A "Network Pruning Network" Approach to Deep Model Compression

Abstract

We present a filter pruning approach for deep model compression, using a multitask network. Our approach is based on learning a a pruner network to prune a pre-trained target network. The pruner is essentially a multitask deep neural network with binary outputs that help identify the filters from each layer of the original network that do not have any significant contribution to the model and can therefore be pruned. The pruner network has the same architecture as the original network except that it has a multitask/multi-output last layer containing binary-valued outputs (one per filter), which indicate which filters have to be pruned. The pruner's goal is to minimize the number of filters from the original network by assigning zero weights to the corresponding output feature-maps. In contrast to most of the existing methods, instead of relying on iterative pruning, our approach can prune the network (original network) in one go and, moreover, does not require specifying the degree of pruning for each layer (and can learn it instead). The compressed model produced by our approach is generic and does not need any special hardware/software support. Moreover, augmenting with other methods such as knowledge distillation, quantization, and connection pruning can increase the degree of compression for the proposed approach. We show the efficacy of our proposed approach for classification and object detection tasks.

Network Collaborative Pruning Method for Hyperspectral Image Classification Based on Evolutionary Multi-Task Optimization

“…. , m L [40]. m i represents the pruning mask of each layer of the network, which is usually represented by a binary matrix with the same dimension as w i .…”

Absorption Pruning of Deep Neural Network for Object Detection in Remote Sensing Imagery

“…As remote sensing imagery has become easier to obtain, many researchers have tried to use remote sensing data to drive intelligent applications based on DCNNs [3]. However, the over-parametrization problem of DCNNs hinders their application on resource-constrained remote sensing platforms [12][13][14]. Moreover, the work in the paper [22] has shown that the redundant parameters of DCNNs are necessary during network training.…”

“…In addition, spaceborne or airborne platforms usually do not have high-speed Internet, and transmitting many remote sensing images to ground servers for processing will significantly consume valuable communication bandwidth. Therefore, the amount of parameters in DCNNs has become a bottleneck for object detection on resource-constrained platforms [11][12][13][14][15][16].…”