Rectification of Corrupted Neural Networks

Imagine if the data set provided for training an artificial neural network turns out to be corrupted. This paper presents a method that can be used to rectify the said neural network after it has been trained but on some corrupted data. In order to rectify the neural network we are provided with a replacement data set for the corrupted data. The proposed method uses the old weights of the corrupted neural network to determine the new weights of the rectified neural network model. Moreover, the proposed method is compared with the present typical method for solving the above stated problem.