humpday
Humpday documentation (install)
A package to help you in two main ways:
- Choose a Python global optimizer package without learning dozens of quirky usage conventions
- Optimize on a simplex (see how) regardless of the optimizer chosen
Basic usage
Import an optimizer or three, and run them:
from humpday.objectives.classic import CLASSIC_OBJECTIVES
from humpday.optimizers.dlibcube import DLIB_OPTIMIZERS
for objective in CLASSIC_OBJECTIVES:
print(' ')
print(objective.__name__)
for optimizer in DLIB_OPTIMIZERS:
f_best, x_best, actual_num_trials = optimizer(objective, n_trials=500, n_dim=34, with_count=True)
print({optimizer.__name__:f_best})
Find what you want in humpday/optimizers, then in every definition file (e.g. hyperoptcube.py) you’ll find an example of how to run the optimizers. There are more patterns in basic usage examples, maybe.
Packages leveraged
Ax-Platform, bayesian-optimization, DLib, HyperOpt, NeverGrad, Optuna, Platypus, PyMoo, PySOT, Scipy classic and shgo, Skopt, nlopt, Py-Bobyaq, UltraOpt, FreeLunch and maybe others in a common calling syntax.
Elo ratings
See article.
Articles:
- A New Family of Diffeomorphisms from the Simplex to the Cube
- The Python Optimizer Form Guide
- Comparing Python Global Optimizer Packages
- Interpreting the Python Global Optimizer Elo Ratings
