量化、投影和剪枝

通过执行量化、投影或剪枝来压缩深度神经网络

将 Deep Learning Toolbox™ 和 Deep Learning Toolbox Model Quantization Library 支持包结合使用，通过以下方式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。然后，您可以从这个剪枝过的网络中生成 C/C++ 或 CUDA^® 代码。
对层进行投影，先使用代表训练数据的数据集对层激活执行主成分分析 (PCA)，然后对层的可学习参数应用线性投影。当您使用无库的 C/C++ 代码生成将网络部署到嵌入式硬件时，投影的深度神经网络的前向传导通常会更快。
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，您可以从这个量化的网络中生成 C/C++、CUDA 或 HDL 代码。
为了实现 C/C++ 和 CUDA 代码生成，软件通过将卷积层的权重、偏置和激活量化为 8 位定标整数数据类型，为卷积深度神经网络生成代码。量化是通过向 codegen (MATLAB Coder) 命令提供由 calibrate 函数生成的标定结果文件来执行的。
代码生成不支持由 quantize 函数生成的量化深度神经网络。

函数

剪枝

`taylorPrunableNetwork`	Network that can be pruned by using first-order Taylor approximation (自 R2022a 起)
`forward`	Compute deep learning network output for training (自 R2019b 起)
`predict`	Compute deep learning network output for inference (自 R2019b 起)
`updatePrunables`	Remove filters from prunable layers based on importance scores (自 R2022a 起)
`updateScore`	Compute and accumulate Taylor-based importance scores for pruning (自 R2022a 起)
`dlnetwork`	Deep learning neural network (自 R2019b 起)

投影

`compressNetworkUsingProjection`	Compress neural network using projection (自 R2022b 起)
`neuronPCA`	Principal component analysis of neuron activations (自 R2022b 起)
`unpackProjectedLayers`	Unpack projected layers of neural network (自 R2023b 起)
`ProjectedLayer`	Compressed neural network layer using projection (自 R2023b 起)
`gruProjectedLayer`	Gated recurrent unit (GRU) projected layer for recurrent neural network (RNN) (自 R2023b 起)
`lstmProjectedLayer`	Long short-term memory (LSTM) projected layer for recurrent neural network (RNN) (自 R2022b 起)

量化

`dlquantizer`	Quantize a deep neural network to 8-bit scaled integer data types (自 R2020a 起)
`dlquantizationOptions`	Options for quantizing a trained deep neural network (自 R2020a 起)
`calibrate`	Simulate and collect ranges of a deep neural network (自 R2020a 起)
`quantize`	Quantize deep neural network (自 R2022a 起)
`validate`	Quantize and validate a deep neural network (自 R2020a 起)
`quantizationDetails`	Display quantization details for a neural network (自 R2022a 起)
`estimateNetworkMetrics`	Estimate network metrics for specific layers of a neural network (自 R2022a 起)
`equalizeLayers`	Equalize layer parameters of deep neural network (自 R2022b 起)

App

深度网络量化器

Quantize deep neural network to 8-bit scaled integer data types (自 R2020a 起)

主题

剪枝

Parameter Pruning and Quantization of Image Classification Network
Use parameter pruning and quantization to reduce network size.
Prune Image Classification Network Using Taylor Scores
This example shows how to reduce the size of a deep neural network using Taylor pruning.
Prune Filters in a Detection Network Using Taylor Scores
This example shows how to reduce network size and increase inference speed by pruning convolutional filters in a you only look once (YOLO) v3 object detection network.
Prune and Quantize Convolutional Neural Network for Speech Recognition
This example shows how to compress a convolutional neural nework (CNN) to prepare it for deployment on an embedded system.

投影和知识蒸馏

Compress Neural Network Using Projection
This example shows how to compress a neural network using projection and principal component analysis.
Compress Network for Estimating Battery State of Charge
This example shows how to compress a neural network for predicting the state of charge of a battery using projection and principal component analysis. (自 R2023b 起)
Train Smaller Neural Network Using Knowledge Distillation
This example shows how to reduce the memory footprint of a deep learning network by using knowledge distillation. (自 R2023b 起)

量化

Quantization of Deep Neural Networks
Understand effects of quantization and how to visualize dynamic ranges of network convolution layers.
Quantization Workflow Prerequisites
Products required for the quantization of deep learning networks.
Prepare Data for Quantizing Networks
Supported datastores for quantization workflows.
Quantize Multiple-Input Network Using Image and Feature Data
Quantize Multiple Input Network Using Image and Feature Data

GPU 目标的量化

Generate INT8 Code for Deep Learning Networks (GPU Coder)
Quantize and generate code for a pretrained convolutional neural network.
Quantize Residual Network Trained for Image Classification and Generate CUDA Code
This example shows how to quantize the learnable parameters in the convolution layers of a deep learning neural network that has residual connections and has been trained for image classification with CIFAR-10 data.
Quantize Layers in Object Detectors and Generate CUDA Code
This example shows how to generate CUDA® code for an SSD vehicle detector and a YOLO v2 vehicle detector that performs inference computations in 8-bit integers for the convolutional layers.
Quantize Semantic Segmentation Network and Generate CUDA Code
Quantize Convolutional Neural Network Trained for Semantic Segmentation and Generate CUDA Code

FPGA 目标的量化

Quantize Network for FPGA Deployment (Deep Learning HDL Toolbox)
Reduce the memory footprint of a deep neural network by quantizing the weights, biases, and activations of convolution layers to 8-bit scaled integer data types.
Classify Images on FPGA Using Quantized Neural Network (Deep Learning HDL Toolbox)
This example shows how to use Deep Learning HDL Toolbox™ to deploy a quantized deep convolutional neural network (CNN) to an FPGA.
Classify Images on FPGA by Using Quantized GoogLeNet Network (Deep Learning HDL Toolbox)
This example show how to use the Deep Learning HDL Toolbox™ to deploy a quantized GoogleNet network to classify an image.

CPU 目标的量化

Generate int8 Code for Deep Learning Networks (MATLAB Coder)
Quantize and generate code for a pretrained convolutional neural network.
Generate INT8 Code for Deep Learning Network on Raspberry Pi (MATLAB Coder)
Generate code for deep learning network that performs inference computations in 8-bit integers.

精选示例

Prune Image Classification Network Using Taylor Scores

Reduce the size of a deep neural network using Taylor pruning. By using the taylorPrunableNetwork function to remove convolution layer filters, you can reduce the overall network size and increase the inference speed.

打开实时脚本

Prune Filters in a Detection Network Using Taylor Scores

Reduce network size and increase inference speed by pruning convolutional filters in a you only look once (YOLO) v3 object detection network.

打开实时脚本

Compress Neural Network Using Projection

Compress a neural network using projection and principal component analysis.

打开实时脚本

Quantize Residual Network Trained for Image Classification and Generate CUDA Code

Quantize the learnable parameters in the convolution layers of a deep learning neural network that has residual connections and has been trained for image classification with CIFAR-10 data.

打开实时脚本

Prune and Quantize Semantic Segmentation Network

Reduce the memory footprint of a semantic segmentation network and speed-up inference by compressing the network using pruning and quantization.

打开实时脚本

Explore Quantized Semantic Segmentation Network Using Grad-CAM

Compare the predictions of a quantized semantic segmentation network to the original network using the gradient-weighted class activation mapping (Grad-CAM) interpretability method.

打开实时脚本