search

tdeboissiere / DeepLearningImplementations

Implementation of recent Deep Learning papers
MIT License
1.81k stars 652 forks source link

Can not reproduce the result #60

Closed chaoyan1037 closed 6 years ago

chaoyan1037 commented 6 years ago

For dense net. Could you please give a detailed parameters description which generated the reported result? I used the default parameters except for those specified in your README file. Also, I changed the number of the epoch of 300. But only got at most 75% train accuracy. I did not change anything else. What is the matter?

chaoyan1037 commented 6 years ago

When using SGD as the optimizer, the learning rate should be 0.1, while Adam 0.001.