List your company
ResearchFormerly WizWand

Image Classification on Skewed MNIST

95.89Macro AccuracyBounded Adaptive
85.469288.174690.8893.5854Jun 2, 2025
Updated 4mo ago

Evaluation results

MethodLinks
Bounded Adaptiveε=4.0, Architecture=two-layer CNN2025.0695.8978.89
Bounded Adaptiveε=2.0, Architecture=two-layer CNN2025.0695.7578.18
Fix-Boundedε=4.0, Architecture=two-layer CNN2025.0695.6577.05
AUTOε=2.0, Architecture=two-layer CNN2025.0695.2374.23
Bounded Adaptiveε=4.0, Backbone=ResNet-182025.0695.0771.57
Fix-Boundedε=2.0, Architecture=two-layer CNN2025.0695.0775.28
AUTOε=4.0, Architecture=two-layer CNN2025.0695.0573.04
Constant Clippingε=4.0, Architecture=two-layer CNN2025.0694.9872.37
Bounded Adaptiveε=1.0, Architecture=two-layer CNN2025.0694.1970.07
Unbounded Adaptiveε=2.0, Architecture=two-layer CNN2025.0694.1971.48
Bounded Adaptiveε=2.0, Backbone=ResNet-182025.0694.1168.87
Fix-Boundedε=1.0, Architecture=two-layer CNN2025.0694.1168.3
Constant Clippingε=2.0, Architecture=two-layer CNN2025.0694.1167.65
Constant Clippingε=4.0, Backbone=ResNet-182025.0693.4659.73
Fix-Bounded Adaptiveε=4.0, Backbone=ResNet-182025.0693.1764.41
Unbounded Adaptiveε=4.0, Architecture=two-layer CNN2025.0693.1664.29
Unbounded Adaptiveε=1.0, Architecture=two-layer CNN2025.0692.8864.61
AUTOε=1.0, Architecture=two-layer CNN2025.0692.5365.66
Fix-Bounded Adaptiveε=2.0, Backbone=ResNet-182025.0692.2363.36
Unbounded Adaptiveε=4.0, Backbone=ResNet-182025.0691.9945.52
Constant Clippingε=1.0, Architecture=two-layer CNN2025.0691.5262.55
Unbounded Adaptiveε=2.0, Backbone=ResNet-182025.0691.4348.69
Constant Clippingε=2.0, Backbone=ResNet-182025.0691.0446.61
AUTOε=4.0, Backbone=ResNet-182025.0689.827.66
AUTOε=2.0, Backbone=ResNet-182025.0688.1219.96
Bounded Adaptiveε=1.0, Backbone=ResNet-182025.0687.3211.13
AUTOε=1.0, Backbone=ResNet-182025.0686.839.49
Constant Clippingε=1.0, Backbone=ResNet-182025.0686.633.53
Unbounded Adaptiveε=1.0, Backbone=ResNet-182025.0686.124.83
Fix-Bounded Adaptiveε=1.0, Backbone=ResNet-182025.0685.8710.62