What type of study is this?

September 10, 2025Open Access

ECA110-Pooling: A Comparative Analysis of Pooling Strategies in Convolutional Neural Networks

DCDoru ConstantinUniversitatea Națională de Știință și Tehnologie Politehnica București CBCostel BălcăuUniversitatea Națională de Știință și Tehnologie Politehnica București

Key Points

ECA110-Pooling outperformed traditional pooling methods, achieving higher accuracy and lower error rates.
In experiments across datasets like ImageNet and CIFAR-10, ECA110-Pooling demonstrated superior performance and efficiency.
Comparative benchmarking against state-of-the-art models revealed ECA110-Pooling's competitive accuracy with fewer parameters.
ECA110-Pooling's rule-based design offers practical applicability, balancing computational efficiency with accuracy.

Abstract

In this paper, we introduce and validate ECA110-Pooling, a novel rule-based pooling operator for Convolutional Neural Networks inspired by elementary cellular automata. A systematic comparative study is conducted, benchmarking ECA110-Pooling against conventional pooling methods (MaxPooling, AveragePooling, MedianPooling, MinPooling, KernelPooling) as well as state-of-the-art (SOTA) architectures. Experiments on three benchmark datasets-ImageNet (subset), CIFAR-10, and Fashion-MNIST-across training horizons ranging from 20 to 50,000 epochs demonstrate that ECA110-Pooling consistently achieves higher Top-1 accuracy, lower error rates, and stronger F1-scores than traditional pooling operators, while maintaining computational efficiency comparable to MaxPooling. Furthermore, in direct comparison with SOTA models, ECA110-Pooling delivers competitive accuracy with substantially fewer parameters and reduced training time. These results establish ECA110-Pooling as a validated and principled approach for image classification, bridging the gap between fixed pooling schemes and complex deep architectures. Its interpretable, rule-based design underscores both theoretical relevance and practical applicability in scenarios requiring a balance of accuracy, efficiency, and scalability.

AIに質問

Bookmark

View Full Paper