<  Back to the Polytechnique Montréal portal

VirtualGAN: Reducing mode collapse in generative adversarial networks using virtual mapping

Adel Abusitta, Omar Abdul Wahab and Benjamin C. M. Fung

Paper (2021)

Document published while its authors were not affiliated with Polytechnique Montréal

An external link is available for this item
Show abstract
Hide abstract

Abstract

This paper introduces a new framework for reducing mode collapse in Generative adversarial networks (GANs). The problem occurs when the generator learns to map several various input values (z) to the same output value, which makes the generator fail to capture all modes of the true data distribution. As a result, the diversity of synthetically produced data is lower than that of the real data. To address this problem, we propose a new and simple framework for training GANs based on the concept of virtual mapping. Our framework integrates two processes into GANs: merge and split. The merge process merges multiple data points (samples) into one before training the discriminator. In this way, the generator would be trained to capture the merged-data distribution rather than the (unmerged) data distribution. After the training, the split process is applied to the generator's output in order to split its contents and produce diverse modes. The proposed framework increases the chance of capturing diverse modes through enabling an indirect or virtual mapping between an input z value and multiple data points. This, in turn, enhances the chance of generating more diverse modes. Our results show the effectiveness of our framework compared to the existing approaches in terms of reducing the mode collapse problem.

Uncontrolled Keywords

training; neural networks; generative adversarial networks; generators

PolyPublie URL: https://publications.polymtl.ca/51753/
Conference Title: International Joint Conference on Neural Networks (IJCNN 2021)
Conference Location: Shenzhen, China
Conference Date(s): 2021-07-18 - 2021-07-22
Publisher: IEEE
DOI: 10.1109/ijcnn52387.2021.9533656
Official URL: https://doi.org/10.1109/ijcnn52387.2021.9533656
Date Deposited: 18 Apr 2023 14:59
Last Modified: 05 Apr 2024 11:56
Cite in APA 7: Abusitta, A., Abdul Wahab, O., & Fung, B. C. M. (2021, July). VirtualGAN: Reducing mode collapse in generative adversarial networks using virtual mapping [Paper]. International Joint Conference on Neural Networks (IJCNN 2021), Shenzhen, China (6 pages). https://doi.org/10.1109/ijcnn52387.2021.9533656

Statistics

Dimensions

Repository Staff Only

View Item View Item