SensiPlot: Difference between revisions

Revision as of 04:18, 20 May 2023

Screenshot of a SensiPlot run with two parameters

Description

SensiPlot is not an optimization algorithm but instead can be used to carry out a sensitivity analysis by varying one or more optimization parameters.

Required input files

same as with PES

Usage

Example of different parameter sampling modes for two parameters

Select an App, a Dataset and then select "SensiPlot" as the Method.
Select one or more Optimization parameters to vary.
Select an Objective function to show in the main diagram.
Select a Mode:
- Randomly distributed: Each parameter combination is completely random
- Evenly distributed: All possible parameter combinations are tested (requires n_steps ^ n_params simulations!)
- Latin Hypercube Sampling: Uses Latin hypercube sampling to generate a near-random sample of parameter values that attempts to cover the entire parameter space.
Set the No. of steps: Number of different values to test for each parameter.
The option Show time series in Wave will display all simulation results in BlueM.Wave once the SensiPlot run is finished.
The option Save individual datasets stores each simulated dataset in a subfolder named sensiplot_<N> of the dataset.

The main diagram will only show up to two optimization parameters and the one selected objective function. To see the values of other optimization parameters and objective functions, use the scatterplot and custom plot features (buttons in the toolbar).

Notes

Optimization parameters that are not selected will be set to their start values as defined in the OPT-file.
Constraints are currently not considered (#177)

@@ Line 2: / Line 2: @@
 __NOTOC__
 {{BlueM.Opt nav}}
-[[Bild:SensiPlot screenshot.png|thumb|Screenshot einer SensiPlot Berechnung mit zwei Parametern]]
+[[Bild:SensiPlot screenshot.png|thumb|Screenshot of a SensiPlot run with two parameters]]
-==Beschreibung==
+==Description==
-'''SensiPlot''' führt eine Sensitivitätsanalyse durch und kann dazu benutzt werden, die Sensitivität von einem oder zwei Optimierungsparametern hinsichtlich einer Zielfunktion zu untersuchen.
+'''SensiPlot''' is not an optimization algorithm but instead can be used to carry out a sensitivity analysis by varying one or more optimization parameters.
-==Erforderliche Eingabedateien==
+==Required input files==
-wie bei [[PES]]
+same as with [[PES]]
-==Anwendung==
+==Usage==
-# ''Anwendung'', ''Datensatz'' und ''Methode'' ("SensiPlot") auswählen. Es erscheint ein Dialogfenster.
+[[File:SensiPlot ParameterSampling.png|thumb|right|Example of different parameter sampling modes for two parameters]]
-# Einen oder zwei ''Optimierungsparameter'' auswählen <br/>(1 Parameter &rarr; Punkt-Diagramm; 2 Parameter &rarr; Oberflächendiagramm)
+# Select an ''App'', a ''Dataset'' and then select "SensiPlot" as the ''Method''.
-# Eine ''Zielfunktion'' auswählen
+# Select one or more ''Optimization parameters'' to vary.
-# Auswählen, ob der/die Optimierungsparameter ''diskret'' oder ''gleichverteilt'' variiert werden sollen, und in welcher Auflösung (''Anzahl Schritte'')
+# Select an ''Objective function'' to show in the main diagram.
-# Die Option ''Wave anzeigen'' führt dazu, dass im Anschluss an SensiPlot alle durchgeführten Simulationen in [[Wave]] dargestellt werden (Vorsicht bei einer größeren Anzahl von Schritten!)
+# Select a ''Mode'':
+#* ''Randomly distributed'': Each parameter combination is completely random
+#* ''Evenly distributed'': All possible parameter combinations are tested (requires <code>n<sub>steps</sub> ^ n<sub>params</sub></code> simulations!)
+#* ''Latin Hypercube Sampling'': Uses [https://en.wikipedia.org/wiki/Latin_hypercube_sampling Latin hypercube sampling] to generate a near-random sample of parameter values that attempts to cover the entire parameter space.
+# Set the ''No. of steps'': Number of different values to test for each parameter.
+# The option ''Show time series in Wave'' will display all simulation results in [[BlueM.Wave]] once the SensiPlot run is finished.
+# The option ''Save individual datasets'' stores each simulated dataset in a subfolder named <code>sensiplot_<N></code> of the dataset.
-==Hinweise==
+The main diagram will only show up to two optimization parameters and the one selected objective function. To see the values of other optimization parameters and objective functions, use the scatterplot and custom plot features (buttons in the toolbar).
-* Bei jeder neuen Parameterkombination werden die Modellparameter auch für die nicht ausgewählten OptParameter in die Modell-Eingabedateien geschrieben, und zwar mit den in der [[OPT-Datei]] angegebenen Startwerten.
-* Constraints werden derzeit nicht berücksichtigt (Bug 253)
-* Lösungsauswahl im Diagramm momentan nicht möglich (Bug 327) - anstatt dessen muss die Lösungsdatenbank verwendet werden
-[[Kategorie:BlueM.Opt Anwendung]]
+==Notes==
+* Optimization parameters that are not selected will be set to their start values as defined in the [[OPT-file]].
+* Constraints are currently not considered ([https://github.com/bluemodel/BlueM.Opt/issues/173 #177])
+[[Category:BlueM.Opt]]