• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>International Conference on Artificial Neural Networks >On Practical Approach to Uniform Quantization of Non-redundant Neural Networks
【24h】

On Practical Approach to Uniform Quantization of Non-redundant Neural Networks

机译：非冗余神经网络均匀量化的实用方法

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

The neural network quantization is highly desired procedure to perform before running the neural networks on mobile devices. Quantization without fine-tuning leads to accuracy drop of the model, whereas commonly used training with quantization is done on the full set of the labeled data and therefore is both time- and resource-consuming. Real life applications require simplification and acceleration of the quantization procedure that will maintain the accuracy of full-precision neural network, especially for modern mobile neural network architectures like Mobilenet-vl, MobileNet-v2 and MNAS. Here we present two methods to significantly optimize the training with the quantization procedure. The first one is introducing the trained scale factors for discretization thresholds that are separate for each filter. The second one is based on mutual rescaling of consequent depth-wise separable convolution and convolution layers. Using the proposed techniques, we quantize the modern mobile architectures of neural networks with the set of train data of only ～10% of the total ImageNet 2012 sample. Such reduction of the train dataset size and a small number of trainable parameters allow to fine-tune the network for several hours while maintaining the high accuracy of the quantized model (the accuracy drop was less than 0.5%).
机译：在移动设备上运行神经网络之前，执行神经网络量化是非常需要执行的过程。在没有微调的情况下进行量化会导致模型的准确性下降，而通常使用量化进行的训练是在完整的标记数据集上进行的，因此既浪费时间又浪费资源。现实生活中的应用要求简化和加速量化过程，以保持全精度神经网络的准确性，尤其是对于像Mobilenet-vl，MobileNet-v2和MNAS这样的现代移动神经网络体系结构。在这里，我们提出两种方法来通过量化程序显着优化训练。第一个方法是针对离散化阈值引入经过训练的比例因子，该离散度阈值对于每个滤波器都是独立的。第二个是基于结果深度可分离卷积和卷积层的相互缩放。使用提出的技术，我们使用仅占ImageNet 2012总样本约10％的训练数据集来量化神经网络的现代移动体系结构。火车数据集大小的这种减少和少量的可训练参数允许在保持量化模型的高精度（精度下降小于0.5％）的同时对网络进行数小时的微调。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号