Skip to content

A PyTorch implementation of an Adaptive Memory Multi-Batch L-BFGS

License

Notifications You must be signed in to change notification settings

fedezocco/AdaMemLBFGS-PyTorch

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

21 Commits
 
 
 
 
 
 

Repository files navigation

An Adaptive Memory Multi-Batch L-BFGS in PyTorch

The repository contains the source code of the paper "An adaptive memory multi-batch L-BFGS algorithm for neural network training" authored by Federico Zocco and Seán McLoone. The paper has been presented at the 21st IFAC World Congress, held in Berlin in July 12-17, 2020.

If you use the code, please consider citing the corresponding paper: Federico Zocco, Seán McLoone, An Adaptive Memory Multi-Batch L-BFGS Algorithm for Neural Network Training, IFAC-PapersOnLine, Volume 53, Issue 2, 2020, Pages 8199-8204, ISSN 2405-8963, https://doi.org/10.1016/j.ifacol.2020.12.1996.

Note that the datasets are not uploaded in this folder because too large. See Table 1 of my paper if you need information to find the datasets.

Dependencies of folders and files

The script "Algorithms_comparison.py" should be executed by the user to do the Monte Carlo simulations that compare the training algorithms on different datasets. It uses all the modules inside the same folder of "Algorithms_comparison.py". It also uses the modules inside the folder "functions".

The datasets for Experiments 1, 4 and 5 are downloaded by "Algorithms_comparison.py" from the Internet.

The folder "Others" contains 2 scripts demonstrating how to do image rescaling and data augmentation. They are not used by the script "Algorithms_comparison.py". I have created and used this 2 scripts to rescale and augment the datasets as detailed in my paper.

About

A PyTorch implementation of an Adaptive Memory Multi-Batch L-BFGS

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages