AnalysisAI ModelsAugust 12, 2026

Adam breaks rotational invariance of factored-model loss

In factored models W = UV^T, the loss is invariant to rotations (U,V)→(UQ,VQ); gradient descent respects that, but Adam's per-coordinate second moment depends on the basis and does not.

1 source

Daily brief

Get tomorrow's AI brief in your inbox

More stories today

Open the live feed
Adam breaks rotational invariance of factored-model loss — AIBriefs