Closed danbraunai-apollo closed 11 months ago
Closes #206
TODO: Compare run with numerical integrated gradients (previous commit) to this run. They should be the same.
NOTE: POSTPONED AFTER STOCHASTIC SOURCES DISCOVERY. Analytic integrated gradients for a plain linear layer
Description
Related Issue
Closes #206
Motivation and Context
How Has This Been Tested?
TODO: Compare run with numerical integrated gradients (previous commit) to this run. They should be the same.
Does this PR introduce a breaking change?