Config:SequentialDesign

From SUMOwiki
Revision as of 11:05, 24 April 2008 by Kcrombec (talk | contribs)

SampleSelector

empty

Dont select any new samples, useful when modeling multiple outputs in parallel.

<SampleSelector type="EmptySampleSelector" combineOutputs="false"/>

random

Each sampling iterations new samples are selected randomly.

<SampleSelector type="RandomSampleSelector" combineOutputs="false"/>

combo

Allows you combine multiple sample selector algorithms.

<SampleSelector type="CombinedSampleSelector" combineOutputs="false">
   <SampleSelector type="RationalPoleSuppressionSampleSelector" combineOutputs="false">
      <!-- Currently no options are available, if the model is a rational model, and a
       - - pole is found (using a simple search strategy), the pole is returned -->
   </SampleSelector>
 
   <SampleSelector type="DelaunaySampleSelector" combineOutputs="false">
      <!-- One of all, data -->
      <Option key="sampleSelect" value="all"/>
      <!-- Integer between 2 and 20 -->
      <Option key="nLastModels" value="2"/>
      <!-- One of densityBased, differenceBased, weightedLinear, weightedGeometric -->
      <Option key="scoreFunction" value="weightedLinear"/>
      <!-- Weighting for weightedLinear -->
      <Option key="lambda" value="0.5"/>
      <!-- Weighting for weightedGeometric -->
      <Option key="mu" value="0.5"/>
      <!-- One of none, max, cap, capmax -->
      <Option key="volumeScaling" value="max"/>
      <Option key="differenceScaling" value="capmax"/>
      <!-- Boolean flag, if set all points closer than snapThreshold to the boundary of 
       - - the domain are clipped to the boundary -->
      <Option key="snapToEdge" value="enable"/>
      <Option key="snapThreshold" value=".05"/>
   </SampleSelector>
</SampleSelector>

delaunay

An adaptive sample selection algorithm that does a trade-off between error and density.

<SampleSelector type="DelaunaySampleSelector" combineOutputs="false">
   <!-- One of all, data -->
   <Option key="sampleSelect" value="all"/>
   <!-- Integer between 2 and 20 -->
   <Option key="nLastModels" value="2"/>      
   <!-- One of densityBased, differenceBased, weightedLinear, weightedGeometric -->
   <Option key="scoreFunction" value="weightedLinear"/>
   <Option key="lambda" value="0.5"/>
   <Option key="mu" value="0.5"/>
   <!-- One of none, max, cap, capmax -->
   <Option key="volumeScaling" value="max"/>
   <Option key="differenceScaling" value="capmax"/>
   <!-- Boolean flag -->
   <Option key="snapToEdge" value="enable"/>
   <Option key="snapThreshold" value=".2"/>
</SampleSelector>

density

A simple density based sample selection algorithm.

<SampleSelector type="DensitySampleSelector" combineOutputs="false"/>

error

An adaptive sample selection algorithm (error based), driven by the evaluation of your model on a dense grid.

<SampleSelector type="ErrorSampleSelector" combineOutputs="false">
   <!-- Integer between 2 and 20 -->
   <Option key="nLastModels" value="4"/>      
   <!-- One of none, max, cap, capmax -->
   <Option key="differenceScaling" value="none"/>
   <!-- Gridsize to evaluate on, one of int or array of dimension length -->
   <Option key="gridSize" value="50"/>
   <!-- Maximum total points to evaluate, distributed over dimensions -->
   <Option key="maxGridSize" value="100000"/>
   <!-- Closeness threshold, Double -->
   <Option key="closenessThreshold" value="0.2"/>
   <!-- Set a % of the maximumSamples to randomly chosen -->
   <Option key="randomPercentage" value="20"/>
</SampleSelector>

gradient

A highly adaptive sampling algorithm, error and density based.

<SampleSelector type="GradientSampleSelector" combineOutputs="false">
   <!-- Integer between 2 and 20 -->
   <Option key="neighbourhoodSize" value="2"/>
</SampleSelector>

isc

A sampling algorithm aimed at optimization problems.

<SampleSelector type="InfillSamplingCriterion" combineOutputs="false">
   <!-- A criterion determines the interesting regions to sample -->
   <!-- Choose 1 from the following: -->
   <Option key="criterion" value="gei"/> <!-- generalized expected improvement -->
   <Option key="g" value="1"/> <!-- balanced local-global search -->
 
   <!--<Option key="criterion" value="wei" />
--> <!-- weighted expected improvement -->
   <!--<Option key="w" value="0.5" />
--> <!-- weight, 0 is global search, 1 is local search -->
 
   <!--<Option key="criterion" value="ei" />
--> <!-- expected improvement -->
 
   <!--<Option key="criterion" value="kushner" />
--> <!-- kushner -->
   <!--<Option key="eps" value="0.001" />
-->
 
   <!--<Option key="criterion" value="lcb" />
--> <!-- lower confidence bound function -->
   <!--<Option key="lcb" value="2.0" />
-->
   <!--<Option key="criterion" value="maxvar" />
--> <!-- maximizes variation -->
 
   <!-- Watson and Barnes criterions -->
   <!--<Option key="criterion" value="wb1" />
--> <!-- threshold-bounded extreme -->
   <!--<Option key="criterion" value="wb2" />
--> <!-- regional extreme -->
 
   <!--<Option key="criterion" value="crowdedness" />
--> <!-- crowdedness function -->
 
   <!-- This criterion has to be solved to choose new samples, one can choose the optimizer used here -->
   <Optimizer type="DirectOptimizer">
      <Option key="maxevals" value="1000"/>
      <Option key="maxits" value="300"/>
   </Optimizer>
 
   <!--
   <Optimizer type="MatlabGA">

   </Optimizer>
   -->
 
   <!--
   when debug is 'on' a contour plot of the ISC function is drawn every iteration.
   Together with the current samples and the chosen samples
   -->
   <Option key="debug" value="on"/>
</SampleSelector>