sxjscience opened a new issue #9830: Macro F1 score depends on the size of the minibatch URL: https://github.com/apache/incubator-mxnet/issues/9830 I find our current implementation of the macro F1 score is dependent to the size of the minibatch. Our strategy is to average the F1 scores of each minibatch. However, the result would be different if we choose different batch size and order of the samples (https://github.com/apache/incubator-mxnet/blob/master/python/mxnet/metric.py#L623-L626). This will certainly confuse the users. I suggest marking the macro F1 as "Not Recommended" and set micro F1 as the default.
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services