Mireille Louys et al., OV-France Theory WG meeting, LYON, June 2007 1 DALIA : an observation vs...

15
DALIA : an observation vs simulation comparison frame work What could be a Model of simulation code worth for ? What could be a Model of simulation code worth for ? Frédéric Boone Mireille Louys Marie-Lise Dubernet François Bonnarel Nicolas Moreau

Transcript of Mireille Louys et al., OV-France Theory WG meeting, LYON, June 2007 1 DALIA : an observation vs...

Page 1: Mireille Louys et al., OV-France Theory WG meeting, LYON, June 2007 1 DALIA : an observation vs simulation comparison frame work What could be a Model.

DALIA : an observation vs simulation comparison frame work What could be a Model of simulation code worth for ?

What could be a Model of

simulation code worth for ?

Frédéric Boone

Mireille Louys Marie-Lise Dubernet

François Bonnarel Nicolas Moreau

Page 2: Mireille Louys et al., OV-France Theory WG meeting, LYON, June 2007 1 DALIA : an observation vs simulation comparison frame work What could be a Model.

Mireille Louys et al., OV-France Theory WG meeting , LYON, June 2007 2

Motivation

• Standardize description of numerical codes used to modelize astronomical objects

• Specify all the information required by a user to run the code: inputs, outputs, control parameters

• Adjust Simulation codes parameters in iterations for simulated and observed data comparisons and interpretation

Page 3: Mireille Louys et al., OV-France Theory WG meeting, LYON, June 2007 1 DALIA : an observation vs simulation comparison frame work What could be a Model.

Mireille Louys et al., OV-France Theory WG meeting , LYON, June 2007 3

A generic framework for modeling

Graphical User Interface

(Web Service)

modelcodes

Observations

spectroscopicdata

Best fit parameters

with error bars

Optimization engine

Use existing formatsand visualization tools

Define a standard

Query existing databases

Use existing algorithms

Page 4: Mireille Louys et al., OV-France Theory WG meeting, LYON, June 2007 1 DALIA : an observation vs simulation comparison frame work What could be a Model.

Mireille Louys et al., OV-France Theory WG meeting , LYON, June 2007 4

DALIA

Page 5: Mireille Louys et al., OV-France Theory WG meeting, LYON, June 2007 1 DALIA : an observation vs simulation comparison frame work What could be a Model.

Mireille Louys et al., OV-France Theory WG meeting , LYON, June 2007 5

DALIA within the VOOld data from archives

New observations

Old optimalmodel instancefrom archives

Model code (algorithm)from archives

Page 6: Mireille Louys et al., OV-France Theory WG meeting, LYON, June 2007 1 DALIA : an observation vs simulation comparison frame work What could be a Model.

Mireille Louys et al., OV-France Theory WG meeting , LYON, June 2007 6

DALIA

Page 7: Mireille Louys et al., OV-France Theory WG meeting, LYON, June 2007 1 DALIA : an observation vs simulation comparison frame work What could be a Model.

Mireille Louys et al., OV-France Theory WG meeting , LYON, June 2007 7

Code Example

• a code that simulates observations of a galactic disk– it takes some parameters as input – It produces a data cube (position, position, velocity)

of flux density for a given molecular or atomic transition (e.g. Halpha, CO(1-0),...)

• To allow a user (or an application) to use the code (run it and read the outputs) the code “provider” needs to describe the various kind of metadata

Page 8: Mireille Louys et al., OV-France Theory WG meeting, LYON, June 2007 1 DALIA : an observation vs simulation comparison frame work What could be a Model.

Different types of information required

• Inputs related to the astrophysical source inclination, size of disk etc...

• Inputs related to the algorithm used in the codenumber of particles

• Inputs related to instrumental effects psf, sampling, etc ...

• Inputs related to general knowledge in physicsfrequency(ies) of the transition(s) considered

(Halpha or CO etc...) for the emission

• Output dataset(s)number of axes, nature of axes, ...

Characterisation or Spectrum data models

Simulation Model

Characterisation , Lev 1-4

Atomic Line data model

Simulation

Page 9: Mireille Louys et al., OV-France Theory WG meeting, LYON, June 2007 1 DALIA : an observation vs simulation comparison frame work What could be a Model.

Mireille Louys et al., OV-France Theory WG meeting , LYON, June 2007 9

Parameter Model Requirements

• To structure the parameter set (Sections, Subsections...) --> The code provider may structure the parameters according to their physical meaning to help the user to catch the physics behind

• Allow for dynamical description – Hierarchy (a parameter determines a subset of parameters)– Tables (several parameters can be vectors of variable length)

• Allow information on the numerical code (program) to be included (native language, variable names used in the code)--> help user/application to implement protocol for data exchange with code (generate pieces of codes in the native language)

Page 10: Mireille Louys et al., OV-France Theory WG meeting, LYON, June 2007 1 DALIA : an observation vs simulation comparison frame work What could be a Model.

