Capabilities
RAVEN
RAVEN provides a set of basic and advanced capabilities that ranges from data generation, data processing and data visualization.
A full set of RAVEN computational capabilities are listed below.
Computing Capabilities
- Parallel computation capabilities (multi-thread and multi-core)
- Supported operating systems: MAC, Linux and Windows
- Workstation and high performance computing (HPC) systems
Multi-steps Analyses
RAVEN analyses are performed through a series of simulation steps. Each simulation step allows the user to perform a set of basic actions:
- Multi-Run
- Training of a ROM
- Post-Process
- IOStep
More complex analyses are performed by simply assembling and linking a series of steps listed above.
Creation & Use of Reduced Order Models (scikit-learn, TensorFlow and CROW libraries)
- SVM
- Gaussian process models
- Linear models
- Multi-class classifiers
- Decision trees
- Naive Bayes
- Neighbors classifiers and regressors
- Multi-dimensional interpolators
- High dimension model reduction (HDMR)
- Morse-Smale complex
- Dynamic Mode Decomposition
- Neural Networks (Deep Learning)
Forward Propagation of Uncertainties
- MonteCarlo sampling
- Grid sampling
- Stratified sampling
- Factorial design
- Response surface design
- Multi-variate analysis
- Generic Polynomial Chaos expansion
Advance Sampling Methods
- Sobol index sampling
- Adaptive sampling
- Sparse grid collocation sampling
- Dynamic event trees (DETs)
Model Capabilities
- Generic interface with codes
- Custom code interfaces
- Custom ad-hoc external models
Data Post-Processing Capabilities
- Data clustering
- Data regression
- Data dimensionality reduction
- Custom generic post-processors
- Time-dependent data analysis
- Data plotting
Model Parameter Optimization
- Simultaneous perturbation stochastic approximation method
- Finite difference gradient based optimization
- Conjiugate gradient based optimization
Data Management
- Data importing and exporting
- Databases creation
