Skip to content

Original code for the paper "Multi-Objective Evolutionary Optimization of Virtualized Fast Feedforward Networks", accepted at EvoStar 2025

Notifications You must be signed in to change notification settings

DIOL-UniTN/MOE-VFFF

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

11 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Multi-Objective Evolutionary Optimization of Virtualized Fast Feedforward Networks

This is the repo associated to the paper Multi-Objective Evolutionary Optimization of Virtualized Fast Feedforward Networks, to be published at EvoStar 2025.

To reproduce the experiments, you can run:

Virtualization via NSGA-II:

python src/main.py --multirun loader=mnist,har,sc exp=movirt 

Virtualizatin via random search:

python src/main.py --multirun loader=mnist,har,sc exp=randomvirt 

Landscape/Virtualization via manual tuning:

python src/main.py --multirun loader=mnist,har,sc exp=weightvirt exp.page_size=0,1,2,3,4 exp.amount=0.5,0.6,0.7,0.8,0.9 

Note that these scripts run multiple experiments. If you want to run a single experiment, you can run:

python src/main.py loader=loader-name exp=experiment-name 
  • Options for loader-name: mnist, har, sc
  • Options for expriment-name: train, movirt, randomvirt, weightvirt, generate_fim

Please see YAML files inside conf directory to see more details regarding configurations. Code, data (including pretrained models and their corresponding fisher information matrices), and results are available at: https://dagshub.com/berab/MOE-VFFF.

About

Original code for the paper "Multi-Objective Evolutionary Optimization of Virtualized Fast Feedforward Networks", accepted at EvoStar 2025

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages