Amazon SageMaker Experiments now supports common chart types to visualize model training results
SageMaker Experiments now supports granular metrics and graphs to help you better understand results from training jobs performed on SageMaker. Amazon SageMaker Experiments is a capability of Amazon SageMaker that lets you organize, track, compare and evaluate machine learning (ML) experiments. With this launch, you can now view precision and recall (PR) curves, receiver operating characteristics (ROC curve), and confusion matrix. You can use these curves to understand false positives/negatives, and tradeoffs between performance and accuracy for a model trained on SageMaker. You can also better compare multiple training runs and identify the best model for your use-case.
SageMaker Experiments now supports granular metrics and graphs to help you better understand results from training jobs performed on SageMaker. Amazon SageMaker Experiments is a capability of Amazon SageMaker that lets you organize, track, compare and evaluate machine learning (ML) experiments. With this launch, you can now view precision and recall (PR) curves, receiver operating characteristics (ROC curve), and confusion matrix. You can use these curves to understand false positives/negatives, and tradeoffs between performance and accuracy for a model trained on SageMaker. You can also better compare multiple training runs and identify the best model for your use-case.
To get started, use the python SDK to log metrics for your trials from your training script. To view charts for your trials, navigate to the chart tab in Experiments UI in SageMaker Studio. Starting today, this feature is available in all regions where SageMaker Experiments is available. To get started with SageMaker Experiments, see the document page or access SageMaker Experiments within SageMaker Studio.