1 / 15

Ontology & Graphical Representations

Ontology & Graphical Representations. NET Meeting – Ottawa – July 7, 2010 Subproject report By Anya Okhmatovskaia. Current strategy. Build the ontology bottom-up: Examine existing POHEM models Identify common concepts, principles, relations

keziah
Télécharger la présentation

Ontology & Graphical Representations

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Ontology & Graphical Representations NET Meeting – Ottawa – July 7, 2010 Subproject report By Anya Okhmatovskaia

  2. Current strategy • Build the ontology bottom-up: • Examine existing POHEM models • Identify common concepts, principles, relations • Create a POHEM-specific ontology, later extend it • Sources of knowledge on POHEM models: • Existing model diagrams • POHEM code • “Data items” spreadsheets for Cancer models

  3. Graphical representation • Helpful exercise in organizing POHEM concepts: leverages ontology development • Progress since first version: • Attempted to separate causal view from temporal • More explicit and (hopefully) more consistent • Web-interactive • Prototype of automatic generator under way • Diagrams are descriptive: we do not intend to do reasoning over graphs

  4. Causal view Health Determinants Diseases Mortality Interventions RF interventions, Treatment Health Utility Costs Direct, Indirect Notation: Nodes = random vars Arcs = causal effect

  5. Causal view: Nodes • Nodes = random variables • Define an agent’s state* • Have an associated function used to compute the node’s value • regression model • look-up table • probabilistic mapping • Input nodes: • have no parent nodes • initialized from data • Different colors = categories of variables in POHEM * There’s no “system state in POHEM, but possible for other simulation models

  6. Causal view: Links • Links = causal effects • Have associated data • directly used in a function to compute a node • used in analysis to derive parameters • used for calibration • Suggested diagrams are not DAGs • Cycles in the graph reflect flattening temporal aspect Disease t1 Disease t2 Intervention t1 Intervention t1

  7. Encoding time Important dimensions: • constant vs. changing over time • time between updates • regular change: fixed intervals (e.g. BMI, Smoking) • dynamically computed time to event (e.g. AMI) • restrictions on temporal dynamics • unconstrained (e.g. BMI?) • monotonic increase/decrease (e.g. Age, Costs) • restricted state transitions (e.g. Disease progression, Death)

  8. Encoding time • State transitions is the most elaborate/detailed way to specify change over time • state diagrams define restrictions on transitions • for each state causal view can be different • It is possible to cast different kinds of temporal change to state transitions • not necessarily the best solution • Node’s temporal dynamics representation is related to variable type • continuous variables would have to be discretized to apply state transition restriction • parameterized functions are a more compact and complete specification

  9. Encoding time Health Determinants (regular change): categorical, logical, rank, numeric (discrete, continuous) etc … Diseases (restricted transitions): categorical Mortality (restricted transitions): logical (T/F) Interventions (possibly constant): logical (?) Health Utility: numeric (continuous) Costs (monotonic increase): numeric (continuous)

  10. Automatic diagrams So far for causal view only Ontology (OWL) Formatted graph (SVG) Java SVG translator or XSLT Protégé plug-in Graph definition file (DOT) GraphViz dot utility Graph image (GIF) Interactive diagram (HTML)

  11. Future work plan Goals & timeline

  12. Goals: Development 1. Ontology development • Refine POHEM-specific ontology • Encode existing POHEM models • Review literature and terminology • Extend/generalize ontology • Encode non-POHEM model(s) • Formally evaluate the ontology 2. Ontology applications • Automatic model documentation • Diagram generator ? Model consistency checking ? Literature search aid

  13. Goals: Training & KT 3. Training • Ontology development and evaluation:CS/Epi Master Thesis • Applications development:CS Master summer Project (1-2) 4. Publications • Conference papers: • initial version of generalized ontology • ontology evaluation • application(s) • Journal paper: Putting everything together

  14. Diagraming application summer project Onto-based documentation summer project Ontology evaluation complete Hire a student for ontology thesis (optimistic scenario) Terminology review done, 1st version of generalized ontology POHEM ontology refinement & model encoding complete Generalized ontology refinement, encoding of non-POHEM model(s) complete Timeline

  15. Conf. paper:initial description of ontology + POHEM examples Conf. paper:ontology application Conf. paper:ontology evaluation Timeline: Publications Diagraming application summer project Onto-based documentation summer project Ontology evaluation complete Hire a student for ontology thesis (optimistic scenario) Submission deadlines: AIME: January AMIA: March EKAW: April ISWC: June Terminology review done, 1st version of generalized ontology POHEM ontology refinement & model encoding complete Generalized ontology refinement, encoding of non-POHEM model(s) complete Journal paper:generalized ontology, examples, evaluation, applications

More Related