5 Hits in 1.6 sec

On Adversarial Mixup Resynthesis [article]

Christopher Beckham, Sina Honari, Vikas Verma, Alex Lamb, Farnoosh Ghadiri, R Devon Hjelm, Yoshua Bengio, Christopher Pal
2019 arXiv   pre-print
We explore models that are capable of combining the attributes of multiple inputs such that a resynthesised output is trained to fool an adversarial discriminator for real versus synthesised data.  ...  Figure 2 : 2 The unsupervised version of adversarial mixup resynthesis (AMR).  ...  AE+GAN = adversarial reconstruction auto-encoder (Equation 2); AMR = adversarial mixup resynthesis (ours); ACAI = adversarially constrained auto-encoder interpolation (Berthelot* et al., 2019)) Method  ... 
arXiv:1903.02709v4 fatcat:5cpnsyxp75fnrfynxrayer5j5a

On the benefits of defining vicinal distributions in latent space [article]

Puneet Mangla, Vedant Singh, Shreyas Jayant Havaldar, Vineeth N Balasubramanian
2021 arXiv   pre-print
Our empirical studies on CIFAR-10, CIFAR-100, and Tiny-ImageNet demonstrate that models trained by performing mixup in the latent manifold learned by VAEs are inherently more robust to various input corruptions  ...  We propose a new approach - VarMixup (Variational Mixup) - to better sample mixup images by using the latent manifold underlying the data.  ...  Adversarial Mixup Resynthesis (Beckham et al., 2019) attempted mixing latent codes used by autoencoders through an arbitrary mixing mechanism that can recombine codes from different inputs to produce  ... 
arXiv:2003.06566v4 fatcat:c2jsf7bpb5cx3krvac3fxxs5bi

How Does Mixup Help With Robustness and Generalization? [article]

Linjun Zhang, Zhun Deng, Kenji Kawaguchi, Amirata Ghorbani, James Zou
2021 arXiv   pre-print
Mixup is a popular data augmentation technique based on taking convex combinations of pairs of examples and their labels.  ...  For robustness, we show that minimizing the Mixup loss corresponds to approximately minimizing an upper bound of the adversarial loss.  ...  Mixup Resynthesis The derived regularization terms are then used to demonstrate why Mixup has improved generalization and robustness against one-step adversarial examples.  ... 
arXiv:2010.04819v4 fatcat:ftapdhfseffffci2ztxcq546re

Interpolation Consistency Training for Semi-supervised Learning

Vikas Verma, Alex Lamb, Juho Kannala, Yoshua Bengio, David Lopez-Paz
2019 Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence  
Our experiments show that ICT achieves state-of-the-art performance when applied to standard neural network architectures on the CIFAR-10 and SVHN benchmark dataset.  ...  In section 2.3, I describe how adversarial training leads to poor performance on unperturbed samples and the justification and summary of Interpolated Adversarial Training (Publication III), which is a  ...  In section 2.2, I briefly describe the working principle of the Auto-Encoders, which is followed by the motivation and summary of Adversarial Mixup Resynthesizer (Publication II).  ... 
doi:10.24963/ijcai.2019/504 dblp:conf/ijcai/VermaLKBL19 fatcat:bucjagfaybei5boup7b56yqngu

Multi-SpectroGAN: High-Diversity and High-Fidelity Spectrogram Generation with Adversarial Style Combination for Speech Synthesis [article]

Sang-Hoon Lee, Hyun-Wook Yoon, Hyeong-Rae Noh, Ji-Hoon Kim, Seong-Whan Lee
2020 arXiv   pre-print
from text sequences with only adversarial feedback.  ...  While generative adversarial networks (GANs) based neural text-to-speech (TTS) systems have shown significant improvement in neural speech synthesis, there is no TTS system to learn to synthesize speech  ...  Similar to (Beckham et al. 2019) interpolating the hidden state of the autoencoder for adversarial mixup resynthesis, we use two types of mixing, binary selection between style embeddings, and manifold  ... 
arXiv:2012.07267v1 fatcat:cms2ugs23jhpvnneq57yjrk2fy