(a, b)-grouping pooling

Last updated on Jul 17, 2023
More
Comparison on the effect of (a, b)-grouping pooling with respect to classic pooling operators.

This project offers an implementation for (a, b)-grouping pooling, a generalization of max pooling which focuses on high feature values and improves gradient flow. Implementation for the paper: Rodriguez-Martinez, I., da Cruz Asmus, T., Dimuro, G. P., Herrera, F., Takáč, Z., & Bustince, H. (2023). Generalizing max pooling via (a, b)-grouping functions for Convolutional Neural Networks. Information Fusion, 101893.

Documentation is available in English and Spanish: https://github.com/iosurodri/overlapsAndGroupings

Deep Learning
Iosu Rodríguez Martínez
Iosu Rodríguez Martínez
PHD student on Artificial Intelligence

My research interests include Convolutional Neural Networks, aggregation theory and fuzzy logic.

Publications

Generalizing max pooling via (a, b)-grouping functions for Convolutional Neural Networks
Despite numerous studies on pooling algorithms, max-pooling remains the standard choice in CNNs. In this work, we introduce (a, b)-grouping functions, an extension of grouping functions tailored for real-valued data in CNNs. We present various construction methods for (a, b)-grouping functions and empirically demonstrate their effectiveness by replacing max-pooling in popular CNN architectures, yielding promising results. Our findings highlight the potential of grouping functions as efficient alternatives to max-pooling in CNN feature downsampling.
Iosu Rodríguez Martínez, Tiago Asmus, Graçaliz Dimuro, Francisco Herrera, Zdenko Takáč, Humberto Bustince
Cite Project DOI