This repository contains an implementation of the following algorithms:
LIPO+ and AdaLIPO+ are empirical enhancements introduced of LIPO and AdaLIPO, introduced in the paper Experimental Improvements of Global Optimization Algorithms for Lipschitz Functions. A demo of these algorithms is available on the IPOL website.
You need to create a class for your function to maximize. This class must be named Function
and follow the following interface:
import numpy as np
class Function:
def __init__(self) -> None:
self.bounds = bounds # (min, max) tuple for each dimension (numpy array)
self.k = k # Lipschitz constant (float)
pass
def __call__(self, x: np.ndarray) -> float:
# Closed form of the function to maximize
pass
The file containing the class must be in the functions
folder.
Several examples are available in the functions
folder.
Then, you can run the algorithm with the following command:
python src/main.py --function <path_to_your_function_class> -n <number_of_function_eval>
Some optional arguments are available, you can see them with the --help
flag.