Config:SequentialDesign

From SUMOwiki
(Redirected from Config:SampleSelector)
Jump to: navigation, search

Generated for SUMO toolbox version 7.0. We are well aware that documentation is not always complete and possibly even out of date in some cases. We try to document everything as best we can but much is limited by available time and manpower. We are are a university research group after all. The most up to date documentation can always be found (if not here) in the default.xml configuration file and, of course, in the source files. If something is unclear please dont hesitate to ask.

Contents

SampleSelector

empty

Don't select any new samples, useful when modeling multiple outputs, and you don't want to involve one of these outputs in the sampling process.

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

random

Selects new samples randomly in the design space.

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

delaunay

This sample selector uses a Delaunay triangulation of the data to select samples in locations far from previous samples, or in locations where the estimated model error is largest. This algorithm uses QHull, which is very slow for high dimensions, so you should only use this sample selector for less than 6D and for less than 1000 samples.

<SampleSelector type="PipelineSampleSelector" combineOutputs="false">
 
   <CandidateGenerator type="DelaunayCandidateGenerator"/>
 
        <CandidateRanker type="modelDifference">
           <Option key="criterion_parameter" value="2"/>
        </CandidateRanker>
        <CandidateRanker type="delaunayVolume"/>
 
        <MergeCriterion type="WeightedAverage" weights="[1 1]"/>
 
</SampleSelector>

density

A space-filling sampling algorithm which uses an approximation of the Voronoi tessellation of the design space. Will only sample within the "allowed" areas if constraints are specified.

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

error

An adaptive sample selection algorithm (error based), driven by the evaluation of your model on a dense grid, which selects samples in locations where the model error is estimated to be the largest.

<SampleSelector type="PipelineSampleSelector" combineOutputs="false">
 
   <CandidateGenerator type="GridCandidateGenerator"/>
 
   <CandidateRanker type="modelDifference">
           <Option key="criterion_parameter" value="4"/>
        </CandidateRanker>
 
        <MergeCriterion type="ClosenessThreshold">
 
      <!-- Closeness threshold, Double -->
      <Option key="closenessThreshold" value="0.05"/>
      <!-- Set a % of the maximumSamples to randomly chosen -->
      <Option key="randomPercentage" value="20"/>
 
      <Option key="debug" value="off"/>
   </MergeCriterion>
</SampleSelector>

lola-voronoi

A highly adaptive sampling algorithm which performs a trade-off between exploration (filling up the design space as equally as possible) and exploitation (selecting data points in highly nonlinear regions). lola-voronoi is the only sample selector which currently supports multiple outputs, auto-sampled inputs and constraints.

<SampleSelector type="LOLAVoronoiSampleSelector" combineOutputs="false">
   <!-- Integer between 2 and 20 -->
   <Option key="neighbourhoodSize" value="2"/>
   <!-- Number of frequency values returned for each submitted sample. Only used with auto-sampled inputs. -->
   <Option key="frequencies" value="11"/>
</SampleSelector>

rationalPoleSupression

A sampling algorithm aimed at supressing poles in rational models by sampling them (only for Rational models)

<SampleSelector type="OptimizeCriterion" combineOutputs="false">
 
   <!-- This criterion has to be solved to choose new samples, one can choose the optimizer used here -->
   <Optimizer>patternsearch</Optimizer>
 
   <CandidateRanker type="rationalPoleSupression" scaling="none"/>
   <CandidateRanker type="modelDifference" scaling="none"/>
 
   <!--
   when debug is 'on' a contour plot of the criterion function is drawn every iteration.
   Together with the current samples and the chosen samples
   -->
   <Option key="debug" value="off"/>
</SampleSelector>

expectedImprovement

A sampling algorithm aimed at optimization problems (only for Kriging and RBF)

<SampleSelector type="OptimizeCriterion" combineOutputs="false">
 
   <!-- This criterion has to be solved to choose new samples, one can choose the optimizer used here -->
   <Optimizer>patternsearch</Optimizer>
 
   <CandidateRanker type="expectedImprovement" scaling="none">
   </CandidateRanker>
   <CandidateRanker type="maxvar" scaling="none"/>
 
   <!--
   when debug is 'on' a contour plot of the criterion function is drawn every iteration.
   Together with the current samples and the chosen samples
   -->
   <Option key="debug" value="off"/>
</SampleSelector>

extremaLOLA

LOLA-Voronoi sample selector supplemented with 1 sample at the minimum and maximum

<SampleSelector type="CombinedSampleSelector" combineOutputs="false">
   <!-- A highly adaptive sampling algorithm, error and density based -->
   <SampleSelector weight="0.8">lola-voronoi</SampleSelector>
   <SampleSelector weight="0.1">sampleMinimum</SampleSelector>
   <SampleSelector weight="0.1">sampleMaximum</SampleSelector>
 
   <MergeCriterion type="ClosenessThreshold">
 
      <!-- Closeness threshold, Double -->
      <Option key="closenessThreshold" value="0.05"/>
      <!-- Set a % of the maximumSamples to randomly chosen -->
      <Option key="randomPercentage" value="0"/>
 
      <Option key="debug" value="off"/>
   </MergeCriterion>
</SampleSelector>

sampleMinimum

Selects one sample at the minimum of the model.

<SampleSelector type="OptimizeCriterion" combineOutputs="false">
   <Optimizer>patternsearch</Optimizer>
   <CandidateRanker type="minmodel" scaling="none"/>
</SampleSelector>

sampleMaximum

Selects one sample at the maximum of the model.

<SampleSelector type="OptimizeCriterion" combineOutputs="false">
   <Optimizer>patternsearch</Optimizer>
   <CandidateRanker type="maxmodel" scaling="none"/>
</SampleSelector>

default

LOLA sample selector combined with error based sample selector

<SampleSelector type="CombinedSampleSelector" combineOutputs="false">
   <SampleSelector weight="0.7">lola-voronoi</SampleSelector>
   <SampleSelector weight="0.3">error</SampleSelector>
 
   <MergeCriterion type="ClosenessThreshold">   
      <!-- Closeness threshold, Double -->
      <Option key="closenessThreshold" value="0.05"/>
      <!-- Set a % of the maximumSamples to randomly chosen -->
      <Option key="randomPercentage" value="0"/>
 
      <Option key="debug" value="off"/>
   </MergeCriterion>
</SampleSelector>
Personal tools