vrakesh commented on issue #14917: Changing the learning rate affect the accuracy URL: https://github.com/apache/incubator-mxnet/issues/14917#issuecomment-491069897 Hi @CynthiaProtector turns out this is expected behavior, once learning rate is set to zero , the training essentially starts, running training on the next batch of data with learning rate set to zero basically means you are being shown test accuracy rather than training accuracy. I would suggest a non zero small amount of decay instead, to see if that helps. But setting learning rate to zero essentially means stopping gradient descent, no further improvements on weights are made on new batches of data If this answer , satisfies your query, Feel free to close this issue
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
