A jamming transition from under- to over-parametrization affects loss landscape and generalization

October 22, 2018 ยท Declared Dead ยท ๐Ÿ› Journal of Physics A: Mathematical and Theoretical

๐Ÿ‘ป CAUSE OF DEATH: Ghosted
No code link whatsoever

"No code URL or promise found in abstract"

Evidence collected by the PWNC Scanner

Authors Stefano Spigler, Mario Geiger, Stรฉphane d'Ascoli, Levent Sagun, Giulio Biroli, Matthieu Wyart arXiv ID 1810.09665 Category cs.LG: Machine Learning Cross-listed cond-mat.dis-nn, stat.ML Citations 160 Venue Journal of Physics A: Mathematical and Theoretical Last Checked 4 months ago
Abstract
We argue that in fully-connected networks a phase transition delimits the over- and under-parametrized regimes where fitting can or cannot be achieved. Under some general conditions, we show that this transition is sharp for the hinge loss. In the whole over-parametrized regime, poor minima of the loss are not encountered during training since the number of constraints to satisfy is too small to hamper minimization. Our findings support a link between this transition and the generalization properties of the network: as we increase the number of parameters of a given model, starting from an under-parametrized network, we observe that the generalization error displays three phases: (i) initial decay, (ii) increase until the transition point --- where it displays a cusp --- and (iii) slow decay toward a constant for the rest of the over-parametrized regime. Thereby we identify the region where the classical phenomenon of over-fitting takes place, and the region where the model keeps improving, in line with previous empirical observations for modern neural networks.
Community shame:
Not yet rated
Community Contributions

Found the code? Know the venue? Think something is wrong? Let us know!

๐Ÿ“œ Similar Papers

In the same crypt โ€” Machine Learning

Died the same way โ€” ๐Ÿ‘ป Ghosted