Geometric Regularization from Overparameterization
Abstract
A geometric explanation is proposed for double descent, linking implicit regularization to how the volume of possible weight updates shrinks as intrinsic dimensionality changes near global minima.
The volume of the distribution of weight sets associated with a loss value may be the source of implicit regularization from overparameterization due to the phenomenon of contracting volume with increasing dimensions for geometric figures demonstrated by hyperspheres. We introduce the geometric regularization conjecture and extract to an explanation for the double descent phenomenon by considering a similar property resulting from shrinking intrinsic dimensionality of the distribution of potential weight set updates available along training path, where if that distribution retracts across a volume verses dimensionality curve peak when approaching the global minima we could expect geometric regularization to re-emerge. We illustrate how data fidelity representational complexity may influence model capacity double descent interpolation thresholds. The existence of epoch and model capacity double descent curves originating from different geometric forms may imply universality of closed n-manifolds having dimensionally adjusted n-sphere volumetric correspondence.
Community
As a parting gift to this platform the author offers a 2022 preprint in which was proposed a physics grounded mechanism for the double descent phenomenon that had been surveyed in the NeurIPS literature. He never found domain experts to share a conversation about the culmination of years of volunteer research. The paper was fully human authored without the use of language models. To the author's knowledge it has never been cited in any formal capacity.
Get this paper in your agent:
hf papers read 2202.09276 Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash Models citing this paper 0
No model linking this paper
Datasets citing this paper 0
No dataset linking this paper
Spaces citing this paper 0
No Space linking this paper
Collections including this paper 0
No Collection including this paper