Advanced deep architecture pruning using single-filter performance

Pruning the parameters and structure of neural networks reduces computational complexity, energy consumption, and latency during inference. Recently, an underlying mechanism for successful deep learning (DL) was presented based on a method that quantitatively measures the single-filter performance in each layer of a DL architecture, and a unique comprehensive mechanism of how deep learning works was presented. This statistical mechanics inspired viewpoint enables one to reveal the macroscopic behavior of the entire network from the microscopic performance of each filter and its cooperative behavior. Herein we demonstrate how this understanding paves the path to high quenched dilution of the convolutional layers of deep architectures without affecting their overall accuracy using the applied filter's cluster connections (AFCC). AFCC is exemplified on VGG-11 and EfficientNet-B0 architectures trained on CIFAR-100, and its high pruning outperforms other techniques using the same pruning magnitude. Additionally, this technique is broadened to single-nodal performance and high pruning of fully connected layers, suggesting a possible implementation to considerably reduce the complexity of overparametrized AI tasks.