Analyzing Deep Neural Networks Using Attribution

Speaker: Dr. Mukund Sundararajan
Principal Research Scientist at Google

Bio:
There once was an RS called MS,
The models he studies are a mess,
A director at Google.
Accurate and frugal,
Explanations are what he likes best.

Abstract:
Predicting cancer from XRays seemed great
Until we discovered the true reason.
The model, in its glory, did fixate
On radiologist markings – treason!

We found the issue with attribution:
By blaming pixels for the prediction (1,2,3,4,5,6).
A complement'ry way to attribute,
is to pay training data, a tribute (1).

If you are int'rested in FTC,
counterfactual theory, SGD
Or Shapley values and fine kernel tricks,
Please come attend, unless you have conflicts.

Should you build deep models down the road,
Use attributions. Takes ten lines of code!

Note: We also have additional sign-up slots to speak 1:1 with Dr. Mukund Sundararajan after the event. If your schedule doesn’t allow you to participate in the 11:30 a.m. event or you would like additional time with Dr. Mukund Sundararajan, you can sign up for a smaller meeting after the event here.