The In­sti­tu­te for Sta­tis­tics and Ma­the­ma­tics is plea­sed to in­vi­te you to the next re­se­arch se­mi­nar, ta­king place on cam­pus:

Sirio Le­gra­man­ti(De­part­ment of Eco­no­mics, Uni­ver­si­ty of Ber­ga­mo)
Le­ver­aging Co­va­ria­tes in Baye­si­an Non­pa­ra­metric Clus­te­ring: An Ap­p­li­ca­ti­on to Trans­por­ta­ti­on Net­works
Wed­nes­day, De­cem­ber 18, 2024, 17:30, in Buil­ding D4, Room D4.0.127.

Abs­tract:

In clus­te­ring, ob­ser­ved in­di­vi­du­al data are often ac­com­pa­nied by co­va­ria­tes that can as­sist the clus­te­ring pro­cess its­elf. This is the case, for examp­le, of trans­por­ta­ti­on net­works, where each node has spa­ti­al co­or­di­na­tes, and it is often de­si­ra­ble that clus­ters of nodes are spa­ti­al­ly co­he­si­ve. In fact, the ob­tai­ned clus­ters may be used to in­form pu­blic po­li­cy de­cis­i­ons, and it may be pre­fer­ra­ble that such po­li­ci­es are uni­form over neigh­bo­ring areas. Na­tu­ral­ly, de­pen­ding on the ap­p­li­ca­ti­on, dif­fe­rent no­ti­ons of clo­sen­ess can be used to de­fi­ne such neigh­bor­hoods, thus po­ten­ti­al­ly re­qui­ring pro­per trans­for­ma­ti­ons of the spa­ti­al co­va­ria­tes. Mo­ti­va­ted by real-​world data about the month­ly sub­scrip­ti­ons to the pu­blic trans­por­ta­ti­on sys­tem of the Ber­ga­mo pro­vin­ce (Italy), we show how to in­cor­po­ra­te properly-​transformed spa­ti­al co­va­ria­tes into a state-​of-the-art sto­chastic block model, while al­lo­wing to weight the con­tri­bu­ti­on of co­va­ria­tes.

(joint work with Va­len­ti­na Ghi­di­ni and Raf­fae­le Ar­gien­to)

We aim to stream all on-​campus talks via Zoom. A di­rect link to the stream will be posted on our web­site.

For fur­ther in­for­ma­ti­on and the se­mi­nar sche­du­le, plea­se see:
www.wu.ac.at/en/stat­math/re­se­arch/res­se­mi­nar