Path classification by stochastic linear recurrent neural networks

DOI10.1186/S13662-022-03686-9MaRDI QIDQ6374664zbMATH OpenFDO

Authors Wiebke Bartolomaeus, Youness Boutaib, Sandra Nestler, Holger Rauhut

Publication date 6 August 2021

Full work available at URL https://doi.org/10.1186/S13662-022-03686-9

Mathematics Subject Classification ID

Signatures and data streams (60L10) Computational learning theory (68Q32) Artificial neural networks and deep learning (68T07) Neural networks for/in biological studies, artificial life and related topics (92B20)

Abstract: We investigate the functioning of a classifying biological neural network from the perspective of statistical learning theory, modelled, in a simplified setting, as a continuous-time stochastic recurrent neural network (RNN) with identity activation function. In the purely stochastic (robust) regime, we give a generalisation error bound that holds with high probability, thus showing that the empirical risk minimiser is the best-in-class hypothesis. We show that RNNs retain a partial signature of the paths they are fed as the unique information exploited for training and classification tasks. We argue that these RNNs are easy to train and robust and back these observations with numerical experiments on both synthetic and real data. We also exhibit a trade-off phenomenon between accuracy and robustness.

This page was built for publication: Path classification by stochastic linear recurrent neural networks

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6374664)