37e44c4b5321605735be9761f9b758fc-Supplemental-Conference.pdf

Supplementary Materials for "RecursiveMix: Mixed Learning with History" The proposed method is trained and will predict the results based on the statistics learned from the training datasets with potential biases that would include possible negative societal impacts. These issues warrant further research and consideration when using this technology. In this section, we review works that "Learning with History" in detail. Specifically, we use the superscript t (e.g., p The superscript t can be epoch-wise, iteration-wise, or customized period-wise. We then demonstrate a detailed review of existing representative approaches according to the abovementioned elements.