Using High-Performance Computing to Scale Generative Adversarial Networks
Generative adversarial networks(GANs) are methods that can be used for data augmentation, which helps in creating better detection models for rare or imbalanced datasets. They can be difficult to train due to issues such as mode collapse. We aim to improve the performance and accuracy of the Lipizza...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Published: |
Massachusetts Institute of Technology
2022
|
| Online Access: | https://hdl.handle.net/1721.1/139311 |
| _version_ | 1826191151297724416 |
|---|---|
| author | Flores, Diana J. |
| author2 | Hemberg, Erik |
| author_facet | Hemberg, Erik Flores, Diana J. |
| author_sort | Flores, Diana J. |
| collection | MIT |
| description | Generative adversarial networks(GANs) are methods that can be used for data augmentation, which helps in creating better detection models for rare or imbalanced datasets. They can be difficult to train due to issues such as mode collapse. We aim to improve the performance and accuracy of the Lipizzaner GAN framework by taking advantage of its distributed nature and running it at very large scales. Lipizzaner was implemented for robustness, but has not been tested at scale in high performance computing(HPC) systems. We believe that by utilizing HPC technologies, we can scale up Lipizzaner and observe performance enhancements. This thesis achieves this scale up, using Oak Ridge National Labs’ Summit Supercomputer. We observed improvements in the performance of Lipizzaner, especially when run with poorer network architectures, which implies Lipizzaner is able to overcome network limitations through scale. |
| first_indexed | 2024-09-23T08:51:03Z |
| format | Thesis |
| id | mit-1721.1/139311 |
| institution | Massachusetts Institute of Technology |
| last_indexed | 2024-09-23T08:51:03Z |
| publishDate | 2022 |
| publisher | Massachusetts Institute of Technology |
| record_format | dspace |
| spelling | mit-1721.1/1393112022-01-15T04:04:16Z Using High-Performance Computing to Scale Generative Adversarial Networks Flores, Diana J. Hemberg, Erik Toutouh, Jamal O’Reilly, Una-May Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science Generative adversarial networks(GANs) are methods that can be used for data augmentation, which helps in creating better detection models for rare or imbalanced datasets. They can be difficult to train due to issues such as mode collapse. We aim to improve the performance and accuracy of the Lipizzaner GAN framework by taking advantage of its distributed nature and running it at very large scales. Lipizzaner was implemented for robustness, but has not been tested at scale in high performance computing(HPC) systems. We believe that by utilizing HPC technologies, we can scale up Lipizzaner and observe performance enhancements. This thesis achieves this scale up, using Oak Ridge National Labs’ Summit Supercomputer. We observed improvements in the performance of Lipizzaner, especially when run with poorer network architectures, which implies Lipizzaner is able to overcome network limitations through scale. M.Eng. 2022-01-14T15:03:08Z 2022-01-14T15:03:08Z 2021-06 2021-06-17T20:13:13.241Z Thesis https://hdl.handle.net/1721.1/139311 In Copyright - Educational Use Permitted Copyright MIT http://rightsstatements.org/page/InC-EDU/1.0/ application/pdf Massachusetts Institute of Technology |
| spellingShingle | Flores, Diana J. Using High-Performance Computing to Scale Generative Adversarial Networks |
| title | Using High-Performance Computing to Scale Generative Adversarial Networks |
| title_full | Using High-Performance Computing to Scale Generative Adversarial Networks |
| title_fullStr | Using High-Performance Computing to Scale Generative Adversarial Networks |
| title_full_unstemmed | Using High-Performance Computing to Scale Generative Adversarial Networks |
| title_short | Using High-Performance Computing to Scale Generative Adversarial Networks |
| title_sort | using high performance computing to scale generative adversarial networks |
| url | https://hdl.handle.net/1721.1/139311 |
| work_keys_str_mv | AT floresdianaj usinghighperformancecomputingtoscalegenerativeadversarialnetworks |