launching the Lipizzaner framework to train Generative Adversarial Networks
by using spatially distributed co-evolutionary algorithms. Most of the
information is on our website at http://lipizzaner.csail.mit.edu/
We have designed and developed Lipizzaner framework to provide resilient
and robust GAN training. Lipizzaner is open-source and it is freely
available in our GitHub repository (
https://github.com/ALFA-group/lipizzaner-gan).
An overview of Lipizzaner can be found in the following Medium post:
https://medium.com/@jamaltoutouh/lipizzaner-a-framework-for-co-evolutionary-distributed-gan-training-7a725fb17e49
The main features of Lipizzaner are:
- Fast convergence because the weights are updated by gradient-based steps
- Improved convergence due to training hyperparameters evolution
- Diverse sample generation because the method returns a mixture of
generators trained by using EC
- Scalability due to spatial distribution topology and asynchronous
parallelism
- Robustness and resilience
It includes a number of datasets ready to be used: CIFAR-10, MNIST,
MNIST-Fashion, CelebA, SVHN, Chest X-Ray COVID-19 images, etc.
In the following days, we will post a series of tutorials and new
documentation.
Get updated. Follow us on twitter: @LipizzanerGAN
PS: We are tweaking Lipizzaner 2.0.... so, we will launch really soon
PPS: See a list of most relevant papers related to Lipizzaner:
J. Toutouh, E. Hemberg, U. O'Reilly. Analyzing the Components of
Distributed Coevolutionary GAN Training. In The Sixteenth International
Conference on Parallel Problem Solving from Nature (PPSN XVI). pages. 10,
2020. arxiv.org/abs/2008.01124
J. Toutouh, E. Hemberg, U. O'Reilly. Re-purposing Heterogeneous Generative
Ensembles with Evolutionary Computation. In Genetic and Evolutionary
Computation Conference (GECCO '20), pages. 10, 2020. DOI:
10.1145/3377930.3390229
J. Toutouh, E. Hemberg, U. O'Reilly. Data Dieting in GAN Training. In: Iba
H., Noman N. (eds) Deep Neural Evolution. Natural Computing Series.
Springer, Singapore. DOI: 10.1007/978-981-15-3685-4_14
Jamal Toutouh, Erik Hemberg, and Una-May O'Reilly. Spatial Evolutionary
Generative Adversarial Networks. In Genetic and Evolutionary Computation
Conference (GECCO '19), July 13–17, 2019, Prague, Czech Republic. ACM, New
York, NY, USA, 9 pages. https://doi.org/10.1145/3321707.3321860
A. Al-Dujaili, T. Schmiedlechner, E. Hemberg, U. O'Reilly. Towards
distributed coevolutionary GANs. In AAAI 2018 Fall Symposium, 2018.
T. Schmiedlechner, I. Ng Zhi Yong, A. Al-Dujaili, E. Hemberg, U. O'Reilly.
Lipizzaner: A System That Scales Robust Generative Adversarial Network
Training. In NeurIPS 2018 Workshop on System for Machine Learning, 2018.
----------------------------------------------
Jamal Toutouh, PhD
- Postdoctoral Fellow at MIT CSAIL
+ MIT Postdoctoral Association Board Member
+ ECUSA Association Board Member
email: toutouh@mit.edu
website: jamal.es
twitter: @jamtou
**********************************************************
*
* Contributions to be spread via DMANET are submitted to
*
* DMANET@zpr.uni-koeln.de
*
* Replies to a message carried on DMANET should NOT be
* addressed to DMANET but to the original sender. The
* original sender, however, is invited to prepare an
* update of the replies received and to communicate it
* via DMANET.
*
* DISCRETE MATHEMATICS AND ALGORITHMS NETWORK (DMANET)
* http://www.zaik.uni-koeln.de/AFS/publications/dmanet/
*
**********************************************************
Posts
