Closed TheaperDeng closed 3 months ago
Thanks. Could you also help test arnoldi using the noisy label script?
Thanks. Could you also help test arnoldi using the noisy label script?
I have done that for MNIST+MLP, it performs well.
[(0, 0), (100, 61), (200, 79), (300, 88), (400, 89), (500, 91), (600, 91), (700, 91), (800, 92), (900, 92)]
Checked Data Sample Found flipped Sample
--------------------------------------------------
0 0
100 61
200 79
300 88
400 89
500 91
600 91
700 91
800 92
900 92
Description
1. Motivation and Context
Support "arnoldi" in IFAttributor.
2. Summary of the change
3. What tests have been added/updated for the change?