Abstract:Original data usually lies on a set of disconnected manifolds rather than a smooth connected manifold. This causes the problem of mode collapse in the training of vanilla Generative Adversarial Network (GAN). There are many existing GAN variants that attempt to address this problem, but they result in limitations. The existing variants either produce simulated instances with low quality or generate identical simulated instances. In this study, we propose a new approach to training GAN utilizing multiple generators, a classifier and a discriminator to address mode collapse. The classifier outputs the statistical probabilities of generated data belonging to a specific category. These probabilities implicitly reflect which manifolds are captured by generators, and the correlation between generators is quantified by mutual information. Our idea views the mutual information values as a constraint to guide generators in learning different manifolds. Specifically, we traverse the generators, calculating the mutual information between each generator and the others. The calculated values are integrated into the generator loss to form a new generator loss and to update the corresponding generator's parameters, using back-propagation. We minimize the mutual information to reduce the correlation between generators while also minimizing the generator loss. This ensures generators capture different manifolds while updating their parameters. A new minimax formula is established to train our approach in a similar spirit to vanilla GAN. We term our approach Mutual Information Multi-generator GAN (MIM-GAN). We conduct extensive experiments utilizing the MNIST, CIFAR10 and CelebA datasets to demonstrate the significant performance improvement of MIM-GAN in both achieving the highest Inception Scores and producing diverse generated data at different resolutions.

Novelty Detection for Multimode Process Using GANs with Learning Disentangled Representation

Dual Distribution Matching GAN

Generative Adversarial Network Based Novelty Detection Usingminimized Reconstruction Error

Non-Adversarial Novelty Detection with Generative Latent Nearest Neighbors

Novelty Detection via Non-Adversarial Generative Network

TransductGAN: a Transductive Adversarial Model for Novelty Detection

MDGAN: Boosting Anomaly Detection Using \\Multi-Discriminator Generative Adversarial Networks

Dual Discriminator Weighted Mixture Generative Adversarial Network for image generation

Latent Dirichlet Allocation Based Generative Adversarial Networks.

MTS-DVGAN: Anomaly Detection in Cyber-Physical Systems using a Dual Variational Generative Adversarial Network

Multi-generator GAN learning disconnected manifolds with mutual information

Partial Modal Conditioned GANs for Multi-modal Multi-label Learning with Arbitrary Modal-Missing

GAN Ensemble for Anomaly Detection

Detecting Anomalies using Generative Adversarial Networks on Images

MFGAN: Multimodal Fusion for Industrial Anomaly Detection Using Attention-Based Autoencoder and Generative Adversarial Network

Directed Chain Generative Adversarial Networks

Anomaly Detection by Latent Regularized Dual Adversarial Networks.

Multiple Granularities Generative Adversarial Network for Recognition of Wafer Map Defects

Label-Removed Generative Adversarial Networks Incorporating with K-Means

Generative Models for Novelty Detection: Applications in abnormal event and situational change detection from data series

MRD‐GAN: Multi‐representation discrimination GAN for enhancing the diversity of the generated data