Different questions.
One pursuit: physical trust.
From symmetry-preserving models to grounded scientific reasoning, each part of my research examines a different reason a prediction can fail.
Physically
trustworthy
machine learning
Enforce · Measure · Verify
Symmetry
Is the prediction physically allowed?
Scale
Does it hold as the system grows?
Conditions
Does it hold when conditions change?
Verification
Can we detect a wrong claim?