Universität Rostock, 2020
Abstract: In the field of machine learning, a distinction is usually made between the training and deployment phases when training neural networks. However, if the data base changes with respect to a domain, the training phase for a neural network has to be performed completely new and already learned knowledge is completely ignored. This thesis deals with alternative learning methods, where the goal is to make the learning of a neural network more efficient with respect to different parameters, such as training time or required training examples.
master thesis free access