Skip to content
/ movado Public
forked from dpaletti/movado

Black-box fitness function approximation python library

License

Notifications You must be signed in to change notification settings

necst/movado

 
 

Repository files navigation

Table of Contents

  1. Installation
  2. Usage
  3. Parameters

Installation

Movado needs python 3.6 or higher. Install it through pip (on many linux systems you need to use pip3 to force python3 installation):

pip3 install --user --no-cache movado

Usage

Movado exposes just an annotation which you need to apply to the fitness function you want to approximate:

@approximate(outputs=3)
def fitness(point: List[Number]) -> List[Number]:
    # my fitness logic

class SomeClass:
    def __init__():
        ...
    @approximate(outputs=1)
    def fitness(self, point: List[Number]) -> List[Number]:
        # this is also allowed

The only mandatory parameter to approximate is the number of objectives of your optimization problem, in the above case the fitness function returns a list of 3 floats and takes as input a list of numbers. Fitness Function Structure:

  1. the fitness function must take only a list of numbers as input. If we approximate a method in a class self is allowed as first parameter but it must be strictly followed by the list of numbers
  2. the fitness function must output a list of numbers with constant length between fitness evaluation

Parameters

Other than outputs several other optional parameters are available

Parameter Description Default Value
disabled if True the approximation False
  is disabled  
stochastic if True caching of fitness False
  calls is disabled  
estimator estimator to use HoeffdingAdaptiveTree
controller controller to user Mab

Associated Publication

@inproceedings{paletti2021online,
  title={Online Learning RTL Synthesis for Automated Design Space Exploration},
  author={Paletti, Daniele and Peverelli, Francesco and Conficconi, Davide and Santambrogio, Marco D},
  booktitle={2022 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW)},
  year={2022},
  organization={IEEE}
}

About

Black-box fitness function approximation python library

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%