Representative Datasets: The Perceptron Case.
Year: 2019
Type: article
Abstract: One of the main drawbacks of the practical use of neural networks is the long time needed in \nthe training process. Such training process consists in an iterative change of parameters trying to \nminimize a loss function. These changes are driven by a dataset, which can be seen as a set of \nlabeled points in an n-dimensional space. In this paper, we explore the concept of representative \ndataset which is smaller than the original dataset and satisfies a nearness condition independent of \nisometric transformations. The representativeness is measured using persistence diagrams due to its \ncomputational efficiency. We also prove that the accuracy of the learning process of a neural network \non a representative dataset is comparable with the accuracy on the original dataset when the neural \nnetwork architecture is a perceptron and the loss function is the mean squared error. These theoretical \nresults accompanied with experimentation open a door to the size reduction of the dataset in order to \ngain time in the training process of any neural network.
Language: English
Subfield: Computational Theory and Mathematics
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2923289908
