Directed-Weighting Group Lasso for Eltwise Blocked CNN Pruning
This work addresses a specific bottleneck in CNN pruning for multi-branch networks, offering incremental improvements in compression and speed for deep learning applications.
The paper tackled the problem of limited compression rates in filter-wise pruning for eltwise layers in CNNs by proposing Directed-Weighting Group Lasso (DWGL), which enforces index-wise incremental coefficients to boost pruning efficiency. It achieved compression rates up to 75.34% on ResNet-56 with minimal error increments (e.g., 0.94% on CIFAR-10) and sped up networks by up to 4 times on ImageNet.
Eltwise layer is a commonly used structure in the multi-branch deep learning network. In a filter-wise pruning procedure, due to the specific operation of the eltwise layer, all its previous convolutional layers should vote for which filters by index should be pruned. Since only an intersection of the voted filters is pruned, the compression rate is limited. This work proposes a method called Directed-Weighting Group Lasso (DWGL), which enforces an index-wise incremental (directed) coefficient on the filterlevel group lasso items, so that the low index filters getting high activation tend to be kept while the high index ones tend to be pruned. When using DWGL, much fewer filters are retained during the voting process and the compression rate can be boosted. The paper test the proposed method on the ResNet series networks. On CIFAR-10, it achieved a 75.34% compression rate on ResNet-56 with a 0.94% error increment, and a 52.06% compression rate on ResNet-20 with a 0.72% error increment. On ImageNet, it achieved a 53% compression rate with ResNet-50 with a 0.6% error increment, speeding up the network by 2.23 times. Furthermore, it achieved a 75% compression rate on ResNet-50 with a 1.2% error increment, speeding up the network by 4 times.