Submitted by noronen on November 30, 2009 - 00:06
M.Sc. Petri Kontkanen will defend his doctoral dissertation "Computationally Efficient Methods for MDL-Optimal Density Estimation and Data Clustering" on Monday, November 30, at 12 o'clock noon in Auditorium CK112 (downstairs by the cafeteria) of the Exactum building at the Kumpula Campus. The Opponent is Professor Kenji Yamanishi from University of Tokyo, and the Custos is Professor Petri Myllymäki.
Submitted by noronen on November 27, 2009 - 21:12
When: Friday 27.11.2009 10am- 1pm.
Where: Place: Pluto cabinet at Pilotti Building, 5th floor, Spektri
We have an international visitor working with Multitouch and other interesting novel applications and have arranged an informal workshop for 10am -1pm noon on Friday 27th November with one 45 minute presentation and 4 X 15 minute presentations and discussions.
Our visiting speaker is Hyowon Lee from Dublin City University
http://www.compapp.dcu.ie/~hlee/
Doctoral dissertation: mathematical techniques for computing NML efficiently
Fri, 27.11.2009
30 Nov 2009 - In the doctoral dissertation, Petri Kontkanen presents mathematical techniques for computing Normalized Maximum Likelihood (NML) distribution efficiently for some model families involving discrete data. Kontkanen also shows how these techniques can be used to apply MDL in two practical applications: histogram density estimation and clustering of multi-dimensional data.
Read more from E-thesis »
You can make a direct webcast, recorded video or podcast from your lecture or event.
Direct webcasts:
Recorded videos:
Submitted by jmalahti on November 25, 2009 - 16:41
Description:
Schedule:
2009-11-25 16:00 - 23:00
Duration:
7 h
Affected services:
HIIT Wiki
Reason:
Platform will be upgrated to Debian 5.0 "Lenny" and Confluence will be upgrated to 3.0.
UPDATE: Upgrade postponed because of meeting.
UPDATE 2: Upgrade postponed to 2009-11-25 because of backup server installation.
UPDATE 3: Upgrade took longer than expected due to problems with tomcat. Memory of the platform was doubled to 8 GB.
Pages