BROWSE
Detail
Conference Paper
Generative Adversarial Network for Robust Regression using Continuous Dataset
Cited 2 time in 
Share 
Share
- Authors
- Yu-Lim Min, Seung-Jin Hong, Hye-jin Kim, Seung-Ik Lee
- Issue Date
- 2020-10
- Citation
- International Conference on Information and Communication Technology Convergence (ICTC) 2020, pp.1209-1211
- Publisher
- IEEE
- Language
- English
- Type
- Conference Paper
- Abstract
- Recently, advanced neural network, which is implementing technical method, has focus on dealing with image classification problems. Unlike classification problems, regression provides a value of output in complex and sophisticated continuous datasets. Though nonlinear models can perform regression better than linear model as Linear Regression(LR), the difficulty to make robust model still remain. In this paper, our purpose is to design training architecture for robust regression. We approach Neural Network known as nonlinear regression to solve limitation of Linear Regression. Additionally, Our architecture uses a new artificial Neural Network(NN) based on adversarial architecture by using the Generator(G) and Discriminator(D). The Discriminator shows the better performance while competing with the Generator and learning regression problem as same time. In evaluation experiments, we compare our proposed model with baseline models including Linear Regression and Neural Network using continuous real world data. We split four datasets into train and test sets as 90:10 and evaluate them by using Mean Squared Error(MSE) function. In summary, our model trained with Generative Adversarial Network(GAN) shows better performance than the baseline models.
- KSP Keywords
- Artificial Neural Network, Classification problems, Image Classification, Nonlinear model, Nonlinear regression, Proposed model, Real-world, Robust model, Robust regression, generative adversarial network, mean square error(MSE)