Platform For AI:Lasso Regression Training

Limits

Supported computing engines: MaxCompute, Flink, or DLC.

Algorithm principles

Lasso regression builds a refined model by creating a penalty function. This function shrinks some regression coefficients by forcing the sum of their absolute values to be less than a fixed value, and sets other coefficients to zero. This method retains the benefits of subset shrinkage and provides biased estimation for handling multicollinear data.

Configure component parameters

• Input ports

  Input port	Data type	Recommended upstream component	Required
  Training data	None	Read Table Feature Engineering Data Preprocessing	Yes
  Base model	Lasso model	Read Table Lasso Regression Training	No (for incremental training)

• Component parameters

  Tab	Parameter	Description
  Field Settings	Target column name	Name of the target column in the input table.
  	Feature column array	Cannot be configured if Vector column name is specified. Names of feature columns used for training. Note Feature column array and Vector column name are mutually exclusive. Use only one to specify input features for the algorithm.
  	Vector column name	Cannot be configured if Feature column array is specified. Name of the vector column. Note Feature column array and Vector column name are mutually exclusive. Use only one to specify input features for the algorithm.
  	Weight column name	Name of the weight column.
  Parameter Settings	Penalty factor: lambda	Coefficient of the regularization term. Data type: DOUBLE.
  	Convergence threshold	Threshold to determine whether the iterative method has converged. Default value: 1.0E-6.
  	Learning rate	Controls the speed at which parameters are updated during model training. Default value: 0.1.
  	Maximum number of iterations	Maximum number of iterations. Default value: 100.
  	Optimization method	Optimization method used to solve the problem. Valid values: LBFGS GD Newton SGD OWLQN
  Execution Tuning	Number of workers	Used with Memory per worker. Must be a positive integer from 1 to 9999.
  	Memory per worker (MB)	Value ranges from 1024 MB to 64 × 1024 MB.

• Output ports

  Output port	Data type	Downstream component
  Trained model	Regression model	Lasso Regression Prediction
  Model information	None	None
  Feature importance	None	None
  Linear model weight coefficients	None	None

Configure the component by using code

from pyalink.alink import *

def main(sources, sinks, parameter):
    batchData = sources[0]
    ridge = LassoRegTrainBatchOp()\
        .setLambda(0.1)\
        .setFeatureCols(["f0","f1"])\
        .setLabelCol("label")
    model = batchData.link(ridge)
    model.link(sinks[0])
    BatchOperator.execute()