5 papers
INN: A Method Identifying Clean-annotated Samples via Consistency Effect in Deep Neural Networks
Dongha Kim, Yongchan Choi, Kunwoong Kim +1
In many classification problems, collecting massive clean-annotated data is not easy, and thus a lot of researches have been done to handle data with noisy labels. Most recent stat…
Kernel-convoluted Deep Neural Networks with Data Augmentation
Minjin Kim, Young-geun Kim, Dongha Kim +2
The Mixup method (Zhang et al. 2018), which uses linearly interpolated data, has emerged as an effective data augmentation tool to improve generalization performance and the robust…
Understanding and Improving Virtual Adversarial Training
Dongha Kim, Yongchan Choi, Yongdai Kim
In semi-supervised learning, virtual adversarial training (VAT) approach is one of the most attractive method due to its intuitional simplicity and powerful performances. VAT finds…
On variation of gradients of deep neural networks
Yongdai Kim, Dongha Kim
We provide a theoretical explanation of the role of the number of nodes at each layer in deep neural networks. We prove that the largest variation of a deep neural network with ReL…
Fast convergence rates of deep neural networks for classification
Yongdai Kim, Ilsang Ohn, Dongha Kim
We derive the fast convergence rates of a deep neural network (DNN) classifier with the rectified linear unit (ReLU) activation function learned using the hinge loss. We consider t…