[Forum SIS] Bocconi DEC Seminar: Nicolò Cesa-Bianchi - May 27th
Anna Simoni
anna.simoni a unibocconi.it
Ven 21 Maggio 2010 16:50:25 CEST
Dear colleagues,
The Department of Decision Sciences (DEC) of Bocconi University is pleased to invite you to the seminar:
==================================================================
DEC Seminar
Università Bocconi,
Room 3-E4-SR03
Via Roentgen 1 - 3rd floor
Time: 12:30pm
The DEC seminar schedule is available at http://www.unibocconi.eu/statseminar
Thursday, May 27th
Nicolò Cesa-Bianchi
(Università degli Studi di Milano)
"Scalable algorithms for prediction on a graph "
Abstract:
Networked data are found in a variety of domains: Web, social networks, biological networks, and many others. In learning tasks, networked data are often represented as a weighted graph whose edge weights reflect the similarity between incident nodes.
In this talk, we consider the problem of classifying in the game-theoretic mistake bound model the nodes of an arbitrary given graph. We characterize the optimal predictive performance in terms of the cutsize of the graph's random spanning tree, and describe a randomized prediction algorithm achieving the optimal performance while running in expected time sublinear in the graph size (on most graphs).
These results are then extended to the active learning model, where training labels are obtained by querying nodes selected by the algorithm. We describe a fast query placement strategy that, in the special case of trees, achieves the optimal number of mistakes when classifying the non-queried nodes.
Joint work with: Claudio Gentile, Fabio Vitale and Giovanni Zappella
==============================================================================
Sincerely,
Anna Simoni
-----------------------------
Anna Simoni
Assistant Professor
Department of Decision Sciences
Università Bocconi
via Roentgen, 1
20136 Milano - Italy
Email:anna.simoni a unibocconi.it
Webpage: http://faculty.unibocconi.eu/annasimoni/
---
Il messaggio che segue e' inserito automaticamente dal server di posta
dell'Universita' Bocconi.
Il 5 per mille per gli studenti meritevoli della Bocconi.
E' un atto volontario, non costa nulla e non sostituisce l'8 per mille.
Scegli Bocconi: codice fiscale 80024610158.
Please note that the above message is addressed only to individuals
filing Italian income tax returns.
---
-------------- parte successiva --------------
Un allegato HTML è stato rimosso...
URL: <http://www.stat.unipg.it/pipermail/sis/attachments/20100521/1d56041d/attachment.html>
Maggiori informazioni sulla lista
Sis