Analysis of convolutional neural network image classifiers in a rotationally symmetric model: Implementations of the estimates and links to image data sets

This repository contains the Python code required to reproduce the simulation part of the paper "Analysis of convolutional neural network image classifiers in a rotationally symmetric model" from Kohler and Walter (2023) referenced below. The Python version used is Python 3.9.7. This work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under project number 449102119. The mnist-rot image dataset consisting of the real images from which the classes "four" and "nine" were used can be downloaded from the link given below. The paper by Larochelle et al. (2007) linked below describes the dataset in more detail. The link for the original mnist dataset has also been linked below.

Identifier
Source https://tudatalib.ulb.tu-darmstadt.de/handle/tudatalib/3871
Related Identifier https://doi.org/10.1109/TIT.2023.3262745
Related Identifier https://sites.google.com/a/lisa.iro.umontreal.ca/public_static_twiki/variations-on-the-mnist-digits
Related Identifier https://doi.org/10.1145/1273496.1273556
Related Identifier http://yann.lecun.com/exdb/mnist/
Metadata Access https://tudatalib.ulb.tu-darmstadt.de/oai/openairedata?verb=GetRecord&metadataPrefix=oai_datacite&identifier=oai:tudatalib.ulb.tu-darmstadt.de:tudatalib/3871
Provenance
Creator Kohler, Michael; Walter, Benjamin
Publisher TU Darmstadt
Contributor TU Darmstadt
Publication Year 2023
Rights Creative Commons Attribution 4.0; info:eu-repo/semantics/openAccess
OpenAccess true
Contact https://tudatalib.ulb.tu-darmstadt.de/page/contact
Representation
Language English
Resource Type Software
Format application/zip
Discipline Other