Difference between revisions of "Main Page"
Line 4:  Line 4:  
 Contents = 
 Contents = 

The Surrogate Modeling Toolbox (SUMO Toolbox) is a Matlab toolbox that automatically builds accurate [http://en.wikipedia.org/wiki/Surrogate_model surrogate models] (also known as metamodels or [http://en.wikipedia.org/wiki/Response_surface_methodology response surface models]) of a given data source (simulation code, data set, script, ...) within the accuracy and time constraints set by the user. In doing so the toolbox minimizes the number of data points (which it chooses automatically) since they are usually expensive. The toolbox tries to be as adaptive and autonomous as possible, requiring no user input besides some initial configuration. 
The Surrogate Modeling Toolbox (SUMO Toolbox) is a Matlab toolbox that automatically builds accurate [http://en.wikipedia.org/wiki/Surrogate_model surrogate models] (also known as metamodels or [http://en.wikipedia.org/wiki/Response_surface_methodology response surface models]) of a given data source (simulation code, data set, script, ...) within the accuracy and time constraints set by the user. In doing so the toolbox minimizes the number of data points (which it chooses automatically) since they are usually expensive. The toolbox tries to be as adaptive and autonomous as possible, requiring no user input besides some initial configuration. 

+  
+  [[Image:sumotask.pngcenterSUMOToolbox : Generating an approximation for a reference model]] 

However, since there is no such thing as a ''onesizefitsall'', the toolbox has been designed to be fully pluggable and extensible using standard [http://en.wikipedia.org/wiki/Object_orientation object oriented] design patterns. Implementations of the different components (model types, sampling strategies, model selection criteria, hyperparameter optimization algorithms,...) can be pluggedin, compared, or replaced by custom implementations. In this way the SUMO Toolbox provides a common platform to easily test and benmark different sampling and approximation strategies while easily integrating in the engineering design process. [[Aboutmore information....]] 
However, since there is no such thing as a ''onesizefitsall'', the toolbox has been designed to be fully pluggable and extensible using standard [http://en.wikipedia.org/wiki/Object_orientation object oriented] design patterns. Implementations of the different components (model types, sampling strategies, model selection criteria, hyperparameter optimization algorithms,...) can be pluggedin, compared, or replaced by custom implementations. In this way the SUMO Toolbox provides a common platform to easily test and benmark different sampling and approximation strategies while easily integrating in the engineering design process. [[Aboutmore information....]] 
Revision as of 11:24, 26 June 2009
SUrrogate MOdeling (SUMO) Toolbox 
The Surrogate Modeling Toolbox (SUMO Toolbox) is a Matlab toolbox that automatically builds accurate surrogate models (also known as metamodels or response surface models) of a given data source (simulation code, data set, script, ...) within the accuracy and time constraints set by the user. In doing so the toolbox minimizes the number of data points (which it chooses automatically) since they are usually expensive. The toolbox tries to be as adaptive and autonomous as possible, requiring no user input besides some initial configuration. However, since there is no such thing as a onesizefitsall, the toolbox has been designed to be fully pluggable and extensible using standard object oriented design patterns. Implementations of the different components (model types, sampling strategies, model selection criteria, hyperparameter optimization algorithms,...) can be pluggedin, compared, or replaced by custom implementations. In this way the SUMO Toolbox provides a common platform to easily test and benmark different sampling and approximation strategies while easily integrating in the engineering design process. more information.... 
Open Positions



We are constantly looking for highly motivated PhD candidates or postdocs, that have a strong interest in one of our research topics. Our research area is highly multidisciplinary, and requires strong mathematical, physical and computer science knowledge. Please contact tom.dhaene@ugent.be 
 This website is under ongoing construction. Since it is a Wiki, you can help by contributing