This paper introduces channel gating, a dynamic, fine-grained, and hardware efficient pruning scheme to reduce the computation cost for convolutional neural networks (CNNs). Channel gating identifies regions in the features that contribute less to the classification result, and skips the computation on a subset of the input channels for these ineffective regions. Unlike static network pruning, channel gating optimizes CNN inference at run-time by exploiting input-specific characteristics, which allows substantially reducing the compute cost with almost no accuracy loss. We experimentally show that applying channel gating in state-of-the-art networks achieves 2.7-8.0x

This paper introduces channel gating, a dynamic, fine-grained, and hardware－efficient pruning scheme to reduce the computation cost for convolutional neural networks (CNNs). Channel gating identifies regions in the features that contribute less to the classification result, and skips the computation on a subset of the input channels for these ineffective regions. Unlike static network pruning, channel gating optimizes CNN inference at run-time by exploiting input-specific characteristics, which allows substantially reducing the compute cost with almost no accuracy loss. We experimentally show that applying channel gating in state-of-the-art networks achieves 2.7-8.0x Combining our method with knowledge distillation reduces the compute cost of ResNet-18 by 2.6x without accuracy drop on ImageNet.

Originality: The proposed idea is not very different from other dynamic pruning methods. In my opinion the main contribution is the reduced amount of extra computation needed for the pruning that allows interesting computational gains and the GPU friendly way of pruning based on channels. The use of channel grouping to avoid a biased selection of the channels is also quite interesting. In dynamic pruning authors should also cite: [Convolutional Networks with Adaptive Inference Graphs. Quality: The proposed contribution makes sense and is justified by interesting experiments on CIFAR10 and Imagenet.

The paper presents a simple yet effective way to reduce computation by only computing a sub-part of the inner-products. This idea results in realization speedups as confirmed by an ASIC design. Given the similarity with some existing work on dynamic pruning, I recommend acceptance as a poster.

This paper introduces channel gating, a dynamic, fine-grained, and hardware－efficient pruning scheme to reduce the computation cost for convolutional neural networks (CNNs). Channel gating identifies regions in the features that contribute less to the classification result, and skips the computation on a subset of the input channels for these ineffective regions. Unlike static network pruning, channel gating optimizes CNN inference at run-time by exploiting input-specific characteristics, which allows substantially reducing the compute cost with almost no accuracy loss. We experimentally show that applying channel gating in state-of-the-art networks achieves 2.7-8.0x Combining our method with knowledge distillation reduces the compute cost of ResNet-18 by 2.6x without accuracy drop on ImageNet.

This paper introduces channel gating, a dynamic, fine-grained, and hardware－efficient pruning scheme to reduce the computation cost for convolutional neural networks (CNNs). Channel gating identifies regions in the features that contribute less to the classification result, and skips the computation on a subset of the input channels for these ineffective regions. Unlike static network pruning, channel gating optimizes CNN inference at run-time by exploiting input-specific characteristics, which allows substantially reducing the compute cost with almost no accuracy loss. We experimentally show that applying channel gating in state-of-the-art networks achieves 2.7-8.0x Combining our method with knowledge distillation reduces the compute cost of ResNet-18 by 2.6x without accuracy drop on ImageNet.