Chen Bo

It seems that this may be the result of an overfit found in the search stage, resulting a little degradation in the re-train stage.

Thank you, you can refer to the FM model's code and add an extra MLP tower to get deepFM model.

Thank you. Each sample of the frappe dataset contains 10 features, including 8 features of the user field (before ',') and 2 features of the item field. All the features...

2-1011是user-feature (user ID and the last music ID)，12-16063是item-feature (music ID and artist ID)，数字只是one-hot表达

对的，last music ID只是最后一条记录，如果有多条就可以先pooling

使用的tensorflow版本为'1.13.1'，这个代码应该对版本没有太多要求的

报错是因为没有pretrain文件，可以先用FM预训练。MovieLens数据集可以根据文中所提方法处理