操作

开发自定义深度学习函数

对于大多数任务，您可以使用内置层。如果没有您的任务所需的内置层，则可以定义您自己的自定义层。您可以使用自定义输出层指定自定义损失函数，并定义具有可学习参数和状态参数的自定义层。定义自定义层后，您可以检查该层是否有效，是否与 GPU 兼容，以及是否输出正确定义的梯度。要了解详细信息，请参阅定义自定义深度学习层。要查看支持的层的列表，请参阅深度学习层列表。

如果 trainingOptions 函数不提供任务所需的训练选项，或者自定义输出层不支持所需的损失函数，则您可以定义自定义训练循环。对于层图不支持的模型，您可以将自定义模型定义为函数。要了解详细信息，请参阅定义自定义训练循环、损失函数和网络。

使用深度学习运算为自定义层、训练循环和模型函数开发 MATLAB^® 代码。

函数

全部展开

自动微分

`dlarray`	Deep learning array for customization (自 R2019b 起)
`dims`	`dlarray` 的维度标签 (自 R2019b 起)
`finddim`	Find dimensions with specified label (自 R2019b 起)
`stripdims`	Remove `dlarray` data format (自 R2019b 起)
`extractdata`	从 `dlarray` 中提取数据 (自 R2019b 起)
`isdlarray`	Check if object is `dlarray` (自 R2020b 起)

深度学习运算

`dlconv`	Deep learning convolution (自 R2019b 起)
`dltranspconv`	Deep learning transposed convolution (自 R2019b 起)
`lstm`	Long short-term memory (自 R2019b 起)
`gru`	Gated recurrent unit (自 R2020a 起)
`attention`	Dot-product attention (自 R2022b 起)
`embed`	Embed discrete data (自 R2020b 起)
`fullyconnect`	Sum all weighted input data and apply a bias (自 R2019b 起)
`dlode45`	Deep learning solution of nonstiff ordinary differential equation (ODE) (自 R2021b 起)
`batchnorm`	Normalize data across all observations for each channel independently (自 R2019b 起)
`crosschannelnorm`	Cross channel square-normalize using local responses (自 R2020a 起)
`groupnorm`	Normalize data across grouped subsets of channels for each observation independently (自 R2020b 起)
`instancenorm`	Normalize across each channel for each observation independently (自 R2021a 起)
`layernorm`	Normalize data across all channels for each observation independently (自 R2021a 起)
`avgpool`	Pool data to average values over spatial dimensions (自 R2019b 起)
`maxpool`	Pool data to maximum value (自 R2019b 起)
`maxunpool`	Unpool the output of a maximum pooling operation (自 R2019b 起)

激活

`relu`	应用修正线性单元激活 (自 R2019b 起)
`leakyrelu`	Apply leaky rectified linear unit activation (自 R2019b 起)
`gelu`	Apply Gaussian error linear unit (GELU) activation (自 R2022b 起)
`softmax`	Apply softmax activation to channel dimension (自 R2019b 起)
`sigmoid`	应用 sigmoid 激活 (自 R2019b 起)

损失运算

`crossentropy`	Cross-entropy loss for classification tasks (自 R2019b 起)
`l1loss`	L₁ loss for regression tasks (自 R2021b 起)
`l2loss`	L₂ loss for regression tasks (自 R2021b 起)
`huber`	Huber loss for regression tasks (自 R2021a 起)
`mse`	Half mean squared error (自 R2019b 起)
`ctc`	Connectionist temporal classification (CTC) loss for unaligned sequence classification (自 R2021a 起)

函数加速

`dlaccelerate`	Accelerate deep learning function for custom training loops (自 R2021a 起)
`AcceleratedFunction`	Accelerated deep learning function (自 R2021a 起)
`clearCache`	Clear accelerated deep learning function trace cache (自 R2021a 起)

主题

自动微分

List of Functions with dlarray Support
View the list of functions that support dlarray objects.
Automatic Differentiation Background
Learn how automatic differentiation works.
Use Automatic Differentiation In Deep Learning Toolbox
How to use automatic differentiation in deep learning.

模型函数

Train Network Using Model Function
This example shows how to create and train a deep learning network by using functions rather than a layer graph or a dlnetwork.
Update Batch Normalization Statistics Using Model Function
This example shows how to update the network state in a network defined as a function.
Make Predictions Using Model Function
This example shows how to make predictions using a model function by splitting data into mini-batches.
Initialize Learnable Parameters for Model Function
Learn how to initialize learnable parameters for custom training loops using a model function.

深度学习函数加速

Deep Learning Function Acceleration for Custom Training Loops
Accelerate model functions and model loss functions for custom training loops by caching and reusing traces.
Accelerate Custom Training Loop Functions
This example shows how to accelerate deep learning custom training loop and prediction functions.
Check Accelerated Deep Learning Function Outputs
This example shows how to check that the outputs of accelerated functions match the outputs of the underlying function.
Evaluate Performance of Accelerated Deep Learning Function
This example shows how to evaluate the performance gains of using an accelerated function.