/2019-Global-Optimization-UNIFI

Comparison of two algorithms for hyperparameters global optimization: Radial Basis Functions and Bayesian gaussian processes.

Primary LanguagePython

2019 Global Optimization @ UniFi

logo-UniFi

Introduction

This repo contains codes and documentations for the "Global Optimization" course at University of Florence.

The projects has been conducted together with @fedem96.

The report file (in Italian) can be found here.

Hyperparameters Optimization

The aim of this project is to compare two algorithms for hyperparameters global optimization, one based on Radial Basis Functions and the other based on Bayesian gaussian processes.

System requirements

  • download Bonmin, Ipopt and other binaries: from https://ampl.com/dl/open/ download the binaries for your operating system
  • extract the binaries, and add extracted folder path to your $PATH environment variable
  • install pytorch
  • install these Python 3 packages too: rbfopt, bayesian-optimization, tensorboardX

Usage

  • download the project: git clone https://github.com/emanuelevivoli/Hyperparameters_Optimization.git
  • enter in the project directory: cd Hyperparameters_Optimization
  • (modify and) run evaluation.py: python3 evaluate.py

Contacts

If you are interested and have some questions, don't hesitate to contact us or open an issue.