Training accuracy converges very fast

raghakot / keras-resnet

Residual networks implementation using Keras-1.0 functional API

Other

1.39k stars 617 forks source link

Training accuracy converges very fast #29

Closed pszyu closed 7 years ago

pszyu commented 7 years ago

Hello everyone, I have run the cifar10.py (set batch_size=128 as in the paper, and set data_augmentation=false, using sgd). My result shows that the training accuracy converges to 1 after 30 epochs, but the testing accuracy converges to 0.6. I do not understand why training accuracy converges so fast (much faster than the results in paper). Does anyone has an idea of the reason?

Thank you!

raghakot commented 7 years ago

You are overfitting.

Try a lower learning rate.
Use l2 regularization.

raghakot commented 7 years ago

We made some changes to cifar10 script. Should work out of the box. Reopen if you are still seeing the issue.