ISCA'18 Prediction based execution on Deep Neural Networks
提出了一种基于预测（prediction based）的DNN执行模型，无精度损失。主要预测做法是在计算时对I或W的高低位拆分，利用高位预测，利用低位计算。 USPE：统一的串行PE，解决了高低位宽度不同的问题，减小面积开销 scale-out Design：我的理解是探索合适的复用方法，既能有比较大的利用率，又能使得memory access不是瓶颈。 图1：理论上可以跳过的0。我对此表示疑惑，
相关文章
- 1. comments on"An Experimental Study on Speech Enhancement Based on Deep Neural Networks"(一)
- 2. An Experimental Study on Speech Enhancement Based on Deep Neural Networks学习摘要
- 3. Flow Prediction in Spatio-Temporal Networks Based on Multitask Deep Learning
- 4. Weighted-Entropy-based Quantization for Deep Neural Networks
- 5. DeepRebirth: Accelerating Deep Neural Network Execution on Mobile Devices
- 6. Inductive Matrix Completion Based on Graph Neural Networks
- 7. 13、Probabilistic Short-term Wind Power Forecasting Based on Deep Neural Networks
- 8. Neural Networks and Deep Learning -- Class 4: Deep Neural Networks
- 9. CVPR 2017 On human motion prediction using recurrent neural networks 论文解读
- 10. Improving Deep Neural Networks
- 更多相关文章...
- • Docker 安装 MySQL - Docker教程
- • Docker 安装 MongoDB - Docker教程
- • Flink 数据传输及反压详解
- • RxJava操作符（一）Creating Observables
相关标签/搜索
每日一句
-
每一个你不满意的现在，都有一个你没有努力的曾经。
欢迎关注本站公众号,获取更多信息
相关文章
- 1. comments on"An Experimental Study on Speech Enhancement Based on Deep Neural Networks"(一)
- 2. An Experimental Study on Speech Enhancement Based on Deep Neural Networks学习摘要
- 3. Flow Prediction in Spatio-Temporal Networks Based on Multitask Deep Learning
- 4. Weighted-Entropy-based Quantization for Deep Neural Networks
- 5. DeepRebirth: Accelerating Deep Neural Network Execution on Mobile Devices
- 6. Inductive Matrix Completion Based on Graph Neural Networks
- 7. 13、Probabilistic Short-term Wind Power Forecasting Based on Deep Neural Networks
- 8. Neural Networks and Deep Learning -- Class 4: Deep Neural Networks
- 9. CVPR 2017 On human motion prediction using recurrent neural networks 论文解读
- 10. Improving Deep Neural Networks