How Far is Adam from Natural Gradient Descent?

תקציר מקורי באנגליתarXiv:2610.00004v1 Announce Type: new Abstract: Adam is the standard optimizer in deep learning, yet its geometric relationship to natural gradient descent (NGD) contains unresolved questions. We study Adam's full update rule, including momentum, as a diagonal empirical Fisher approximation subject to diagonal truncation, empirical label substitution, and temporal lag. Using the scale-invariant $\gamma(\Delta\theta)$ metric, we measure Adam's geometric deviation from true NGD across four loss landscapes: well-conditioned linear regression, ill-conditioned linear regression, logistic regression, and a non-convex small neural network. Adam's geometric trajectory is context-dependent. Deviation remains low in well-conditioned settings but rises significantly under ill-conditioning, reaching m