==> identify commonalities from the
peculiarities to "factorize" ,
- maximize modularity
Decomposing too much a problem may lead to poor performances.
It may be necessary to find compromizes:
Adopted design:
Some technical detail about implementations:
Archiving in XML
Contexts and Perspectives:
The model is intended for radio-telescopes in general (single-dish, interferometers).
It is migrating from the ALMA SDM (ASDM) to the SDM (the plan is to use it for the eVLA).
Toward a standard for the radioastronomy? (VO?).
The ALMA project must be able to provide data using an export data format (TBD).
R&D is required to extend the SDM to be able to manage the informations contained in the astronomical data themselves at the level of the database itself.
The most direct approach is to search for schemes in the Fourier domain for the cross-correlations for the observed visibilities and the image-plane for the auto-correlations.
It requires extending the relational model for similarity queries (as opposed to exact matching) to be able to handle scoring functions.
For some schemes extensive simulations will be required to help understanding the high level of abstraction in the data characterization.
Example using the brightness distribution of a simulated compressive turbulence
Consider the distribution of the phase increments and define the entropy in its distribution:
The entropy for a Kolmogorov distribution is:
The amount of structure present in the distribution could be defined by:
The histogram of the phase increment can be fitted by a van Mises distribution:
In this context this quantity of structure is given by (F. Levrier):
and we may consider to introduce this as an attribute of the data model, a property of the cross-correlation data themselves at some level in the project structure.
Francois Viallefond Last modified: Wed May 11 12:34:40 CEST 2005
==> push to strong cohesion in the entities and to loose coupling between
them.
It may be necessary to find compromizes:
- how frequently the data will be used for different views?
- which parts of the data base require maintenance?
- etc...
Adopted design:
- It is a relational model extended to handle objects (collections)
- The principle of unicity is applied
- All the relations have keys
- A key may be composed of more than one field
- The depth of the structures are provided using foreign keys
- An attribute in a relation may be optional
- The relations are normalized
- The parameter space is composed of:
- multi-dimensional structures (meta and auxiliary data)
- a tree structure (project structure)and
- a connector(flexible allowing reconfigurations at any time)).
- Allow 1:1 and 1:many relationships between the tables
Some technical detail about implementations:
- Reference document: a set of relational tables (key and data sections).
(Re-use MS model presentation).
- The model can be fully described using XML-Schema
(Can use standard RDBMS processors)
- The model can be fully described using an UML class diagram
- Prototype implementation using XSLT for automated code generation
- Now implemented from UML/ArchitectureWare for automated code generation
Archiving in XML
Contexts and Perspectives:
The model is intended for radio-telescopes in general (single-dish, interferometers).
It is migrating from the ALMA SDM (ASDM) to the SDM (the plan is to use it for the eVLA).
Toward a standard for the radioastronomy? (VO?).
The ALMA project must be able to provide data using an export data format (TBD).
Extension of the data model for the heuristics and data analysis:
The SDM can be considered as a classic relational model:- Standard management database operations
for data filtering and data combination can provide various views.
- The SDM attributes used in the
queries fall in the category of meta-data (and auxiliary data)
and the project meta-information.
R&D is required to extend the SDM to be able to manage the informations contained in the astronomical data themselves at the level of the database itself.
The most direct approach is to search for schemes in the Fourier domain for the cross-correlations for the observed visibilities and the image-plane for the auto-correlations.
It requires extending the relational model for similarity queries (as opposed to exact matching) to be able to handle scoring functions.
For some schemes extensive simulations will be required to help understanding the high level of abstraction in the data characterization.
Example:
Consider the distribution of the phase increments and define the entropy in its distribution:
The entropy for a Kolmogorov distribution is:
The amount of structure present in the distribution could be defined by:
The histogram of the phase increment can be fitted by a van Mises distribution:
In this context this quantity of structure is given by (F. Levrier):
and we may consider to introduce this as an attribute of the data model, a property of the cross-correlation data themselves at some level in the project structure.
Francois Viallefond Last modified: Wed May 11 12:34:40 CEST 2005