Skip to content

opendihu/optimization

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

397 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Bayesian Optimization

Description

Bayesian Optimization is a method to find maxima of blackbox functions with a relatively low number of function evaluations. For that we are using the python package botorch. This repository contains files for Bayesian Optimization that can be used standalone or call the neuromuscular simulation framework OpenDiHu to find the optimal parameter(s) of a skeletal muscle simulation.

This project started as Lukas Bauer's bachelor thesis, that led to the first version of the repository. The thesis can be accessed at OPUS, the online publication library from the University Stuttgart.

Dependencies

Python: Python 3.10.12

Required Python libraries: botorch, torch, numpy, matplotlib, subprocess, sys, os, shlex, csv, time, signal

OpenDiHu (if used): Version 1.5

Setup

Inside BayesianOptimization is the setup for a 1D Bayesian Optimization that is set up to optimize an easy dummy function. In the subfolder "nD" is a similar setup for a function from $\mathbb{R}^n$ to $\mathbb{R}$.

Inside opendihu_examples you can find the examples "isotonic_contraction", "isometric_contraction", "paired_muscles" and "prestretch_force_for_given_length". An overview of the different cases described is provided in the ReadMe file.

Inside test_functions there are several made-up functions which can be used to test different Bayesian Optimization models. These functions have different characteristics, so that you can choose a model that works best for the kind of functions you are looking for.

Results

The results of the mentioned Bachelor Thesis are in the test_functions and isotonic_cuboid_contraction files. A discussion of when to use which parameter setup can be found here. Other results can be found in the corresponding ReadMes of the different cases.

About

A Bayesian optimization framework

Resources

Stars

Watchers

Forks

Packages

No packages published

Contributors 2

  •  
  •