Citation
Beth Osyk, Thomas Mandl, Man-Kit Leung. "Model Ontologies: Generalizing Type System Ideas to Arbitrary Model". Talk or presentation, 16, April, 2009; Presented at the 8th Biennial Ptolemy Miniconference.

Abstract
The Ptolemy Hierarchical Orthogonal Multi-Attribute Solver (PtHOMAS) project studies data ontologies in a model by generalizing type system ideas to arbitrary model properties. In particular, we present a static analysis method which extends the efficient type inference engine in Ptolemy II. Model components such as actors, finite state machines, and expressions typically have properties, either for the component itself or its communication channels (ports, for example). A simple example for a port property is whether its incoming/outgoing signal is constant or not. Unfortunately, while property information is valuable to the designer, most components have context-dependent properties (e.g. connections, feedback-loops), which makes it hard for a designer to evaluate them. In order to enable a correct, efficient and large scale property annotation an automatic inference algorithm is required.

By framing this problem mathematically, the Ptolemy II inference engine can automatically infer properties wherever not explicitly specified. The properties are related as a lattice (for example, constant < nonconstant). The relationships within and between components are expressed as a list of inequalities relative to the lattice. For example, inside one actor with a single output port, outputPort ≤ inputPort for all inputPort, which results in the output being nonconstant if any input is nonconstant. Naturally, the same approach can be used to verify models with respect to given properties.

The presentation gives some insight into the basic concepts, a rough overview of the implemented framework in Ptolemy II, and concludes with a concrete demo for constant propagation.

Electronic downloads

• 10-Latronico-PtolemyHierarchicalOrthogonalMultiAttributeSolver-PtHOMAS_Apr09.ppt · application/vnd.ms-powerpoint · 1056 kbytes

• HTML

  Beth Osyk, Thomas Mandl, Man-Kit Leung. <a
  href="http://chess.eecs.berkeley.edu/pubs/558.html"
  ><i>Model Ontologies: Generalizing Type System
  Ideas to Arbitrary Model</i></a>, Talk or
  presentation,  16, April, 2009; Presented at the 8th
  Biennial Ptolemy Miniconference.

• Plain text

  Beth Osyk, Thomas Mandl, Man-Kit Leung. "Model
  Ontologies: Generalizing Type System Ideas to Arbitrary
  Model". Talk or presentation,  16, April, 2009;
  Presented at the 8th Biennial Ptolemy Miniconference.

• BibTeX

  @presentation{OsykMandlLeung09_ModelOntologiesGeneralizingTypeSystemIdeasToArbitrary,
      author = {Beth Osyk and Thomas Mandl and Man-Kit Leung},
      title = {Model Ontologies: Generalizing Type System Ideas
                to Arbitrary Model},
      day = {16},
      month = {April},
      year = {2009},
      note = {Presented at the 8th Biennial Ptolemy
                Miniconference},
      abstract = {The Ptolemy Hierarchical Orthogonal
                Multi-Attribute Solver (PtHOMAS) project studies
                data ontologies in a model by generalizing type
                system ideas to arbitrary model properties. In
                particular, we present a static analysis method
                which extends the efficient type inference engine
                in Ptolemy II. Model components such as actors,
                finite state machines, and expressions typically
                have properties, either for the component itself
                or its communication channels (ports, for
                example). A simple example for a port property is
                whether its incoming/outgoing signal is constant
                or not. Unfortunately, while property information
                is valuable to the designer, most components have
                context-dependent properties (e.g. connections,
                feedback-loops), which makes it hard for a
                designer to evaluate them. In order to enable a
                correct, efficient and large scale property
                annotation an automatic inference algorithm is
                required. <p>By framing this problem
                mathematically, the Ptolemy II inference engine
                can automatically infer properties wherever not
                explicitly specified. The properties are related
                as a lattice (for example, constant <
                nonconstant). The relationships within and between
                components are expressed as a list of inequalities
                relative to the lattice. For example, inside one
                actor with a single output port, outputPort ≤
                inputPort for all inputPort, which results in the
                output being nonconstant if any input is
                nonconstant. Naturally, the same approach can be
                used to verify models with respect to given
                properties. <p> The presentation gives some
                insight into the basic concepts, a rough overview
                of the implemented framework in Ptolemy II, and
                concludes with a concrete demo for constant
                propagation.},
      URL = {http://chess.eecs.berkeley.edu/pubs/558.html}
  }
  

Posted by Christopher Brooks on 17 Apr 2009.
Groups: ptolemy
For additional information, see the Publications FAQ or contact webmaster at chess eecs berkeley edu.