Can I use multiple contexts with Azure Personalizer?

Question

Can I use multiple contexts with Azure Personalizer?

Rman-1501 0

my question is I want to know how I can use Azure Personalizer in two or more different contexts. For example one context would be an article a user would be interested in. Another context would be where is the best place to show ads on a website. I mean two different problems with their different set of actions, action features and context features. Can I use the same azure instance or i have to create separate instances? Wouldn’t it create any conflicts in the models and decrease its accuracy in each context? how can I see the models it created and their performance evaluations? Thanks

Rohit Mungi 49,126 Reputation points Microsoft Employee Moderator

2023-04-17T11:28:17.5433333+00:00

@Rman-1501 Ideally, you should be using a single resource or a learning loop which is created for each subject domain or content area you have. Do not use multiple content areas in the same loop because this will confuse the learning loop and provide poor predictions. If you want Personalizer to select the best content for more than one content area of a web page, use a different learning loop for each.

For the second part of the question, I think you can enable Inference explainability to understand which features of a chosen action are the most and least influential to then model during inference. I also think the resource level metrics available for a personalizer resource should help you in evaluating the resource performance. Thanks!!
Rman-1501 0 Reputation points

2023-04-18T13:34:49.53+00:00

Thank you. I see your recommendation of using one instance per usecase. however that raises a question to me that what’s the point of sending actions list every time on each Rank call? If it shouldn’t change?