Data Augmentation using CA Evolved GANs
Kaitav Nayankumar Mehta, Ziad Kobti, Kathryn Pfaff, Susan Fox · 2019
Mining medical data images have great potential for exploring hidden patterns in the medical domain. Medical data are heterogeneous which involves images to a great extent like MRI, ECG or Stroke effects etc. Knowledge discovery from such data can improve the diagnostic technique. However, to make the machine learn from such datasets requires large data. In the low-data regime, machine learning algorithms work poorly. Data Augmentation alleviates this by using existing data more effectively, but standard data augmentation produces only limited alternative data. Recent developments in Deep Learning field is noteworthy when it comes to learning probability distribution of points through neural networks, and one of key part for such progress is because of Generative Adversarial Networks(GANs). In this paper, we propose an evolutionary training technique using a cultural algorithm(CA) for neuro-evolution of deep task oriented GANs to find the best architecture for the given dataset. This architecture will help in generating similar but completely new data images which can be further used for training diagnostic Neural Networks. We have compared our approach with the Genetic Algorithm(GA) based neuro-evolution of GANs and show that CA based neuro-evolution of GANs evolves architecture which can generate a higher number of stroke-face images with better resolution when there is low data of original stroke faces.