PerforatedCNNs: Acceleration through Elimination of Redundant Convolutions

Abstract

We propose a novel approach to reduce the computational cost of evaluation of convolutional neural networks, a factor that has hindered their deployment in low-power devices such as mobile phones. Inspired by the loop perforation technique from source code optimization, we speed up the bottleneck convolutional layers by skipping their evaluation in some of the spatial positions. We propose and analyze several strategies of choosing these positions. We demonstrate that perforation can accelerate modern convolutional networks such as AlexNet and VGG-16 by a factor of 2x - 4x. Additionally, we show that perforation is complementary to the recently proposed acceleration method of Zhang et al.

NIPS 2016

References in corpus (4)

• Deep Learning with Limited Numerical Precision
• cuDNN: Efficient Primitives for Deep Learning
• Recurrent Models of Visual Attention
• Tensorizing Neural Networks

Cited by in corpus (30)

• Dynamic Channel Pruning: Feature Boosting and Suppression
• Ultimate tensorization: compressing convolutional and FC layers alike
• What is the State of Neural Network Pruning?
• Zero-Cost Proxies for Lightweight NAS
• NISP: Pruning Networks using Neuron Importance Score Propagation
• Insights on representational similarity in neural networks with canonical correlation
• ExpandNets: Linear Over-parameterization to Train Compact Convolutional Networks
• More is Less: A More Complicated Network with Less Inference Complexity
• Pixel-wise Attentional Gating for Parsimonious Pixel Labeling
• QueryDet: Cascaded Sparse Query for Accelerating High-Resolution Small Object Detection
• SegBlocks: Block-Based Dynamic Resolution Networks for Real-Time Segmentation
• SBNet: Sparse Blocks Network for Fast Inference
• Centripetal SGD for Pruning Very Deep Convolutional Networks with Complicated Structure
• Efficient Segmentation: Learning Downsampling Near Semantic Boundaries
• Efficient ConvNets for Analog Arrays
• Spatially Adaptive Inference with Stochastic Feature Sampling and Interpolation
• Learning Versatile Convolution Filters for Efficient Visual Recognition
• Improving Feature Attribution through Input-specific Network Pruning
• Manipulating Identical Filter Redundancy for Efficient Pruning on Deep and Complicated CNN
• PnP-DETR: Towards Efficient Visual Analysis with Transformers
• Perceive, Attend, and Drive: Learning Spatial Attention for Safe Self-Driving
• Dynamic Neural Network Channel Execution for Efficient Training
• SECS: Efficient Deep Stream Processing via Class Skew Dichotomy
• CoDiNet: Path Distribution Modeling with Consistency and Diversity for Dynamic Routing
• Thanks for Nothing: Predicting Zero-Valued Activations with Lightweight Convolutional Neural Networks
• Interpretable Neural Network Decoupling
• Dynamic Multi-path Neural Network
• Channel Pruning Guided by Classification Loss and Feature Importance
• ViP: Virtual Pooling for Accelerating CNN-based Image Classification and Object Detection
• Efficient Structured Pruning and Architecture Searching for Group Convolution