[Forum SIS] Occasional DEC Seminar (Bocconi)- February 15th at 16:00.00

[Isadora Antoniano]

*DEC - Occasional Seminar*

Monday, February 15th

*Università Bocconi,*

Room 3-E4-SR03

Via Rontgen 1 - 3rd floor

Time: *16:00 pm*


*Nick Polson*
(The University of Chicago Booth School of Business)

*"Deep Tree Predictors"*

Abstract

We develop the class of deep tree predictors for the construction of a high
dimensional input-output learning machine. Deep learners have achieved
recent empirical predictive success in high dimensional learning problems.
Deep tree predictors build on this success by directly using a
non-differentiable tree activation functions at the first layer of the
architecture. In doing so, this provides more flexibility than traditional
machine learner and leads to improved predictive performance that handles
the curse of dimensionality. We illustrate our methodology in a
non-parametric regression problem and contrast our fitted machine with a
random forest. Finally, we conclude with directions for future research.

*ATTENTION*

If you are a guest and you do not have a Bocconi ID Card to access the
Bocconi Buildings, please confirm your participation to the event by
sending an email to martina.ercoli a unibocconi.it.
In addition, we suggest that you print, fill out and send the attached form
(to give your consent to personal data handling) via email to the above
address. In this way, the check-in procedure at the entrance will be faster
and easier.


Kind regards,

Raffaella Piccarreta and Isadora Antoniano
-------------- parte successiva --------------
Un allegato HTML è stato rimosso...
URL: <http://www.stat.unipg.it/pipermail/sis/attachments/20160205/78dac906/attachment-0001.html>
-------------- parte successiva --------------
Un allegato non testuale è stato rimosso....
Nome:        ModuloVisitorPass&Privacy.pdf
Tipo:        application/pdf
Dimensione:  56070 bytes
Descrizione: non disponibile
URL:         <http://www.stat.unipg.it/pipermail/sis/attachments/20160205/78dac906/attachment-0002.pdf>
-------------- parte successiva --------------
Un allegato non testuale è stato rimosso....
Nome:        Polson.15.02.16.pdf
Tipo:        application/pdf
Dimensione:  42506 bytes
Descrizione: non disponibile
URL:         <http://www.stat.unipg.it/pipermail/sis/attachments/20160205/78dac906/attachment-0003.pdf>