[Forum SIS] Bocconi DEC Seminar: Jukka Corander - October 28th

Anna Simoni anna.simoni a unibocconi.it
Lun 25 Ott 2010 12:16:13 CEST



Dear colleagues, 



The Department of Decision Sciences (DEC) of Bocconi University is pleased to invite you to the seminar: 


================================================================== 

DEC Seminar 
Universitā Bocconi, 
Room 3-E4-SR03 
Via Rontgen 1 - 3rd floor 
Time: 12:30pm 

The DEC seminar schedule is available at http://www.unibocconi.eu/statseminar 

Thursday, October 28th 

Jukka Corander 
(Department of Mathematics and statistics - University of Helsinki ) 

"Have I seen you before? Principles of Bayesian predictive classification Revisited" 

Abstract: 

Classification of objects into a finite set   of alternative classes based 
on observed features of the objects is a common task in statistical 
machine learning. An important application example familiar to most of 
us is spam filtering of email messages. In this talk we review the 
probabilistic basis ! of generative classification and show how a 
particular in! ductive rule of classification arises from basic principles 
of predictive probabilistic modeling pioneered by Seymour Geisser in 
1960's. The standard practice of classifying objects one by one, which 
follows from an i.i.d. assumption, is demonstrated to be at odds with 
laws of predictive probability and we show also that it can be motivated 
as an asymptotic approximation to a more coherent rule. A novel 
inductive principle of predictive classification is introduced and we 
discuss its properties in relation to other principles. 

============================================================================== 

Sincerely, 
Anna Simoni 



--------------------------- 
Anna Simoni 
Assistant Professor 
Department of Decision Sciences 
Universitā Bocconi 
via Roentgen, 1 
20136 Milano - Italy 
Email: anna.simoni a unibocconi.it 
Webpage: http://faculty.unibocconi.eu/annasimoni/
-------------- parte successiva --------------
Un allegato HTML č stato rimosso...
URL: <http://www.stat.unipg.it/pipermail/sis/attachments/20101025/8ee44c3f/attachment.html>


Maggiori informazioni sulla lista Sis