Multilayer perceptron (MLP) regression is a neural network-based regression algorithm that you can use to resolve nonlinear regression issues. MLP regression maps input features to outputs by using multiple hidden layers and can capture complex patterns and relationships. The training process of MLP regression includes forward propagation, loss calculation, backward propagation, and parameter update. The training process facilitates model learning and optimization and help accurately predict outputs.
Supported computing resources
Inputs and outputs
Input ports
You can use the Read File Data component to read training data files from Object Storage Service (OSS) paths.
You can configure the Train Data Oss path parameter of the MLP Regression (Training) component to select training data files.
Output port
You can save trained models to the path specified by the Output Model Oss Dir parameter of the MLP Regression (Training) component.
The following table describes the parameters of the component.
Configure the component
On the details page of a pipeline in Machine Learning Designer, add the MLP Regression (Training) component to the pipeline and configure the parameters described in the following table.
Tab | Parameter | Required | Default value | Description | |
Field Settings | Train Data Oss path | No | None | If no upstream OSS data is shipped, you must select a training data file. Example: train_data.csv. The .csv file that you select must be a numerical feature file and cannot contain headers. The last column in the file stores values used for training, and other columns store features. | |
Output Model Oss Dir | Yes | None | The OSS path that you can use to save a trained model. | ||
Pretrained Model Oss Path | No | None | The path of the pre-trained model. If you leave this parameter empty, no pre-trained model is loaded. | ||
Parameter Settings | MLP Layer Num | Yes | 3 | The number of MLP layers, except the output layer. | |
List of Hidden Layer Size | Yes | 64,32,16 | The number of output channels at each hidden layer. Separate multiple numbers with commas (,). If you enter a single number, the number is used as the number of output channels for all hidden layers. | ||
List of Dropout Ratio | Yes | 0.5 | The dropout rate of each dropout layer. Separate multiple dropout rates with commas (,). If you enter a single dropout rate, the dropout rate is used as the dropout rate for all dropout layers. | ||
Training Epoch | Yes | 100 | The total number of training epochs. | ||
Learning Rate | Yes | 0.01 | The learning rate. | ||
Training Batchsize | Yes | 32 | The number of training samples used in each iteration. | ||
Model Save Epoch | Yes | 10 | The model is saved after every specified number of training epochs. | ||
Validation Epoch | Yes | 5 | The validation set is evaluated after every specified number of training epochs. | ||
Optimizer Type | Yes | Adam | The optimizer, which is an algorithm used to update model parameters, such as the weight and offset. Valid values: Adam and SGD. | ||
Loss Type | Yes | MSE | The loss function that you can use to measure the difference between the predicted value and the actual value of a model. Valid values: MSE and L1. | ||
Execution Tuning | Select Resource Group | Public Resource Group | No | None | The instance type (CPU or GPU) and virtual private cloud (VPC) that you want to use. |
Dedicated resource group | No | None | The number of CPU cores, memory, shared memory, and number of GPUs that you want to use. | ||
Maximum Running Duration (seconds) | No | None | The maximum period of time for which the component can run. If the specified period of time is exceeded, the job is terminated. | ||