PredictiveNet: An energy-efficient convolutional neural network via zero prediction

By Yingyan Lin, Charbel Sakr, Yongjune Kim, Naresh R Shanbhag. 2017

61 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Efficiency & Compression

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Shiji Song
2. Gao Huang
3. Yulin Wang
4. Zhangyang Wang
5. Le Yang
6. Yingyan Lin
7. Song Han
8. Yuan Xie

In Computer Vision and Pattern Recognition

1. Zhangyang Wang
2. Shiji Song
3. Gao Huang
4. Yulin Wang
5. Le Yang
6. Song Han
7. Yuan Xie
8. Yizeng Han

In Efficiency & Compression

1. Zhangyang Wang
2. Gao Huang
3. Shiji Song
4. Yuan Xie
5. Yingyan Lin
6. Guoqi Li
7. Luping P. Shi
8. By Lei Deng

Citations in the corpus

• Dynamic Neural Networks: A Survey (2021, background): “Many input-dependent operations, including pixel-level dynamic computation [114], [275], [276], adaptive channel pruning [277], [278] and early exiting [279], have also been tailored together with hardware for further improving their practical efficiency.”
• SnaPEA: Predictive Early Activation for Reducing Computation in Deep Convolutional Neural Networks (2018, background): “PredictiveNet [34] proposes a skipping mechanism for CNNs.”
• Boosting the Performance of CNN Accelerators with Dynamic Fine-Grained Channel Gating (2019, background): “Predictive-based execution proposed in [36, 37] predicts zeros by first executing the significant bits.”
• Deep k-Means: Re-Training and Parameter Sharing with Harder Cluster Assignments for Compressing Deep Convolutions (2018, methodology): “• In order to bridge the gap between algorithm and hardware design of CNNs, we propose an improved set of energy-aware metrics based on (Sakr et al., 2017).”
• Model Compression and Hardware Acceleration for Neural Networks: A Comprehensive Survey (2020, background): “The preselection of unimportant neurons can be finished by hashing search [241], negative-activation prediction [242], low-precision estimation [243], [244], prechecking neighboring pixels [245], or dimension-reduction estimation [246].”
• SeerNet: Predicting Convolutional Neural Network Feature-Map Sparsity Through Low-Bit Quantization (2019, background): “In [1, 29, 19], the authors designed customized accelerators for DNNs to make early decisions or predictions of skipping unnecessary computation.”