Continuous Data Summarization on MNIST, CIFAR-10, and Fashion-MNIST
-0.225LossRobust Continuous Greedy Algorithm
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Robust Continuous Greedy AlgorithmAttack Norm=l1, D=1, gamma=1, epsilon=12026.06 | -0.225 | 0.664 | 6.299 | |
| Robust Continuous Greedy AlgorithmAttack Norm=l2, D=1, gamma=1, epsilon=12026.06 | 0.811 | 0.401 | 6.641 | |
| Robust Continuous Greedy AlgorithmAttack Norm=l_infinity, D=1, gamma=1, epsilon=0.52026.06 | 1.588 | 0.649 | 6.747 | |
| Gradient-based Robust BaselineAttack Norm=l_infinity, D=1, gamma=1, epsilon=0.52026.06 | 29.58 | 37.01 | -50.95 | |
| Gradient-based Robust BaselineAttack Norm=l1, D=1, gamma=1, epsilon=12026.06 | 30.66 | 32.33 | -47.53 | |
| Gradient-based Robust BaselineAttack Norm=l2, D=1, gamma=1, epsilon=12026.06 | 31.53 | 29.19 | -45.4 |