Giter VIP home page Giter VIP logo

graphhinge's Introduction

Hi~ Thank you for coming!

  • 🔭 I am a Ph.D. majoring in Computer Science at Shanghai Jiao Tong University starting from Sep. 2019 to Jun. 2024.
  • 🌱 I was a visiting scholar at AI Center of University College London from Aug. 2022 to Jul. 2023.
  • 💬 Happy to help you if I could.
  • 📫 Please reach me via [email protected].
  • ⚡ I am a literature and photography lover :P.

graphhinge's People

Contributors

jinjiarui avatar kounianhuadu avatar

Stargazers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

Watchers

 avatar  avatar  avatar  avatar  avatar  avatar

graphhinge's Issues

Got an unexpected result

I downloaded the ml-100k(MovieLens 100K Dataset) and did the experiments with default args ( 3 epochs), however, the results I got are much higher than those reported in the paper.

Test AUC 0.9506 | ACC 0.8792 | Test F1 0.8764

I am wondering whether it is caused by my data processing scripts or other reason.

About baseline methods

您好,请问能否提供您选用的一些baseline方法的代码呢,因为我发现许多的源码具有特异性,而我刚入门这个领域,对于如何改写这些模型感到困惑
(例如LGRec和MCRec它们的initial embedding的详细生成方法并未提供,而且LGRec的模型代码似乎跑不通;HetGNN模型是hard coded for academic paper,怎么改用到其他数据呢?HAN模型是embedding方法,如果要改用到推荐,该如何实现呢?)
希望能得到您的回复

Confusion about metapath

Hello, it is a pleasure to read your article, I did not quite understand the metapath part of the provided code, about this part of the code
def forward(self, UI, IU, UIUI, IUIU, UIAI1, IAIU1, UIAI2, IAIU2, UIAI3, IAIU3):
user_idx = UI[:,0,0] #B
item_idx = IU[:,0,0] #B
source_feature = self.user_emb(user_idx) #BE
target_feature = self.item_emb(item_idx) #B
E

#UI & IU BLN (batch_size, num_paths, num_nodes)
ui = torch.stack((self.user_emb(UI[:,:,0]), self.item_emb(UI[:,:,1])),3) #BLEN
iu = torch.stack((self.item_emb(IU[:,:,0]), self.user_emb(IU[:,:,1])),3) #B
LEN
Can the metapath set here be changed according to your needs? As a beginner of HIN, I hope to get your answer, thank you very much

sampling path but without omitting existing edge

Imho, omitting 20% of existing edges should be done before constructing the graph when doing 5 fold cross-validation.
Otherwise, meta-path sampling will generate a lot of paths involved with positive instances in the test&eval dataset, which is an information leak and should be unseen.

请教一下有关movielens数据集预处理的相关问题

感谢您在HIN领域作出的贡献!
我在运行您的代码dataloader.py 的 _load_movielens(self)时,想知道您运行时导入的
_data_list = ['movie_genre.dat', 'user_occupation.dat', 'user_age.dat', 'user_movie.dat']
四个文件是如何通过原始数据预处理得到的
对于其中的movie_genre.dat
_movie, _genre = int(_line[0]), int(_line[1])
是如何将类型变为数字的呢?
如果您能为我解答或者将您处理后并且实际测试的movielens数据集提供给我,我将感激不尽,谢谢!

About Potential Data Leakage

I noticed that you used all “user-item” edges to construct the heterogeneous graph. And based on it, you then predict the user-item edges (i.e. recommendation). Would this cause data leakage?

experiments results in KDD paper and "GraphHINGE: Learning Interaction Models of Structured Neighborhood on Heterogeneous Information Network"

您好,感谢您在HIN领域出色的工作!
我了解了您在KDD 2020的paper以及后面这篇paper中与KDDpaper中方法相似的部分,以GraphHINGE_FFT.py中的方法为例,kDD论文中movielens的acc为0.7896,后一篇为0.8837,我看主要方法部分没有很明显的区别(仅指在这个py文件中所涉及的网络结构),而且baseline的相同方法也有一定程度的提升,请问这种提升是因为什么呢?是因为hyperparameter的设置,数据的预处理方式,还是因为一些两篇论文细节方面我没有充分进行理解,在细节上是有区别的(仅指在KDD论文中提到的方法)?
我是这个方向的初学者,如果您能对我的疑惑进行解答,我将非常感谢!

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.