Piecewise-affine networks realize fewer affine regions than their theoretical capacity when trained on higher-complexity tasks, termed expressivity saturation, alongside a new bound on line-segment probes.
Kuan-Lin Chen, Harinath Garudadri, and Bhaskar D
1 Pith paper cite this work, alongside 4 external citations. Polarity classification is still indexing.
1
Pith paper citing it
4
external citations · OpenAlex
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Expressivity Saturation: Reduced Affine Region Usage Under Increasing Task Complexity
Piecewise-affine networks realize fewer affine regions than their theoretical capacity when trained on higher-complexity tasks, termed expressivity saturation, alongside a new bound on line-segment probes.