Confusing about the calculation of accuracy #15
Description
The author's approach to calculating accuracy seems to differ from what I believe. I think the classification should be based on the highest probability prediction, but the author seems to consider a prediction as correct if any of the top k predictions correspond to the target label. This approach may seem inappropriate.