Explicit Composition Constructs in DSLs - The case of the epidemiological language Kendrick
- 1. Explicit Composition Constructs in DSLs The case of the epidemiological language Kendrick Bùi Thị Mai Anh, Nick Papoulias, Mikal Ziane Serge Stinckwich IWST 2016, Prague
- 2. Cameroon Université de Yaoundé 1 Vietnam IFI-VNU & USTH Morocco LMDP, Université Cadi Ayyad de Marrakech Senegal UGB (Université Gaston Berger de Saint- Louis) Senegal UCAD (Université Cheikh Anta Diop de Dakar) 65 members + Phd Students: Professors, Researchers, Associates (26 HDR),1 research engineer, 3 admin staff, 3 post-docts and 45 PhD students Texte International Research Unit UMMISCO Mathematical and Computational Modeling of Complex Systems Laboratory
- 4. UMMISCO East-Central Africa Unit (Yaoundé) Research Activities on Model-Driven Epidemiology 1 Professor (Maurice Tchuenté) 2 researchers from UMMISCO/Bondy (3 months in 2016) 6 PhD students EPICAM KENDRICK Epidemiological Surveillance System Modeling&Simulation of Epidemiology Models Process Mining Ontologies Domain- Specific Languages and Tools Model- Driven Engineering Multi-Agent Simulation
- 5. https://github.com/UMMISCO/EPICAM EPICAM is a Model-Driven Engineering Platform for Epidemiological Surveillance System Deployed on 25 sites around Cameroon 3 PhD students from Yaoundé 1 University involved Applied to Tuberculosis but adaptable to other diseases Model-Driven Engineering that allows automatic generation of the platform
- 6. http://ummisco.github.io/kendrick/ Kendrick is a platform for epidemiological modeling and analysis Visualizations It helps epidemiologists craft custom analyses cheaply BrowsersModels 2 PhD students from Yaoundé 1 University involved
- 7. What is Epidemiology Modeling ? • Building mathematical models to study speed of a disease in a population • Compartmental models
- 9. SIR Model
- 10. Models Simulations • 3 ways to do simulations: • Population-level: deterministic simulation (ODEs solver) • Individual-Level: stochastic simulation (Gillepsie simulation) • Agent-based level
- 11. Multi-concerns Models of Epidemiology • Seasonality • Multi-hosts • Multi-strains • Age/Risk structure • Spatial aspect • Control strategies
- 14. Spatial concerns
- 16. How epidemiologists are building their models ?
- 17. SEIRS model
- 22. Concerns are scattered Concerns are tangled
- 23. Concerns are scattered Concerns are tangled Not modular, not reusable
- 24. Separation of Concerns in Epidemiology Modeling • Decompose highly-coupled monolithic models into modular concerns • define concerns with as few dependencies as possible • Combine concerns as freely as possible
- 25. Solution • Generic mathematical meta-model that provides abstractions to define epidemiological concerns • Models are expressed as Stochastic Automata • Composition operator (tensor sum) • Transforms concerns to modify automata
- 26. Tensor Composition of a spatial concern with SIR model
- 27. Kendrick DSL • Implemented the generic mathematical meta- models in Smalltalk • Embedded DSL in Pharo • Allows definition and composition of concerns • https://github.com/UMMISCO/kendrick
- 30. Kendrick Demo
- 31. If you want to try Kendrick download a MOOSE 6.0 image from http://agilevisualization.com/
- 32. Participants • Fabrice Atrevi (IFI-VN) • Bui Thi Mai Anh (UMMISCO-VN) • Ho Tuong Vinh (UMMISCO-VN) • Cheikhou Oumar Ka (UMMISCO-Senegal) • Nick Papoulias (UMMISCO-France) • Benjamin Roche (UMMISCO-France) • Aboubakar Sidiki (UMMISCO-Cameroon) • Serge Stinckwich (UMMISCO-France) • Mikal Ziane (LIP6, UPMC)
- 33. Conclusions • Kendrick is an embedded DSL for epidemiology modeling/simulations that promote separation of concerns • Chapter 15 in “Agile Visualisation Book”: http:// agilevisualization.com/ • We rely a lot on the Pharo community: PolyMath, Roassal, PetitParser, Moose, STon, SmalltalkCI
- 34. Ongoing Work • GPU (VirtualGPU) implementation of stochastic algorithms - Cheik Oumar Ka • Network concerns - Aboubakar Sidiki (April-May 2016) • Metamorphic tests of Epi models - Herman Mekontso Tchinda • User eXperiments (in collaboration with Nick Papoulias during 2017)