剪枝、投影和量化

压缩深度神经网络，减少网络内存占用，并为生成代码准备网络

将 Deep Learning Toolbox™ 与 Deep Learning Toolbox Model Compression Library 支持包结合使用，以减少深度神经网络的内存占用和计算要求：

  • 使用一阶泰勒逼近从卷积层中对滤波器进行剪枝。

  • 通过对层激活执行主成分分析 (PCA) 来对层进行投影。

  • 将层的权重、偏置和激活量化为精度降低的缩放整数数据类型。

然后，您可以从压缩后的网络生成代码，并将其部署到所需的硬件。

突出显示的链接

Simplified illustration of compression. On the left is a sketch of a large neural network with a label indicating the network is 20 MB. An arrow points to a second sketch on the right, which shows a smaller model inside a box. A label indicates the smaller network is 5 MB.

类别

  • 网络压缩快速入门
    学习 Deep Learning Toolbox Model Compression Library 的基础知识
  • 剪枝
    使用一阶泰勒逼近对网络滤波器进行剪枝；减少可学习参数的数量
  • 投影
    使用主成分分析 (PCA) 对网络层进行投影；减少可学习参数的数量
  • 量化
    将网络参数量化为精度降低的数据类型；为生成定点代码准备深度学习网络
  • 网络压缩应用
    在端到端工作流中探索深度学习模型压缩

精选示例

Analyze and Compress 1-D Convolutional Neural Network

Analyze and Compress 1-D Convolutional Neural Network

Analyze 1-D convolutional network for compression and compress it using Taylor pruning and projection.

Compress Sequence Classification Network for Road Damage Detection

Compress Sequence Classification Network for Road Damage Detection

Compress network to meet memory requirement using pruning, projection, and quantization.

Compress Deep Learning Network for Battery State of Charge Estimation

Compress Deep Learning Network for Battery State of Charge Estimation

Compress a neural network for predicting the state of charge of a battery using projection.