Embed Size (px)
Citation preview
gamar: an R interface to the GAMA agent-based simulation platform
Marc ChoisyLucie ContaminHo Bich HaiNicolas MarilleauJean-Daniel Zucker
population-based individual-based agent-baseddifferential equations Gillespie algorithms
Increased level of detail
deSolve
adaptivetauGillespieSSAdde
diffeqrPBSddesolve
bvpSolve
Agents are• autonomous • heterogeneous • active • adaptive
Population dynamics modeling
population-based individual-based agent-baseddifferential equations Gillespie algorithms
Increased level of detail
deSolve
adaptivetauGillespieSSAdde
diffeqrPBSddesolve
bvpSolve
Agents are• autonomous • heterogeneous • active • adaptive
Population dynamics modeling
RNetLogorrepastNetLogoR
NetLogo
Performance
Ease of use
GAMA
repast.github.io ccl.northwestern.edu/netlogo gama-platform.github.io
since 2007since 1999 since 2000
Agent-based modeling platforms
RNetLogorrepastNetLogoR
NetLogo
Performance
Ease of use
GAMA
repast.github.io ccl.northwestern.edu/netlogo gama-platform.github.io
since 2007since 1999 since 2000
Agent-based modeling platforms
• a language: GAML
• a user-interface
• a fast and parallelized engine
• integration with GIS data
https://gama-platform.github.io
The GAMA simulation platform
• a language: GAML
• a user-interface
• a fast and parallelized engine
• integration with GIS data
https://gama-platform.github.io
The GAMA simulation platform
.gaml
model
outputs
1. design model
2. load
model
3. design experiment plans
5. analyse simulations
4. run experiment
experiment object
The gamar R package
run_experiment()load_experiment()
fullfact(), repl()
makemovie()
.gaml
model
outputs
1. design model
2. load
model
3. design experiment plans
5. analyse simulations
4. run experiment
experiment object
The gamar R package
run_experiment()load_experiment()
fullfact(), repl()
makemovie()
.gaml
model
outputs
1. design model
2. load
model
3. design experiment plans
5. analyse simulations
4. run experiment
experiment object
The gamar R package
run_experiment()load_experiment()
fullfact(), repl()
makemovie()
several columns for parameters values - always first columns - names start with p_
several columns for monitored variables - just after the parameters columns - names start with r_
one column for simulations durations - always penultimate position - named tmax
one column for seeds values - always penultimate position - named seed
The experiment class
one list-column for simulations outputs - always last position - named output
/Users/gamar/models/sir/sir.gaml/Users/gamar/models/sir/output
outputNANANANANANA...
NANANANANANA
output<data.frame[1000,3]><data.frame[1000,3]><data.frame[1000,3]><data.frame[1000,3]><data.frame[1000,3]><data.frame[1000,3]>
.
.
.<data.frame[1000,3]><data.frame[1000,3]><data.frame[1000,3]><data.frame[1000,3]><data.frame[1000,3]><data.frame[1000,3]>
several columns for parameters values - always first columns - names start with p_
several columns for monitored variables - just after the parameters columns - names start with r_
one column for simulations durations - always penultimate position - named tmax
one column for seeds values - always penultimate position - named seed
The experiment class
one list-column for simulations outputs - always last position - named output
/Users/gamar/models/sir/sir.gaml/Users/gamar/models/sir/output
Visualizing simulation resultswith(sir$output[[1]], plot(Step, r_I, ylab = "nb. infected")
path_to_movie <- make_movie(pp$output[[1]], "r_main_display")
