Training dataset reduction on generative adversarial network

In recent years, generative model using neural network (GAN) has become an interesting field in machine learning. However, a study that investigates the effect of reducing training dataset for GAN model has not been conducted, while it is known that collecting images for training dataset requires a lot of human labor work. In this research, series of experiments with various amount of dataset have been conducted to get the idea of how small is the amount of dataset required for a GAN to work. It has been shown that the reduction to around fifty thousand images of dataset has gained a better result than a full amount dataset. Additionally, a new evaluation method for quantifying the performance of GAN network was also proposed, which can be considered later as another evaluation method for GAN framework.