Explainable AI for Science and Medicine

Understanding why a machine learning model makes a certain prediction can be as crucial as the prediction's accuracy in many applications. Here l will present a unified approach to explain the output of any machine learning model. It connects game theory with local explanations, uniting many previous methods. I will then focus specifically on tree-based models, such as random forests and gradient boosted trees, where we have developed the first polynomial time algorithm to exactly compute classic attribution values from game theory. Based on these methods we have created a new set of tools for understanding both global model structure and individual model predictions. These methods were motivated by specific problems we faced in medical machine learning, and they significantly improve doctor decision support during anesthesia. However, these explainable machine learning methods are not specific to medicine, and are now used by researchers across many domains. The associated open source software (github.com/slundberg/shap) supports many modern machine learning frameworks and is very widely used in industry (including at Microsoft).
See more at www.microsoft.com/en-us/resea...

[@YuchengLin]

The audience asked challenging questions because they UNDERSTAND the content. Kudos!

[@behnamplays]

One of the best talks I've heard in 2020. Awesome!

[@AjaySharma-me1sy]

I am currently pursuing the Explainable AI course at UW and read Scott's paper as a class discussion. But I truly only understood it through this lecture, thanks for posting this!

[@griffinheart]

Great presentation, very clearly explained the concept, appreciate the great work!

[@jasonalbia5007]

Very interesting talk! Highly informed audience can really be tough sometimes. Great presentation! :)

[@andrewm4894]

found this talk really great, shared it with everyone!

[@bevansmith3210]

Thanks for sharing. Really informed audience

[@senwang1982]

great work and detailed presentation. thanks for sharing.

[@rdkap42]

My largest concern is the independence of features assumption, but this is a great talk

[@ProfessionalTycoons]

Thank you for this talk!

[@zhaobryan4441]

Great audience! I love the atmosphere there

[@PantelisNatsiavas]

Impressive lecture (and impressive audience too)

[@Muuip]

SHAP summaries should be integreated in all Machine Learning models. Computers can be programmed to learn and programmed to teach what and how they have learned with SHAP summaries ... diminishing inference and diminishing singularity.

[@linlinzhao9085]

excellent presentation, thanks

[@KnowNothingJohnSnow]

Wonderful presentation !!!! Thank u so mcuh

[@vibesnovibes6320]

Amazing discussion 👍

[@dr_flunks]

Now I must have this toy... thank you!!

[@sakuragi9570]

at 35:01 with a caption on, we got a valuable meme material. Thanks Scott! Good presentation btw

[@pinakibhattacharyya7853]

Great talk

[@anilb1076]

can any one say what are the tools/libraries used for xai?

[@lugas2267]

what a nice dude

[@Muuip]

Great presentation, much appreciated! 👍

[@yamacgulfidanalumni6286]

Let the man talk lol

[@sarangakumarapeli4348]

is there any method to evaluate XAI framworks results?

[@stanzhao3606]

Haha, Susan was there as well. I detected her voice ^+^.

[@alphavr1315]

The lady always asking is really annoying...

[@randomguy75]

Thank you

[@karthiksrinivasan4923]

She asks great questions actually!

[@ettoremariotti4280]

so annoying!!!!!! It really breaks the flow of the presentation

[@bholaprasad26]

This is so frustrating. He is saying for 1 min then all the people asking him questions for 10 min. Why the hell they are not letting him finish the presentation and ask questions later. It's good that you are smart but being annoying is not.

[@hoaxuan7074]

A Karen.

[@Wurfenkopf]

19:20 I'm a mathematician and, LOL!

[@Muuip]

Machine Learning models of Multi-omics data in combination with biology physiological and pathological mathematic and 3D models to ascertain causality in order to suggest intervention(s) on a continuous basis.

[@user-ux3wg1xj9s]

14:53 이어보기

[@EC-ve4dw]

Good talk! plse speak slowly and articulate better. not understandable sometimes.