Multidimensional analysis of complex networks — PhD Dissertation full text

The fol­low­ing is the short abstract of my PhD The­sis. For those of you inter­ested in the full PhD dis­ser­ta­tion, please take a look at AMS Dot­torato web­site. Com­ments, sug­ges­tions, feed­backs are always well appreciated!

Com­plex Net­works analy­sis turn out to be a very promis­ing field of research, tes­ti­fied by many research projects and works that span dif­fer­ent fields. Those analy­sis have been usu­ally focused on char­ac­ter­ize a sin­gle aspect of the sys­tem and a study that con­sid­ers many infor­ma­tive axes along with a net­work evolve is lacking.

Con­tinue read­ing

Multidimensional analysis of complex networks

PhD dis­ser­ta­tion abstract

An Australian Virtual Social networkThe analy­sis of Com­plex Net­works turn out to be a very promis­ing field of research, tes­ti­fied by many research projects and works that span dif­fer­ent fields. Until recently, those analy­sis have been usu­ally focused on deeply char­ac­ter­ize a sin­gle aspect of the sys­tem, there­fore a study that con­sid­ers many infor­ma­tive axes along with a net­work evolve is lack­ing.
In this The­sis, we pro­pose a new mul­ti­di­men­sional analy­sis that is able to inspect net­works in the two most impor­tant dimen­sions of a sys­tem, namely space and time. In order to achieve this goal, we stud­ied them sin­gu­larly and inves­ti­gated how the vari­a­tion of the con­sti­tut­ing para­me­ters dri­ves changes to the net­work behav­iour as a whole.

By focus­ing on space dimen­sion, we were able to char­ac­ter­ize spa­tial alter­ation in terms of abstrac­tion lev­els. We pro­pose a novel algo­rithm that, by apply­ing a fuzzi­ness func­tion, can recon­struct net­works under dif­fer­ent level of details. We call this analy­sis tele­scopic as it recalls the mag­ni­fi­ca­tion and reduc­tion process of the lens.
Through this line of research we have suc­cess­fully ver­i­fied that sta­tis­ti­cal indi­ca­tors, that are fre­quently used in many com­plex net­works researches, depends strongly on the gran­u­lar­ity (i.e, the detail level) with which a sys­tem is described and on the class of net­works con­sid­ered. Con­tinue read­ing

Twitter hashtags length over time

Twitter hashtags length over timeThis post is the first of the new Sim­ple Sci­ence cat­e­gory. This thread fea­tures research top­ics that are mainly focused on intu­itive research facts/insight detected in the most impor­tant and well known vir­tual com­mu­ni­ties. How­ever, research on com­plex net­works is very inter­dis­ci­pli­nary thus we do not exclude to extend our find­ings to other than social networks.

Since we want an audi­ence as broad as pos­si­ble and the con­tent to be as read­able as pos­si­ble, we will not use strong sta­tis­ti­cal tools, but instead we adopt the basic and under­stand­able ones. In spe­cial cases where more advanced math­e­mat­i­cal meth­ods are needed, we pro­vide ade­quate and clear explanations.

Con­tinue read­ing

Hacking Openpicus usb-to-serial converter

Usb to serial converter hackingGeeks usu­ally have, in their DIY bags, many usb-to-serial con­vert­ers such as Arduino’s or Open­picus’. These con­vert­ers, for the peo­ple not accus­tomed to micro­con­trollers’ world, are cir­cuits nor­mally used for upload­ing com­piled code from the PC. Addi­tion­ally, they are used to seri­ally com­mu­ni­cate with the PC. They take usb sig­nal and con­vert it to ser­ial TTL.

Even though ser­ial TTL is very impor­tant when com­mu­ni­cat­ing with micro­con­trollers, many home and indus­trial devices exchange data using RS485 pro­to­col. The main ben­e­fits of using RS485 are basi­cally effi­ciency, econ­omy, the long dis­tances allowed between devices (up to 1200 meters) in elec­tri­cally noisy envi­ron­ments. Con­tinue read­ing

Ambiguous author query detection

Ambiguous author query detectionThe name ambi­gu­ity prob­lem is espe­cially chal­leng­ing in the field of bib­li­o­graphic dig­i­tal libraries. The prob­lem is ampli­fied when names are col­lected from het­ero­ge­neous sources. This is the case in the Schol­arom­e­ter sys­tem, which per­forms bib­lio­met­ric analy­sis by cross-correlating author names in user queries with those retrieved from dig­i­tal libraries. The uncon­trolled nature of user-generated anno­ta­tions is very valu­able, but cre­ates the need to detect ambigu­ous names. Our goal is to detect ambigu­ous names at query time by min­ing dig­i­tal library anno­ta­tion data, thereby decreas­ing noise in the bib­lio­met­ric analy­sis. Con­tinue read­ing