| 1 |
Files in this directory demonstrate use cases for some of mystic's "advanced" |
| 2 |
features, such as using dimensional collapse detectors, building a surrogate |
| 3 |
for an unknown surface, or using support vectors in optimization. |
| 4 |
|
| 5 |
|
| 6 |
== Notes on mystic examples3 == |
| 7 |
Naming convention: |
| 8 |
- `3D` denotes use of 3-D toy model in surrogate.py |
| 9 |
- `5D` denotes use of 5-D toy model in toys.py |
| 10 |
- `GP` denotes use of sklearn GP regression instead of RBF interpolation |
| 11 |
- `NN` denotes use of sklearn MLP regression instead of RBF interpolation |
| 12 |
- 'ML' denotes use of iterative surrogate-assisted optimization |
| 13 |
- `c0mean` denotes a (backpropagation) calculation of mean value of an input |
| 14 |
- `collapse` denotes use of dimensional collapse detectors |
| 15 |
- `mean` denotes a calculation of mean value of an output |
| 16 |
- `ouq` denotes use of high-level UQ interface to find bounds, surrogates |
| 17 |
- `pandas` denotes mystic extensions/interface to pandas |
| 18 |
- `pof` denotes a probability of failure calculation |
| 19 |
- `searcher` denotes optimizer-driven discovery of all extrema [DEPRECATED] |
| 20 |
- `sklearn` denotes mystic extensions/interface to sklearn |
| 21 |
- `surface` denotes optimizer-driven surface interpolation [DEPRECATED] |
| 22 |
- `svc` denotes use of support vectors in classification |
| 23 |
- `svr` denotes use of support vectors in regression |
| 24 |
- `sparsity` denotes use of the sparsity sampler |
| 25 |
- `xrd` denotes the lattice parameter toy model |
| 26 |
|
| 27 |
OUQ calculations: |
| 28 |
- test_*_ub_*: find upper bound on (mean of input/output, or PoF) [examples5] |
| 29 |
- test_error*: find model error |
| 30 |
- test_expect: find expected value |
| 31 |
- test_expected*: find bound on expected value |
| 32 |
- test_expected_error* find bound on expected error |
| 33 |
- test_glb*: find greatest lower bound |
| 34 |
- test_gub*: find greatest upper bound |
| 35 |
- test_llb*: find least lower bound |
| 36 |
- test_lub*: find least upper bound |
| 37 |
- test_pof: find probability of failure (PoF) |
| 38 |
- xrd_design*: find average and bounds on expected value and bounds |
| 39 |
|
| 40 |
Miscellaneous: |
| 41 |
- test_cache: demonstrate use and customization of mystic.cache |
| 42 |
- test_improve_score: workflow to retry/update sklearn estimators |
| 43 |
- xrd_opt*: find minimum using ensemble optimization |
| 44 |
- xrd_optML*: find minimum using iterative surrogate improvement |
| 45 |
- xrd_*_db: inspect expected value databases generated with xrd_design* |
| 46 |
|
| 47 |
External dependencies: |
| 48 |
- examples with "svc", "svr", or "sparsity" in the name require `matplotlib`. |
| 49 |
- examples with "surface" and "searcher" require `scipy` and `matplotlib`. |
| 50 |
- examples with "pandas" require `pandas` and `matplotlib`. |
| 51 |
- otherwise, assume `sklearn` and `matplotlib` are required. |
| 52 |
|
| 53 |
In-place dependencies for ouq* calculations: |
| 54 |
- dataset: extensions on dataset operations |
| 55 |
- emulators: toy models emulating XRD calculations |
| 56 |
- estimator: high-level OUQ estimator (requires sklearn) |
| 57 |
- interpolator: high-level OUQ interpolator |
| 58 |
- misc: miscellaneous constraints and definitions for OUQ calculation |
| 59 |
- ml: lower-level interface to sklearn, used by estimator |
| 60 |
- noisy: add noise for OUQ noisy model |
| 61 |
- plotter: high-level interface to plotting learned surrogates |
| 62 |
- spec*: same as misc, used in test_*_ub_* OUQ calculations |
| 63 |
- surrogate: hypervelocity impact model, used in test_*_ub_* OUQ calculations |
| 64 |
- toys: simple toy models for OUQ calculations |