More Like this

1. Robustness of conditional GANs to noisy labels

   Thekumparampil, Kiran Koshy; Khetan, Ashish; Lin, Zinan; Oh, Sewoong (December 2018, Advances in neural information processing systems)

   We study the problem of learning conditional generators from noisy labeled samples, where the labels are corrupted by random noise. A standard training of conditional GANs will not only produce samples with wrong labels, but also generate poor quality samples. We consider two scenarios, depending on whether the noise model is known or not. When the distribution of the noise is known, we introduce a novel architecture which we call Robust Conditional GAN (RCGAN). The main idea is to corrupt the label of the generated sample before feeding to the adversarial discriminator, forcing the generator to produce samples with clean labels. This approach of passing through a matching noisy channel is justified by accompanying multiplicative approximation bounds between the loss of the RCGAN and the distance between the clean real distribution and the generator distribution. This shows that the proposed approach is robust, when used with a carefully chosen discriminator architecture, known as projection discriminator. When the distribution of the noise is not known, we provide an extension of our architecture, which we call RCGAN-U, that learns the noise model simultaneously while training the generator. We show experimentally on MNIST and CIFAR-10 datasets that both the approaches consistently improve upon baseline approaches, and RCGAN-U closely matches the performance of RCGAN. 

   more » « less
2. Robustness of Conditional GANs to Noisy Labels

   Thekumparampil, Kiran Koshy; Khetan, Ashish; Lin, Zinan; Oh, Sewoong (December 2018, Advances in neural information processing systems)

   We study the problem of learning conditional generators from noisy labeled samples, where the labels are corrupted by random noise. A standard training of conditional GANs will not only produce samples with wrong labels, but also generate poor quality samples. We consider two scenarios, depending on whether the noise model is known or not. When the distribution of the noise is known, we introduce a novel architecture which we call Robust Conditional GAN (RCGAN). The main idea is to corrupt the label of the generated sample before feeding to the adversarial discriminator, forcing the generator to produce samples with clean labels. This approach of passing through a matching noisy channel is justified by corresponding multiplicative approximation bounds between the loss of the RCGAN and the distance between the clean real distribution and the generator distribution. This shows that the proposed approach is robust, when used with a carefully chosen discriminator architecture, known as projection discriminator. When the distribution of the noise is not known, we provide an extension of our architecture, which we call RCGAN-U, that learns the noise model simultaneously while training the generator. We show experimentally on MNIST and CIFAR-10 datasets that both the approaches consistently improve upon baseline approaches, and RCGAN-U closely matches the performance of RCGAN. 

   more » « less
3. Flash drought onset over the contiguous United States: sensitivity of inventories and trends to quantitative definitions

   https://doi.org/10.5194/hess-25-565-2021  

   Osman, Mahmoud; Zaitchik, Benjamin F.; Badr, Hamada S.; Christian, Jordan I.; Tadesse, Tsegaye; Otkin, Jason A.; Anderson, Martha C. (January 2021, Hydrology and Earth System Sciences)

   null (Ed.)

   Abstract. The term “flash drought” is frequently invoked to describe droughts thatdevelop rapidly over a relatively short timescale. Despite extensive andgrowing research on flash drought processes, predictability, and trends,there is still no standard quantitative definition that encompasses allflash drought characteristics and pathways. Instead, diverse definitionshave been proposed, supporting wide-ranging studies of flash drought butcreating the potential for confusion as to what the term means and how tocharacterize it. Use of different definitions might also lead to differentconclusions regarding flash drought frequency, predictability, and trendsunder climate change. In this study, we compared five previously publisheddefinitions, a newly proposed definition, and an operational satellite-baseddrought monitoring product to clarify conceptual differences and toinvestigate the sensitivity of flash drought inventories and trends to thechoice of definition. Our analyses indicate that the newly introduced SoilMoisture Volatility Index definition effectively captures flash droughtonset in both humid and semi-arid regions. Analyses also showed thatestimates of flash drought frequency, spatial distribution, and seasonalityvary across the contiguous United States depending upon which definition is used.Definitions differ in their representation of some of the largest and mostwidely studied flash droughts of recent years. Trend analysis indicates thatdefinitions that include air temperature show significant increases in flashdroughts over the past 40 years, but few trends are evident fordefinitions based on other surface conditions or fluxes. These resultsindicate that “flash drought” is a composite term that includes severaltypes of events and that clarity in definition is critical when monitoring,forecasting, or projecting the drought phenomenon. 

   more » « less
4. Twin Auxiliary Classifiers GAN

   Mingming Gong, Yanwu Xu (January 2019, Advances in neural information processing systems)

   null (Ed.)

   Conditional generative models enjoy remarkable progress over the past few years. One of the popular conditional models is Auxiliary Classifier GAN (AC-GAN), which generates highly discriminative images by extending the loss function of GAN with an auxiliary classifier. However, the diversity of the generated samples by AC-GAN tends to decrease as the number of classes increases, hence limiting its power on large-scale data. In this paper, we identify the source of the low diversity issue theoretically and propose a practical solution to solve the problem. We show that the auxiliary classifier in AC-GAN imposes perfect separability, which is disadvantageous when the supports of the class distributions have significant overlap. To address the issue, we propose Twin Auxiliary Classifiers Generative Adversarial Net (TAC-GAN) that further benefits from a new player that interacts with other players (the generator and the discriminator) in GAN. Theoretically, we demonstrate that TAC-GAN can effectively minimize the divergence between the generated and real-data distributions. Extensive experimental results show that our TAC-GAN can successfully replicate the true data distributions on simulated data, and significantly improves the diversity of class-conditional image generation on real datasets. 

   more » « less
5. PMGAN: Paralleled Mix-Generator Generative Adversarial Networks with Balance Control

   Xiao, Xia; Rajasekaran, Sanguthevar (January 2018, The 27th International Conference on Artificial Neural Networks)

   A Generative Adversarial Network (GAN) is an unsupervised generative framework to generate a sample distribution that is identical to the data distribution. Recently, mix strategy multi-generator/discriminator GANs have been shown to outperform single pair GANs. However, the mixed model suffers from the problem of linearly growing training time. Also, imbalanced training among generators makes it difficult to parallelize. In this paper, we propose a balanced mix-generator GAN that works in parallel by mixing multiple disjoint generators to approximate the real distribution. The weights of the discriminator and the classifier are controlled by a balance strategy. We also present an efficient loss function, to force each generator to embrace few modes with a high probability. Our model is naturally adaptive to large parallel computation frameworks. Each generator can be trained on multiple GPUs asynchronously. We have performed extensive experiments on synthetic datasets, MNIST1000, CIFAR-10, and ImageNet. The results establish that our model can achieve the state-of-the-art performance (in terms of the modes coverage and the inception score), with significantly reduced training time. We also show that the missing mode problem can be relieved with a growing number of generators. 

   more » « less