September 16, 2014 - September 16, 2014
Webinar: Add value to your experimental data and perform faster system behavior prediction with Surrogate Modeling in pSeven
This webinar presents pSeven’s surrogate modeling and data analysis capabilities, which allow to:
- Replace time-consuming and computationally expensive models with cheap to evaluate surrogate models.
- Predict your system behavior quickly and easily with surrogate models built using your experimental data.
- Find optimal characteristics of your design.
Webinar showcases the following tools available in pSeven to perform surrogate modeling, coupled with the real-life application examples:
Approximation tool to construct approximation (surrogate) models using data of the same fidelity and assesses their quality. Tool properties:
- In-house developed and popular approximation methods allow to accelerate surrogate modeling and/or make the models more accurate.
- Decision tree intelligent technique selection mechanism (automatic selection) for users
- Manual setup option for experts
- Accuracy evaluation capabilities and possibility to smoothen the approximation before evaluating.
Data fusion tool to construct surrogate models using data of different fidelity and generate a surrogate model for the high fidelity function as a result. Tool properties:
- Can exactly fit the high fidelity training data;
- Allows accuracy evaluation to estimate uncertainty for predictions obtained;
- Estimates quality of obtained models using Internal validation.
Both tools provide both approximation of high fidelity functions and partial derivatives, allow to manage a surrogate model construction time and perfectly handle samples of varied sizes: from tiny to really huge datasets.
Other pSeven capabilities for surrogate modeling and data analysis are mentioned, such as Design of Experiments, Sensitivity and Dependency Analysis, Dimension reduction and Feature Extraction. These capabilities will be subjects of further DATADVANCE webinars.