An ML model trained on SageMaker is overfitting, and the training set includes irrelevant features. To reduce overfitting and lessen the influence of unnecessary features, which action should the engineer take?

L1 regularization (Lasso) adds a penalty equal to the absolute value of the magnitude of the coefficients to the loss function. This encourages sparsity in the model, effectively driving the weights of less important features to zero. By doing so, L1 regularization performs automatic feature selection, reducing the influence of irrelevant features and mitigating overfitting. Increasing or decreasing training iterations might affect convergence or underfitting/overfitting, but doesn't directly address irrelevant features. SageMaker Debugger is for monitoring and debugging training jobs, not for applying regularization techniques during runtime. Regularization parameters are set before training.