神经网络

用于回归的神经网络

神经网络模型由一系列反映大脑处理信息方式的层构成。Statistics and Machine Learning Toolbox™ 中提供的回归神经网络模型是完全连接的前馈神经网络，您可以调整全连接层的大小并更改层的激活函数。

要训练回归神经网络模型，可以使用回归学习器。为了获得更大的灵活性，请在命令行界面中使用 fitrnet 来训练回归神经网络模型。经过训练后，您可以通过将模型和新预测数据传递给 predict 来预测变量新数据的响应。

如果您要创建更复杂的深度学习网络并拥有 Deep Learning Toolbox™，您可以尝试深度网络设计器 (Deep Learning Toolbox)。

App

使用有监督机器学习训练回归模型来预测数据

Predict responses using neural network regression model (自 R2021b 起)

`fitrnet`	Train neural network regression model (自 R2021a 起)
`compact`	Reduce size of machine learning model
`fitrqnet`	Train regression quantile neural network (自 R2024b 起)

`lime`	Local interpretable model-agnostic explanations (LIME)
`partialDependence`	Compute partial dependence
`permutationImportance`	Predictor importance by permutation (自 R2024a 起)
`plotPartialDependence`	Create partial dependence plot (PDP) and individual conditional expectation (ICE) plots
`shapley`	Shapley values (自 R2021a 起)

`crossval`	Cross-validate machine learning model
`kfoldLoss`	Loss for cross-validated partitioned regression model
`kfoldPredict`	Predict responses for observations in cross-validated regression model
`kfoldfun`	Cross-validate function for regression
`kfoldLoss`	Loss for cross-validated partitioned quantile regression model (自 R2025a 起)
`kfoldPredict`	Predict responses for observations in cross-validated quantile regression model (自 R2025a 起)
`kfoldfun`	Cross-validate function for quantile regression (自 R2025a 起)

`loss`	Loss for regression neural network (自 R2021a 起)
`resubLoss`	Resubstitution regression loss
`loss`	Loss for quantile neural network regression model (自 R2024b 起)

`predict`	Predict responses using regression neural network (自 R2021a 起)
`resubPredict`	Predict responses for training data using trained regression model
`predict`	Predict response for quantile neural network regression model (自 R2024b 起)

`RegressionNeuralNetwork`	Neural network model for regression (自 R2021a 起)
`CompactRegressionNeuralNetwork`	Compact neural network model for regression (自 R2021a 起)
`RegressionPartitionedNeuralNetwork`	Cross-validated regression neural network model (自 R2023b 起)
`RegressionQuantileNeuralNetwork`	Quantile neural network model for regression (自 R2024b 起)
`CompactRegressionQuantileNeuralNetwork`	Compact quantile neural network model for regression (自 R2025a 起)
`RegressionPartitionedQuantileModel`	Cross-validated quantile model for regression (自 R2025a 起)

Assess Regression Neural Network Performance
Use fitrnet to create a feedforward regression neural network model with fully connected layers, and assess the performance of the model on test data.
Working with Quantile Regression Models
Estimate prediction intervals and create models that are robust to outliers by using quantile regression models.
Create Prediction Intervals Using Split Conformal Prediction
Create a prediction interval and use conformalized quantile regression to calibrate the prediction interval.
Deploy Neural Network Regression Model to FPGA/ASIC Platform
Predict in Simulink^® using a neural network regression model, and deploy the Simulink model to an FPGA/ASIC platform by using HDL code generation.