The index

Machine learning

5 entries matching the current filters. Sorted by recently added.

Depth separation for functions learnable by gradient descent

Depth-3 networks can express functions depth-2 cannot. Whether training can find them is a different question, and it is open.

Community consensus· 2016

OP-0048Deep learning theoryOpen

Derive neural scaling-law exponents from properties of the data

Loss follows a power law in compute, data and parameters. No theory predicts the exponent from the data distribution.

arXiv preprints· 2020

OP-0047Deep learning theoryOpen

Characterise the implicit bias of gradient descent for non-linear networks

For linear models the answer is a clean minimum-norm statement. For anything deeper, there is no general theory.

Community consensus· 2017

OP-11931OptimizationStatus unassessed

Parallel simple solvers with near-optimal optimization depth

Parallel simple solvers with near-optimal optimization depth. Current research status is unassessed. The dated source poses this question; inclusion does not certify that it remains open today.

SIAM Optimization — Open Problems in Optimization· Source question · editorial summary· 2025

OP-11930OptimizationStatus unassessed

Simple subgradient methods between two oracle-complexity regimes

Simple subgradient methods between two oracle-complexity regimes. Current research status is unassessed. The dated source poses this question; inclusion does not certify that it remains open today.

SIAM Optimization — Open Problems in Optimization· Source question · editorial summary· 2025