UQPyL: The Uncertainty Quantification Python Laboratory provides comprehensive workflows tailored for Uncertainty Quantification and Optimization of computational models and their associated applications (e.g., model calibration, resource scheduling, product design).
- Comprehensive Sensitivity Analysis and Optimization Algorithm: Implements widely used sensitivity analysis methodologies and optimization algorithms.
- Advanced Surrogate Modeling: Integrates diverse surrogate models equipped to solve computationally expensive problems.
- Rich Application Resources: Provides a comprehensive suite of benchmark problems and practical case studies, enabling users to get started quickly.
- Modular and Extensible Architecture: Encourages and facilitates the development of novel methods or algorithms by users, aligning with our commitment to openness and collaboration. (We appreciate and welcome contributions)
- Website: UQPyL Official Site (#TODO: Needs update)
- Source Code: GitHub Repository
- Documentation: ReadTheDocs(#TODO: Being updating )
- Citation Info: [UQPyL 2.0](#TODO: Needs update), UQPyL 1.0
(All methods support surrogate models)
- Sobol'
- Delta Test (DT)
- Extended Fourier Amplitude Sensitivity Test (eFAST)
- Random Balance Designs - Fourier Amplitude Sensitivity Test (RBD-FAST)
- Multivariate Adaptive Regression Splines-Sensitivity Analysis (MARS-SA)
- Morris
- Regional Sensitivity Analysis (RSA)
(* indicates solving computational expensive optimization problem)
- Single Objective Optimization: SCE-UA, ML-SCE-UA, GA, CSA, PSO, DE, ABC, ASMO*, EGO*
- Multi-Objective Optimization: MOEA/D, NSGA-II, RVEA, NSGA-III, MOASMO*
Note: The library is still being updated. If you need other algorithms, please contact us.
- Fully Connected Neural Network (FCNN)
- Kriging (KRG)
- Gaussian Process (GP)
- Linear Regression (LR)
- Polynomial Regression (PR)
- Radial Basis Function (RBF)
- Support Vector Machine (SVM)
- Multivariate Adaptive Regression Splines (MARS)
Recommended (PyPi or Conda):
pip install UQPyLconda install UQPyLAlternatively:
git clone https://github.com/smasky/UQPyL.git
cd UQPyL
pip install .To use UQPyL, define the problem you want to solve. The problem usually contains three important properties:
func(the mapping from X to Y)- The dimensions of decisions and outputs
- The bounds of decisions (ub, lb)
from UQPyL.problems.single_objective import Sphere
problem = Sphere(nInput=10, ub=100, lb=-100)
problem = Sphere(nInput=10, ub=np.ones(10)*100, lb=np.ones(10)*-100)Define the evaluation function:
from UQPyL.problems import PracticalProblem
def func(X):
Y = np.sum(X, axis=1).reshape(-1, 1)
return Y
problem = PracticalProblem(func=func, nInput=10, nOutput=1, ub=100, lb=-100, name="Sphere")Note: The func needs to accept a matrix of X and return a matrix of Y, with columns equal to dimensions and rows equal to samples. X and Y should be np.ndarray.
After defining the problem, you can use any methods in UQPyL.
from UQPyL.sensibility import Sobol
sobol = Sobol() # Instantiate and set hyper-parameters
sobol.analyze(problem)from UQPyL.optimization.single_objective import SCE_UA
sce = SCE_UA()
res = sce.run(problem)
bestDec = res.bestDec
bestObj = res.bestObjfrom UQPyL.DoE import LHS
lhs = LHS(problem)
xTrain = lhs.sample(200, problem.nInput)
yTrain = problem.evaluate(xTrain)
xTest = lhs.sample(50, problem.nInput)
yTest = problem.evaluate(xTest)
from UQPyL.surrogate.rbf import RBF
rbf = RBF()
rbf.fit(xTrain, yTrain)
yPred = rbf.predict(xTest)
from UQPyL.utility.metric import r_square
r2 = r_square(yTest, yPred)For more advanced usage, please refer to the documentation (#TODO).
We welcome contributions to expand our library with more sophisticated UQ methods, optimization algorithms and engineering problems.
For any inquiries or contributions, please contact:
wmtSky
Email: wmtsky@hhu.edu.cn, wmtsmasky@gmail.com
This project is licensed under the MIT License - see the LICENSE file for details.
