|
Artificial neural networks (ANNs) are vital for deep learning but often require many parameters, leading to inefficiency and overfitting. This research presents a novel ANN architecture inspired by biological dendrites and their input sampling, which enhances robustness and achieves comparable or superior performance on image classification tasks with fewer parameters. The improved efficiency stems from a learning strategy where nodes respond to multiple classes, unlike traditional class-specific ANNs. These findings highlight the potential of dendritic features to improve the precision and resilience of artificial intelligence.
|
|