • How many exams you need to pass in order to com­ple­te your degree
  • The list of exams as out­put by Esse3, with the marks that you have obtai­ned
  • The gra­de point ave­ra­ge (voto medio pesa­to)
  • If you have addi­tio­nal expe­rien­ces besi­de the cour­ses at the uni­ver­si­ty, add a CV
  • Your per­so­nal inte­rests in the field of com­pu­ter scien­ce

I pre­fer to super­vi­se the­ses that are either com­ple­te­ly exter­nal (sta­ge + the­sis com­ple­ted in a com­pa­ny) or com­ple­te­ly inter­nal (UniTN intern­ship + the­sis com­ple­ted at DISI). This cor­re­sponds to 15 ECTS cre­di­ts at the Bachelor level (appro­xi­ma­te­ly 2.5–3 mon­ths) and to 30 ECTS cre­di­ts at the master level (appro­xi­ma­te­ly 5–6 mon­ths).

Current ideas (May 2016)

  • Title: Discovering Network Communities from Cascades
    Community detec­tion is an impor­tant pro­blem in com­plex net­work ana­ly­sis. Most exi­sting methods rely on the expli­cit net­work struc­tu­re (link bet­ween nodes) in order to disco­ver com­mu­ni­ties. Community detec­tion is a chal­len­ging task, and most for­mu­la­tions are NP-Hard. In addi­tion to that, often­ti­mes the expli­cit struc­tu­re of the net­work might be latent (hid­den) and that makes the pro­blem even more chal­len­ging. However, we have other sour­ces of infor­ma­tion besi­des the net­work struc­tu­re that allow us to tac­kle this pro­blem, and one inva­lua­ble instan­ce of such case is the dif­fu­sion of a con­ta­gion in a net­work; for exam­ple the dif­fu­sion of a virus in a socie­ty, the dif­fu­sion of a pie­ce of news or meme in onli­ne social net­works and so on. An inte­re­sting pro­per­ty of such even­ts is that most of the time the spread of a con­ta­gion occurs within a clo­se­ly rela­ted or clu­ste­red sub­jec­ts. Even thou­gh this pro­per­ty has been con­fir­med in seve­ral stu­dies, it has not been well exploi­ted to disco­ver the under­ly­ing com­mu­ni­ty struc­tu­re. Thus, the goal of this research is to infer the com­mu­ni­ty struc­tu­re from dif­fu­sion even­ts only, i.e. without having the kno­w­led­ge of the net­work struc­tu­re; and effec­ti­ve­ly addres­sing this pro­blem will have tre­men­dous appli­ca­tions in real world pro­blems such as, epi­de­mio­lo­gy and viral mar­ke­ting.  We seek to address this pro­blem from repre­sen­ta­tion lear­ning per­spec­ti­ve, more par­ti­cu­lar­ly using deep neu­ral net­works. Therefore, the pro­blem requi­res a good back­ground in Machine Learning more impor­tan­tly neu­ral net­works, and also in Social Network Analysis.

Some of my past students here in Trento…

  • Roberto Zandonati wor­ked on his the­sis about the sli­cing pro­blem in peer-to-peer systems. We later coo­pe­ra­ted in wri­ting a paper based on his work. The paper has been accep­ted here:
    Alberto Montresor and Roberto Zandonati. Absolute sli­cing in peer-to-peer
    . In Proc. of the 5th International Workshop on Hot Topics in Peer-to-Peer Systems (HotP2P’08), Miami, FL, USA, April 2008.
  • Alessio Guerrieri wor­ked on his the­sis on DTNs in coo­pe­ra­tion with the Create-Net research cen­ter (here in Povo). A paper based on his work has been accep­ted here:
    Alessio Guerrieri, Alberto Montresor, Iacopo Carreras, Francesco De Pellegrini, and Daniele Miorandi. Distributed esti­ma­tion of glo­bal para­me­ters in delay-tole­rant net­works. In Proceedings of the 3rd IEEE WoWMoM Workshop on Autonomic and Opportunistic Communications (AOC’09), Kos, Greece, June 2009.
    Later, an exten­ded ver­sion of this paper was publi­shed in a jour­nal: Alessio Guerrieri, Iacopo Carreras, Francesco De Pellegrini, Daniele Miorandi, and Alberto Montresor. Distributed esti­ma­tion of glo­bal para­me­ters in delay-tole­rant net­works. Computer Communications, 2010.
    BTW, Alessio also secu­red a scho­lar­ship of 12.000 euros to par­ti­ci­pa­te in a dou­ble degree with GeorgiaTech. He spent the aca­de­mic year 2009/2010 in Atlanta, Georgia, USA. He later publi­shed other papers during his Ph.D. stu­dies under my super­vi­sion.
  • Andrea Dalla Valle wor­ked on a the­sis on par­ti­tion detec­tion in peer-to-peer systems. We have not wor­ked on a paper yet (my fault!); but again, in the mean time Andrea was the second stu­dent to get the Georgiatech scho­lar­ship for 2009/2010.
  • Vinay Sachidananda, one of our stu­den­ts of the “Invest your talent in Italy” pro­gram, wor­ked on an exter­nal the­sis with ArsLogica; I ser­ved as inter­nal tutor. Later, part of his work was publi­shed here: Andrey Somov, Vinay Sachidananda, and Roberto Passerone. A Self-Powered Module with Localization and Tracking System for Paintball. In Proceedings of IWSOS 2008, Vienna, Austria, December 12th 2008. Springer Verlang: LNCS 5343, 182 — 193. As you can guess from the author list, my rile was mar­gi­nal
  • Gabriele Seppi wor­ked on a the­sis about “Popularity-based Caching in Underlying Networks With Client Mobility” wor­king toge­ther with DoCoMo (Germany). Gabriele took part in Double Degree with Georgiatech. The work has been done com­ple­te­ly by Gabriele and the DoCoMo guys.
  • Stella Margonar deve­lo­ped the Java soft­ware that is avai­la­ble on my Algoritmi e Strutture Dati cour­se web page, for the visua­li­za­tion of algo­ri­thms and exer­ci­ses. Her work was later bought by the publi­sher that prin­ted my book on the topic.
  • Simone Miorelli deve­lo­ped a flock simu­la­tor. The idea is that flocks are exam­ples of self-orga­ni­zed distri­bu­ted systems; each flock mem­ber fol­lo­ws very sim­ple rules, whi­le a com­plex, glo­bal beha­vior emer­ges. His work has been spon­so­red by MUSE — the (then) upco­ming museum of natu­ral scien­ce and has inspi­red some of the exhi­bi­ts.
  • Paolo Pandini is an exam­ple that eve­ry­bo­dy should con­si­der even­tual­ly: he is an high-school pro­fes­sor who deci­ded to enroll in our com­pu­ter scien­ce degree after his reti­re­ment. He is beco­ming youn­ger eve­ry year he spend with us! He hel­ped us in desi­gning tea­ching modu­les in com­pu­ter scien­ce for some ele­men­ta­ry schools in Valsugana, based on the “Computer Science Unplugged” book.
  • Federico Scrinzi secu­red a Google Summer of Code scho­lar­ship, wor­king on Euscan (Ebuild Upstream Scanner), a power­ful appli­ca­tion for detec­ting out­da­ted ebuilds in the Gentoo pac­ka­ge mana­ger by loo­king for new upstream ver­sions of the pac­ka­ges. My role was real­ly mini­mal — he com­ple­ted all the work by him­self. He is now a goo­gler in Dublin.