search

xingjunm / dimensionality-driven-learning

Code for paper "Dimensionality-Driven Learning with Noisy Labels" - ICML 2018
58 stars 15 forks source link

LID increase with acc #4

Open lxbin1995 opened 6 years ago

lxbin1995 commented 6 years ago

Thank you for sharing the codes and your paper is very inspiring. But I test 'python train_newmodels.py -d cifar-10 -m d2l -e 50 -b 128 -r 0 ',and find the results is not like that in your paper: the LID increase while the Test acc increase ,which is different from the statement in your paper cifar-10/ 50 epochs /20%noise: 758344f310a6947a273af41e8e8c0bf

cifar-10/ 50 epochs /0%noise: image

cifar-10/150epochs/0%noise: 84ef2c7b08045f2400966b50de84316 the results in your paper: 3431a86e1a1a8fb9136581c261500ea

xingjunm commented 6 years ago

Hi, Issues on CIFRA-10 has been fixed.