Hybrid Pruning: Thinner Sparse Networks for Fast Inference on Edge Devices
arXiv:1811.00482
Abstract
We introduce hybrid pruning which combines both coarse-grained channel and fine-grained weight pruning to reduce model size, computation and power demands with no to little loss in accuracy for enabling modern networks deployment on resource-constrained devices, such as always-on security cameras and drones. Additionally, to effectively perform channel pruning, we propose a fast sensitivity test that helps us quickly identify the sensitivity of within and across layers of a network to the output accuracy for target multiplier accumulators (MACs) or accuracy tolerance. Our experiment shows significantly better results on ResNet50 on ImageNet compared to existing work, even with an additional constraint of channels be hardware-friendly number.
References in corpus (5)
- Very Deep Convolutional Networks for Large-Scale Image Recognition
- Channel Pruning for Accelerating Very Deep Neural Networks
- Soft Filter Pruning for Accelerating Deep Convolutional Neural Networks
- Structured Probabilistic Pruning for Convolutional Neural Network Acceleration
- Layer-compensated Pruning for Resource-constrained Convolutional Neural Networks
Cited by in corpus (10)
- Dual Dynamic Inference: Enabling More Efficient, Adaptive and Controllable Deep Inference
- Efficient Visual Recognition with Deep Neural Networks: A Survey on Recent Advances and New Directions
- Pruning-aware Sparse Regularization for Network Pruning
- SQuantizer: Simultaneous Learning for Both Sparse and Low-precision Neural Networks
- Fractional Skipping: Towards Finer-Grained Dynamic CNN Inference
- Weight Pruning via Adaptive Sparsity Loss
- Manipulating Identical Filter Redundancy for Efficient Pruning on Deep and Complicated CNN
- Deep Asymmetric Networks with a Set of Node-wise Variant Activation Functions
- Efficient architecture for deep neural networks with heterogeneous sensitivity
- Low Precision Constant Parameter CNN on FPGA