Pruning by Active Attention Manipulation
This work addresses the need for efficient model compression in deep learning, offering a novel pruning approach that is incremental but shows strong performance gains over existing methods.
The paper tackles the problem of filter pruning in CNNs by introducing PAAM, a method that learns filter importance scores during training using attention mechanisms, resulting in improved accuracy and significant parameter reduction on datasets like CIFAR-10 and ImageNet, with gains such as 1.02% accuracy increase and 52.3% parameter reduction on ResNet56.
Filter pruning of a CNN is typically achieved by applying discrete masks on the CNN's filter weights or activation maps, post-training. Here, we present a new filter-importance-scoring concept named pruning by active attention manipulation (PAAM), that sparsifies the CNN's set of filters through a particular attention mechanism, during-training. PAAM learns analog filter scores from the filter weights by optimizing a cost function regularized by an additive term in the scores. As the filters are not independent, we use attention to dynamically learn their correlations. Moreover, by training the pruning scores of all layers simultaneously, PAAM can account for layer inter-dependencies, which is essential to finding a performant sparse sub-network. PAAM can also train and generate a pruned network from scratch in a straightforward, one-stage training process without requiring a pre-trained network. Finally, PAAM does not need layer-specific hyperparameters and pre-defined layer budgets, since it can implicitly determine the appropriate number of filters in each layer. Our experimental results on different network architectures suggest that PAAM outperforms state-of-the-art structured-pruning methods (SOTA). On CIFAR-10 dataset, without requiring a pre-trained baseline network, we obtain 1.02% and 1.19% accuracy gain and 52.3% and 54% parameters reduction, on ResNet56 and ResNet110, respectively. Similarly, on the ImageNet dataset, PAAM achieves 1.06% accuracy gain while pruning 51.1% of the parameters on ResNet50. For Cifar-10, this is better than the SOTA with a margin of 9.5% and 6.6%, respectively, and on ImageNet with a margin of 11%.