Method for Generating Training Data for Training a Machine Learning Algorithm

A method is for generating training data for training a machine learning algorithm. The training data respectively include a data point and a data value associated with the data point. The method includes providing first training data for training the machine learning algorithm and approximating a manifold in which at least one part of the data points of the first training data is located. The method further includes determining a structure of the at least one part of the data points of the first training data in the manifold, and generating additional training data based on the determined structure of the at least one part of the data points of the first training data in the manifold.