Efficient machine unlearning using general adversarial network

Abstract

According to a recent study conducted by Forbes, the collective world data is expected to raise by 175 ZB in 2025 accounting to a 61% increase in data generation. As human, we are favoured with the nature of forgetting irrelevant data. The neurons in our brain are designed to filtrate the noise in the information in order to create more space to store relevant data. This neurophysical principle is witnessing its significance in the field of artificial intelligence due to data intensification. There is a great need for developing systems that have the capability to forget the data on the user’s request without compromising its system accuracy as data are the backbone of any deep learning system. Our approach combines the process of batching training data and general adversarial networks for data augmentation networks to achieve a complete and faster data forgetting model.