huangtinglin / ngcf-pytorch Goto Github PK
View Code? Open in Web Editor NEWPyTorch Implementation for Neural Graph Collaborative Filtering
PyTorch Implementation for Neural Graph Collaborative Filtering
It should be python main.py --dataset amazon-book --regs [1e-5] --embed_size 64 --layer_size [64,64,64] --lr 0.0005 --save_flag 1 --pretrain 0 --batch_size 1024 --epoch 200 --verbose 50 --node_dropout [0.1] --mess_dropout [0.1,0.1,0.1]
not python NGCF.py .......
According to the original paper, the authors randomly selected 10% of interactions as validation set to tune hyper-parameters. However, it seems no validation set is constructed to select models and the best performance on the test set is finally reported in your implementation.
Looking forward to your reply.
你好, 我最近也在复现这篇论文. 但是我遇到了一个问题, 我所预处理的数据的统计和论文差别很大. 请问, 你有注意到吗
main.py里包含一些错误,无法运行
Hi,
Thanks for your great implementation in PyTorch!
My GPU runs out of memory when creating the adjacent matrix (create_adj_mat()). May I ask how large the memory of your GPU is when running this code successfully? Thank you very much!
原始数据集(如amazon book)每行内容是 [用户 物品 评分 时间戳]
train.txt中每行是 [用户id 物品1id 物品2id 物品3id ...]
请问:
1、源数据集中用户对物品的评分如何处理?是直接忽略掉,只要有有评分就意味着发生交互吗?
2、用户对物品的交互是按8:2随机划分到训练集和数据集吗?
3、train.txt 中 每行物品的顺序是什么依据?
谢谢!
你好,代码里的norm_adj_mat好像跟NGCF中的D^-0.5AD^-0.5不一致,但是作者的tensorflow代码好像也是这样的。
在上一个问题我看到了“用户对物品的交互是按8:2随机划分到训练集和数据集”,我在切分movielen-1m也是这样切分的,然后用您的NGCF跑的时候,发现跑出来的Precision@10能达到0.5几,但movielen-1m一般这个值再好也应该只在0.15左右吧?但你的计算公式也没错... 所以我很疑惑,不知道到底哪里出问题了...
I wonder is the self-connection part missing in your code?
运行示例之后 没找到预测结果
Hi, I am confused about how the python code runs without bugs since I found that there are no clear pre-defined variables in the main.py. For example, you only define args in the parser.py but not introduce it in main.py. But it seems no problem here. Can you give me some hints or explanations on this so that I can understand the codes more thoroughly. Thanks!
(you can reply in Chinese if you want)
I experimented with AMAZON-BOOK data, and the NDCG result is 0.06892(At K=20), which is much higher than the paper's NDCG value (0.0261).
Is there any difference between this code and the original code?
It is questionable why the NDCG value of this code is higher than the paper value.
According to your paper, there are two transformation matrices in NGCF model: W1 and W2. Adjacent matrix for W1 is added with identity matrix. For W2, however, self connection is not considered. It seems that both adjacent matrices for W1 and W2 are added with identity matrix: https://github.com/huangtinglin/NGCF-PyTorch/blob/master/NGCF/NGCF.py#L105. Looking forward to your reply : )
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.