Hey guys, hope you are enjoying the new Qlik Community and the many relationships it fosters as well as the numerous resources it offers. Today I want to introduce and demonstrate our new machine learning capabilities added to the Qlik Sense November 2018 release. At Qlik we have been talking about augmented intelligence for a while and we have delivered some great innovations, such as our new cognitive engine and the Insight Advisor; both released earlier this year. Now we have taken things to the next level and are yet again leading BI innovation with the release of precedent based machine learning.
With any AI functionality, it requires to be constantly fed more and more information in order to get smarter and smarter. You could think of it as an inquisitive child. With the Qlik Sense November 2018 release, we've released features under the umbrella of what is called precedent-based machine learning. Precedent-based learning allows our cognitive engine to constantly learn from the data it sees, so it becomes a learning model. As it sees more data, it gets smarter and uses that to its advantage to give you better predictions and at the same time solicits user feedback so it can produce significantly smarter insights.
Watch this brief video to learn more and see these new capabilities in action.
Enjoy!
Can't see the video? YouTube blocked by your organization or region? Download the attached .mp4 to watch on your computer or mobile device.
It's these constant game changing improvements that Qlik Sense constantly brings us that inspire and energise me as a developer. I think Qlik need to do more to push their selling partners to push the end users on to Sense rather than Qlik View. I'm still seeing too many companies in the UK on Qlik View and missing out on a much richer user experience.
#1) - It gets stored in the learning database which is part of the enterprise repository and currently is not exposed to users or admins.
#2) - Yes - If you train the model in an app that is unpublished, then all learning will carry over to all users of the app, when this app is published. However, if you train the app AFTER it is published, the training remains in the context of the user who is training the app. This feature allows developers to train the app for all users, but at the same time avoids situations where business users can mess up results for all users by training the engine incorrectly or giving it noise.
Is there any private information sent to the cloud? Who is taking part of the learning exercise (app only, stream only, server only or the world?) Dion
Thanks Mike, but can't I train the model in my app and then re-publish to all of my users?
If it's following your scenario in #2, then wouldn't that mean that my existing, published apps can't leverage PBML and only my new apps can post November 2018 release?
Hi Dverbeke - in regards to private information I am not quite sure what is being asked here.
Ian - if you are training the app in your workspace before published to as stream yes, if you are training using your own copy of an app in your context (creating sheets from an already approved app and sheets) no. Yes correct, you would need to publish new copies of that app and overwrite them in the stream.
