MNIST
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
MNIST uniform label noise η=0.4 (7-fold CV)
96.62Accuracy (7-fold CV)
3
MNIST uniform label noise η=0.3 (7-fold CV)
97.02Accuracy (7-fold CV)
3
MNIST (test)
99.9Test Accuracy
3
MNIST
293.45Equivariance Error (Scaling + Rotation)
3
MNIST (test)
22Delay Gain
3
MNIST (test)
97.62Mean Accuracy
3
MNIST (test)
99.24NAT (Natural Accuracy)
3
MNIST
63.79Sinkhorn Divergence
3
MNIST (test)
5.16Local Training Time (h)
3
MNIST (test)
99.66Accuracy (ACCL)
3
MNIST to SVHN vs KMNIST
0.86AUROC
3
MNIST to SVHN vs FMNIST
0.955AUROC
3
MNIST to SVHN
4.102FID (50K)
3
Digit-MNIST watermarked (random colors) (test)
90.56Accuracy
3
MNIST (first 10 images) (test)
129.1Magnitude Cx
3
Augmented MNIST 1.0 (Retain Set)
32.92FID
3
Augmented MNIST 1.0 (Forget Set)
145.03FID
3
MNIST-Modified (test)
94.52Accuracy
3
MNIST (test)
93.8Accuracy
3
MNIST (test)
93.8Test Accuracy
3
MNIST
501.82Estimated Clusters
3
MNIST
12,000Avg Samples per Episode
3
MNIST (test)
306.9Cross-Entropy Loss
3
MNIST0
63.75F1 Macro
3
MNIST0
65.78F1-macro
3