humpday 0.1.18

Taking the pain out of choosing a Python global optimizer

humpday

A package that helps you choose a Python global optimizer package, and strategy therin, from Ax-Platform, bayesian-optimization, DLib, HyperOpt, NeverGrad, Optuna, Platypus, PyMoo, PySOT, Scipy classic and shgo, Skopt, and UltraOpt.

• 50+ strategies are assigned Elo ratings.

• All are presented in a common calling syntax. By all means contribute more to optimizers.

• Pass the dimensions of the problem, function evaluation budget and time budget to receive suggestions,

• Or call the meta minimizer, which will choose one and run it.

    from humpday import minimize
    best_val, best_x = minimize(f, n_dim=13, n_trials=130 )
  

Here f is intended to be minimized on the hypercube [0,1]^n_dim.

Optimizer suggestions

    from pprint import pprint 
    from humpday import suggest
    pprint(suggest(n_dim=5, n_trials=130,n_seconds=5*60))

where n_seconds is the total computation budget for the optimizer (not the objective function) over all 130 function evaluations. Alternatively you can pass an objective function:

    from humpday import recommend

    def my_objective(u):
        time.sleep(0.01)
        return u[0]*math.sin(u[1])

    recommendations = recommend(my_objective, n_dim=21, n_trials=130)

As this function is very fast, some optimizers will be culled from the list.

Install

pip install humpday

Bleeding edge:

pip install git+https://github.com/microprediction/humpday

File an issue if you have problems. If you get a CMake error, try:

pip install cmake
pip install humpday 

Optional packages

I removed ultraopt from setup so you'll have to install it manually:

pip install cmake
pip install ultraopt

Articles

• (most recent) HumpDay: A Package to Take the Pain Out of Choosing a Python Optimizer.
• Comparing Python Global Optimizers.