Azure PCA-Based Anomaly Detection

Question

I'm trying to create an anomaly detection model to be able to detect odd new entries in my dataset.

I'm using Azure ML Studio and using the PCA-Based Anomaly Detection component. One of the options of this component is called "Training mode" and reading the documentation I should be able to set it to either "SingleParameter" or "ParameterRange". But when I add the component and try to set it up I only see the option "SingleParameter" available.

Did someone know why and if I can unlock it ? I would like to plug the PCA-Based Anomaly Detection component to a sweep parameters component but I think I can't do that is the Training mode is not set up to "ParameterRange".

Answer 1

As per the documentation it seems like the Azure ML Studio only support the SingleParameter training mode. So, you can't use the PCA-Based Anomaly Detection component with the sweep parameters.

However, one possible workaround to this can be with Tune Model Hyperparameters component. With this component you perform a parameter sweep to tune hyper-parameters as this component supports a parameter range and can be used with the PCA-Based Anomaly Detection component.