管理试验
R2026b在多个初始条件下训练网络，以交互方式调整训练选项，并评估结果
使用试验管理器，通过扫描一系列超参数值或使用贝叶斯优化，找到神经网络的最佳训练选项。使用内置函数 trainnet 或定义您自己的自定义训练函数。使用训练图监控进度。使用混淆矩阵和自定义度量函数来评估经过训练的网络。
此页包含有关 AI 工作流试验的信息。有关使用该 App 的一般信息，请参阅试验管理器。
App
| 试验管理器 | Create and run experiments to train and compare deep learning networks |
对象
experiments.Monitor | Update results table and training plots for custom training experiments |
函数
groupSubPlot | Group metrics in experiment training plot |
recordMetrics | Record metric values in experiment results table and training plot |
updateInfo | Update information columns in experiment results table |
yscale | Set training plot y-axis scale (linear or logarithmic) (自 R2024a 起) |
主题
配置试验
- Choose Strategy for Exploring Experiment Parameters
Choose between the exhaustive sweep, random sampling, and Bayesian optimization strategies for exploring your experiment parameters using Experiment Manager. - Tune PyTorch Model Using Experiment Manager
Tune hyperparameters of a PyTorch® model by sweeping over combinations of hyperparameter values using Experiment Manager. (自 R2026b 起) - Debug Deep Learning Experiments
Diagnose problems in your setup, training, and metric functions. (自 R2023a 起) - Keyboard Shortcuts for Experiment Manager
Navigate Experiment Manager using only your keyboard.
使用 trainnet 的试验
- Compare Classification Network Architectures Using Experiment
Train a deep learning network for classification using Experiment Manager. - Compare Dropout Probabilities and Filter Configurations for Image Regression Using Experiment
Train a deep learning network for regression using Experiment Manager. - Evaluate Deep Learning Experiments by Using Metric Functions
Use metric functions to evaluate the results of an experiment.
使用自定义训练函数的试验
- Use Bayesian Optimization in Custom Training Experiments
Create custom training experiments that use Bayesian optimization. - Run a Custom Training Experiment for Image Comparison
Train a twin neural network to identify similar images of handwritten characters. - Custom Training with Multiple GPUs in Experiment Manager
Configure multiple parallel workers to collaborate on each trial of a custom training experiment.
迁移学习
- Try Multiple Pretrained Networks for Transfer Learning
Configure an experiment that replaces layers of different pretrained networks for transfer learning. - Experiment with Weight Initializers for Transfer Learning
Configure an experiment that initializes the weights of convolution and fully connected layers using different weight initializers. - Audio Transfer Learning Using Experiment Manager
Configure an experiment that compares the performance of multiple pretrained networks applied to a speech command recognition task using transfer learning.
缩短试验执行时间
- Run Experiments in Parallel
Run multiple simultaneous trials or one trial at a time on multiple workers. - Offload Experiments as Batch Jobs to a Cluster
Run experiments on a cluster so you can continue working or close MATLAB®. (自 R2022a 起)
精选示例
Tune Experiment Hyperparameters by Using Bayesian Optimization
Find optimal network hyperparameters and training options for convolutional neural networks.
Choose Training Configurations for LSTM Using Bayesian Optimization
Find optimal data architecture and network configurations for sequence-to-sequence regression using Bayesian optimization.
Use Experiment Manager to Train Generative Adversarial Networks (GANs)
Create a custom training experiment to generate images of flowers.
