Skip to main content
Article
Robustness Analysis of Classification Using Recurrent Neural Networks with Perturbed Sequential Input
arXiv
  • Guangyi Liu, Leigh University, PA, United States
  • Arash Amini, Leigh University, PA, United States
  • Martin Takac, Mohamed bin Zayed University of Artificial Intelligence
  • Nader Motee, Leigh University, PA, United States
Document Type
Article
Abstract

For a given stable recurrent neural network (RNN) that is trained to perform a classification task using sequential inputs, we quantify explicit robustness bounds as a function of trainable weight matrices. The sequential inputs can be perturbed in various ways, e.g., streaming images can be deformed due to robot motion or imperfect camera lens. Using the notion of the Voronoi diagram and Lipschitz properties of stable RNNs, we provide a thorough analysis and characterize the maximum allowable perturbations while guaranteeing the full accuracy of the classification task. We illustrate and validate our theoretical results using a map dataset with clouds as well as the MNIST dataset. © 2022, CC BY.

DOI
arXiv:2203.05403v1
Publication Date
3-10-2022
Keywords
  • Classification tasks,
  • Lipschitz property,
  • Robot motion,
  • Robustness analysis,
  • Robustness bound,
  • Streaming images,
  • Voronoi diagram properties,
  • Weight matrices,
  • Recurrent neural networks,
  • Machine Learning (cs.LG)
Comments

Preprint: arXiv

Archived with thanks to arXiv

Preprint License: CC by 4.0

Uploaded 18 May 2022

Citation Information
G. Liu, A. Amini, M. Takac, and N. Motee, "Robustness Analysis of Classification Using Recurrent Neural Networks with Perturbed Sequential Input", 2022, arXiv, doi: 10.48550/arXiv.2203.05403