2020-04-15还需进行的实验和思考的过程

要做的事情：

实验：

对比实验baseline

1) GP-BPR

2) learning binary recommendation for fashion recommendations

3) KNNLS

4) KGCN

5) RippleNet

6) VBPR

7) TBPR

8) VTBPR

8) T-V-E-BPR

Ablation Study

1. Visual, textual, entity 

2. attention vs. average

3. transform v.s. non-transform

4. bpr loss + vse loss + l2 loss + regularization loss

5. dimension of user_embedding, word, visual ,entity to the performance of the system

6. windows sizes of filters and the number of filters m.( referring to dkn and textcnn)

7. consider to add attention

8. Offline embedding methods such as transE, transD, transH.

Case Study

1.找到knowledge graph在推荐中发挥的作用，并给与图示

2. with knowledge graph or without 区别. 如何给予可解释性的说明。

图标的声明，不同number of xxx, 对最终性能的影响。