Transcript Network!

Eötvös University
Budapest
in the
Network
The Team
 Seniors:
• István Csabai (node coordinator):
» Photometric redshift estimation, virtual observatories, science
database technology, SDSS
• Zsolt Frei:
» Galaxy morphology, galaxy mergers, gravitational waves
 Students:
• Norbert Purger, Bence Kocsis, Merse Gáspár,
• Márton Trencséni, László Dobos, Dávid Koronczay
» Working on SDSS related topics
 Network student:
• Oliver Vince (Belgrade)
– Training focus:
» Extragalactic astronomy, cosmology - Network!
» Computational astronomy, VO technology
Focus topics
 Development of datamining and
visualization techniques – SDSS ‘color
space’
 Improving photometric redshift estimation
 Estimation of physical parameters of
galaxies from photometry
 Bulge/disk separation of large SDSS
galaxies
 Virtual Observatory, Spectrum Services
Collaboration with other nodes
 JHU:
• Alex Szalay, Tamas Budavari, Ani Thakar …
• Virtual observatories, SDSS database, photometric redshift
estimation
• Regular visits for seniors ad students
 Paris:
•
•
•
•
Stephane Charlot
Spectral synthesis models for photo-z, spectral models in VO
Oliver Vince visited Paris, and will visit next year
New joint topic involving several nodes: „Optical attenuation
law of nearby galaxies”
Datamining: The Color Space
300 million points
in 5+ dimensions
u
g
r
i
z
Datamining: Spatial Indexing
Datamining: Speed Up Queries
duration (msec)
80000
60000
40000
kd-tree
20000
SQL
0
0
0.05
0.1
0.15
0.2
0.25
ratio of rows returned
0.3
0.35
Datamining: Visualization
Adaptively
fetch data
from
database
Datamining:Integration with Database
TRADITIONAL APPROACH
Flat files, Fortran, C code
+ Complex manipulation of data
- Sequential slow access
VISUALIZATION
Tools using OpenGL, DirectX
+ Fast
- Using files, some tools access
database, but not interactive
INTEGRATE
•Implement in SQL Server
•use for astronomical data-mining
•and for fast interactive visualization
MULTIDIMENSIONAL INDEXING
B-tree, R-tree, K-d tree, BSP-tree …
+ Many for low D, some for high D
+ Fast, tuned for various problems
- Implemented mostly as memory
algorithms, maybe suboptimal in
databases
SQL DATABASES
Oracle, MS SQL Server, …
+ Organize, efficiently access data
- Hard to implement complex algorithms
- Multidimensional indexing (OLAP) is limited to
categorical data
• Joint Eötvös & JHU publication at the
Conference on Innovative Data Systems Research
Estimating physical parameters and
redshift
3-10 DIMENSION
PARAMETRS
age, dust, ...
GALAXY
early type,
late type
LIGHT;
SED
3000 DIM
5 DIMENSION
MAGNITUDE
SPACE
BROADBAND
FILTERS
REDSHIFT
Photometric redshift estimation
100M galaxies with known
ugriz photometry, but no redshift
ugriz
•Find k nearest neighbors
•Use polinomial regression
•Estimate redshift
redshift
1M galaxies with known
photometry and redshift
Photometric redshift estimation
 Joint work between JHU & Eötvös
 Photometric redshift calculated for 300M
SDSS objects
 Included in SDSS DR5 Catalog and Data
Release paper
 Application: targeting MgII absorbers
 collaboration between MPA & Eötvös
 network ER Vivienne Wild involved
Virtual Observatory: Spectrum &
Filter Services
 Developed by Eötvös
student Laszlo Dobos &
JHU researcher Tamas
Budavari
 Several joint publications
 Collaboration with IAP
researcher Stephane
Charlot to include
spectral synthesis
models
Spectrum Services example: similar
spectrum search
Network events
 MAGPOP Virtual
Observatory
Workshop Budapest, Hungary,
2005. April 25-26
 MAGPOP Summer
School - Budapest,
Hungary, 2006.
August 23-25
 Hosting the webpage