Date post: | 19-Jan-2016 |
Category: |
Documents |
Upload: | calvin-lucas |
View: | 212 times |
Download: | 0 times |
OwlViper:Web Ontology Language
Visual Programming Environment for Data
Reduction (of work)
Ed Shaya (PI), Brian Thomas, Peter Teuben, Zhenping Huang (UMD),
OTHERS??
Next Generation Analysis Software
Ingredients: OWL,OWL, Visual ProgrammingVisual Programming,, VOx, Registries, Grid, WebServices, WSDL,
COM, CORBA, Math Packages, existing scientific analysis software.
Functionality Visual Programming - Layer on top of “normal” code and
pipelines them using simple graphical user interface. No user programming or scripting required – Visual entry, as
in click and drag. Draw flow diagram in GUI. Diagram is converted to script,
script is converted to workflow, workflow is interactively managed as it is executed.
Easy to use Scientists’ application/environment to query the VOx distributed data centers, process using the Grid when required, analyze, visualize, and web-publish results.
Various levels of autonomy from manual creation of a workflow to “state the goal and let it rip” via OWL knowledge base.
To make it scientific it must handle unit conversions and error propagation as well.
The proverbial drawing board!
Components Infrastructure
OWL Analyzer Graphical User Interface (GUI) GUI-to-script-to-GUI converter Script-to-Workflow-to-Script converter Executor Data Hunter
Functions and Field Specifics OWL Ontology Callable code or relevant transformations
E.g. – FFT, deconvolution, regression, radiative transfer equation, coordinate transforms, visualization, etc. etc.
Ontology (OWL) Functions and data are placed in class structure (not a
strict hierarchy) plus properties or relationships.
Function
Data Extracter
TransformLogical Function
Local File Reader
ANDDatabase
Query
Coordinate Transform
Integral Transform
Filter
Constraint
output
hasA
Thing1 Thing2isOwnedby
Data object
inputData
objectFunction
Data Object Types Data requests by object plus constraint on
properties All datasets can be seen as sets of physical objects with
properties. Each row of a scientific table provides properties on object
specified by some id (name or id-number). DataType – Star with Brightness, Mass, and Position
Data with extra info can be cast to this datatype. Data with less info can be merged with other data to
create this datatype.
Star
Brightness=10.3 mag
Mass = 12.3 MsunsPosition:
RA = 12:34:43.65DE = 32o12’32”Equinox = 2000
CooSystem = ICRS
Scientists accept their errors Mass hasUnits restricted to MassUnits, hasValue(s),
hasError(s). Kilogram isInstanceOf Mass with prefix=kilo,
consistsOf=gram. ThisMass isanInstanceOf Mass. ThisMass hasUnits kilogram, hasValue=754.3, hasError=2.2
Leave no computer behind. Ontology is a means for knowledge to become computer-
understandable. It allows one to make simple statements:
Subtraction isaSubClassOf Arithmetic. <Class ID=”Subtraction>
<SubClassOf resource=”#Arithmetic”/>
</Class> Subtraction has initialValue range float or Measurement and
isReducedBy range float or Measurement. ThisSubtraction has InitialValue 3.1 and isReducedBy 2.2. MSubtraction isaSubClassOf Subtraction. MSubtraction has InitialValue restricted to Measurement. BMinusV isaSubClassOf MSubtraction with initialValue restricted to B-
Measurements and isReducedBy restricted to V-Measurements. ThisB has Value=18.6 and Error=0.5. ThisV has Value=17.3 and Error=0.34. ThisB-V has initialValue ThisB and isReducedBy ThisV. TheStar named=”alphaBoo” hasColor ThisB-V.
OWL Analyzer Find transforms by name or class-name or (xml)
description (output/input object type). Report closest matches (ontologically). Starting from goal data type, analyzer finds path
back to source data (either locally or in distributed databases). If distributed databases are used, then initial path
reported after registries are searched. Path may be constrained to pass through certain
functions or data types.
Graphical User Interface Drop down menus of Things each with drop down
of its properties. Celestial objects and measurable
Constraint Box for value range on measurable and error Function Menu by class or by output data type.
Entry form for values of flags and input parameters. Pieces flow together by simply adding arrows,
data to function to data to function. Items that don’t logically fit will not allow arrows to hook
in Can work backwards. Enter a final data object,
menus are reduced to only allowed functions. Choose function, menu reduced to only allowed data.
Under the Covers Script
GUI-to-Script, Script-to-GUI Save process for later reuse Publish process for others
A process can be named and added to the function ontology Some users like to script; provide command line.
Workflow (Work details) Script to Workflow, Workflow-to-script Branch abort Subfunctions
Low level database queries (from Data Hunter) Decision Trees
when to use the grid
Executor Script validator Time estimator Logic evaluator Basic mathematics at main level Error propagation Unit analysis, conversion Interactive mode Tests for infinite loops python based? Interfaces: C++, Java, WebServices, IDL, Matlab,
pyRAF, Ftools, Numerical Recipes…
Data Hunter Read Local Files. Check VOx registries for data resources. Download WSDL for services to resources. Convert WSDL to service call code. Transform high level query to service calls.
May require ontology support Add metadata to data if necessary.
Information in the registry or implicit in the query may not be in the data upon arrival. This can be automatically added. Provisions can be made for adding metadata manually.
Demonstrations Distances to Galaxies and Ho N-body simulation and comparison to
observations
Tully-Fisher Relation
Basic Astronomy Knowledge on Distance
hasTFDistance
disjointdisjoint
Parameter Set Instances for TF
The user is presented with these choices for transform. They may either opt to choose one they like, or edit the properties of this transform to create their own.
•The computer now has the following search list of Galaxy objects:SpiralGalaxy (DTF, Cz)SpiralGalaxy (I, W20, inclination, Cz)
Ho Flow on OWLViper CanvasData
Hunter Spiral GalaxyI(Tot)
21cm WidthInclination
Redshift
Best caseFilter
Weighted Average
Spiral GalaxyI(Tot)
21cm WidthInclination
Redshift
Tully-FisherRelation
Plotter
Distance
cz
Ho Table
Spiral GalaxyDistance
LinearRegression
Date Ho err
1-6-08 72.3 14.5
2-6-08 64.2 7.5
3-6-08 53.2 3.2
4-6-08 42.0 0.3