Here is a one page help to help you to describe your numerical experiments (NE). Always start from simple example and complexify. Always compare with SoA methodologies
find an opensource algorithm ? https://the-algorithms.com
Exemple: Joseph Morlier, tuesday the 12th of September 2019
In 3 lines, try to resume the big picture of your NE.
Let's define first the physics embedded in your problem and level of fidelity of your NE: Exemple: Fluid and Structures add basic physical hypothesis, linearity / NL, type of solver/tolerances Control is a physics?
Describe the methodology with full references Example: in Fluid Dynamics / CFD
LO Empiricial equation /regression
L0+ lifting line
L1 DLM/VLM
L2 RANS
L2+ LES etc...
check convergency
Please, Recap also the tuning of the physical solver
Describe the methodology with full references
space/time/both
Describe the methodology with full references
If you are using openMDAO, please use the last version
Describe the methodology with full references (RBF, GP, ...)
Please describe how you constructed and validated your surrogate Precise the Kernel, and hyperparameter tuning
Architecture Number of inputs x Number of ouptus Number of HL Type of activation function
Describe the methodology with full references (SVD, ...)
x can be continuous, discrete, categorial ...hybrid
Describe the methodology (Gradient Based, GA, Bayesian Optimization/EGO ...)
Constrained or unconstrained ? Use derivative or not ? Sensibility computed using ? FD, Complex step, Analytic, Semi Analytic
Clearly define the objective function (mono, multi) and the constraints (equality, inequality)
min f(x)
wrt g(x)<=0
h(x)==0
and bounds for x
Can you write a linear system for constraints?
define the optimizer (SLSQP for example) and associated tolerances and or stopping criteria
Provide X0, Xn, X* (sometimes it's nice to see the evolution at n, please choose n)
Multistart ?
Upload here your codes and precise here the dependencies