Evaluation of the Learning of Convolutional Neural Networks Using Information Theoretical Measures
This repository contains the code regarding our submitted publication: "Evaluating the Learning Procedure of CNNs Through a Sequence of Prognostic Tests Utilizing Information Theoretical Measures" which was submitted to the MDPI J. Entropy.
Preprints of the publication accompanying this code will be available at: https://arxiv.org (publication was submitted to journal and will be updated; please check whether published already before citing).
Usage Note: users need to creat a file folder in the same directory where the Python code is located, with the following directory structure: Figures/kernel_3x3/1-layer, Figures/kernel_3x3/2-layer, ... for all possible number of convolutional layers users want to try. Same folder directory structures should also be created for other kernel sizes, such as 5x5, 7x7 and 9x9 if the users want to test.
Xiyu Shi <x.shi@lboro.ac.uk), and Varuna De-Silva <v.d.de-silva@lboro.ac.uk)
18 December 2021