The UML Schema

Objects to describedynamical parameter sets

Objects to structure the description ofparameter sets

Page 11: Mireille Louys et al., OV-France Theory WG meeting, LYON, June 2007 1 DALIA : an observation vs simulation comparison frame work What could be a Model.

Three levels of documents Level Level

description Description file Who is

preparing itHow it is

used

Parameter collection

List of parameter values for one single code execution

nuagemoleculaire.dal generated via the GUI by

the interface generator of Dalia

By the user , via a gui or by editing a template xml document

Launched by either a user or a workflow

Xml template document for

one code

Structure and description of code parameters for one code

(tree + table combinations)

nuagemoleculaire.xml The « Code provider » when publishing the code in a registry 

By the Dalia interface generator application for instance

General XML schema

Overall structure of possible parameters in any simulation code

astronumericalcodes.xsd

Up to the Theory IG to validate and/or generalise it

By a parser to validate new descriptions of simulation codes

Page 12: Mireille Louys et al., OV-France Theory WG meeting, LYON, June 2007 1 DALIA : an observation vs simulation comparison frame work What could be a Model.

Structured parameter set

<section> <title>Situation</title> <subSection> <title>Coordinates</title> <param xsi:type="amns:ParamCoor"> <name>RA</name> <varName>ra</varName> <unit>hh:mm:s</unit> <description>Right Ascension of dynamic center</description> <default>-10:40:20.00</default> < /param> <param xsi:type="amns:ParamCoor"> <name>Dec</name> <varName>dec</varName> <unit>dd:mm:s</unit> <description>Declination of dynamic center</description> <default>20:40:00.00</default> </param>

UML XML INSTANCE DALIA GUI

Page 13: Mireille Louys et al., OV-France Theory WG meeting, LYON, June 2007 1 DALIA : an observation vs simulation comparison frame work What could be a Model.

<choiceTree> <name>Potential type</name> <varName>pottype</varName> <description>Potential shape</description> <choiceGroup> <choiceval>Logarithmic</choiceval> <param xsi:type="amns:ParamFloat"> <name> Length scale </name> <varName>rp</varName> <unit>pc</unit> <description>Length scale of potential</description> <default>200</default> </param> <param xsi:type="amns:ParamFloat"> <name>Velocity scale</name> <varName>vp</varName> <unit>km/s</unit> <description> velocity scale of potential </description> <default>100</default> </param> </choiceGroup> <choiceGroup implemented="true"> <choiceval>Plummer</choiceval> <param xsi:type="amns:ParamFloat"> <name>Length scale</name>

Hierarchy in a dynamical parameter set

UML XML INSTANCE DALIA GUI

Page 14: Mireille Louys et al., OV-France Theory WG meeting, LYON, June 2007 1 DALIA : an observation vs simulation comparison frame work What could be a Model.

<parvect> <name>Radial distribution</name> <description>Radial distribution of matter</description> <ncolVarName>nur</ncolVarName> <colName>radius</colName> <param xsi:type="amns:ParamFloat"> <name>radius</name> <varName>uradii</varName> <unit>pc</unit> <description>Radius in pc</description> <default>100.</default> </param> <param xsi:type="amns:ParamFloat"> <name>Col. Dens.</name> <varName>udens</varName> <unit>10^22 cm-2</unit> <description>Comlumn density</description> <default>1.</default> </param> <param xsi:type="amns:ParamFloat"> <name>Vel. Disp.</name> <varName>udisp</varName> <unit>km/s </unit> <description>Velocity dispersion </description> <default>10.</default> </param> <parvect>

Dynamical table of parametersUML XML INSTANCE DALIA GUI

Page 15: Mireille Louys et al., OV-France Theory WG meeting, LYON, June 2007 1 DALIA : an observation vs simulation comparison frame work What could be a Model.

Mireille Louys et al., OV-France Theory WG meeting , LYON, June 2007 15

Perspectives

• Commonalities with the Simulation model • Distribute the intermediate level (mybeautifulmodel.xml) + an interface

generator• Share « instances » of models in the sense of an instance of parameter set:

– e.g. result of a model fit = values of the best fit parameters --> need to define a data model for instances of parameter sets

– How do we allow simple identification of output matching between snapshot and observation ?

– Jargon translation ?

==> how to share theoretical knowledge at the same level as the observations