CODE: GRAPH Link Prediction w/ DGL on Pytorch and PyG Code Example | GraphML | GNN

[@deveshlangade5938]

Thanks man been finding this for so long

[@code4AI]

No problem 👍

[@WeeeAffandi]

What would be the steps for performing just the inference ? Do we need to calculate pos_g and neg_g in that case ? All the examples I have seen use some splitting of train / test / valid from the dataset whereas I just want to perform inference on the entire dataset

[@aboudramanediarra7086]

Hello, thank you very much for this beautiful presentation. I work with image data. I'd like to represent them in the form of a graph and predict the links between the pixels. Can you help me with some ideas or a piece of source code? Thanks in advance.

[@VietNguyen-vi3fu]

Hi, would it be possible to use autoencoder-based models for link prediction on directed, heterogenous multi-graphs (multiple edge types, multiple edges between 2 nodes)? I'm stuck at how to incorporate the final prediction task for the problem.

[@harkishanpatil5723]

Hey hello and first of all thanks for the content. Your content is like blessing for me. Because I am in bit of a situation here . I come from a non coding background and I am doing my final year project related to this topic and it will be very helpful if you provide the colab notebook for DGL of the same implementation. Thank you.

[@code4AI]

Glad you find it helpful. For a detailed step-by-step explanation I recommend the official documentation here: docs.dgl.ai/en/0.9.x/tutorials/blitz/4_link_predict.html#sphx-glr-tutorials-blitz-4-link-predict-py and for the Jupyter NB, maybe you try this one: docs.dgl.ai/en/0.9.x/_downloads/3d18f3c819071626ba37897638f36dde/4_link_predict.ipynb

[@code4AI]

Another resource is, of course, as always, GitHub: Theory and code for link prediction: github.com/dmlc/dgl/blob/0.9.x/docs/source/guide/training-link.rst

[@Orneyfish]

Hey man can you share the colab notebook as well?

[@miltonminervino8460]

I cannot understand in the DGL framework how one can do inference on unseen nodes for link prediction, do you have some code snippet showing this?

[@code4AI]

If you cross out the word "unseen" in your q, could you imagine a solution?

[@alishaheen8443]

Hello and thank you for your content. I am working on a project where I am supposed to detect 2d objects in an image and convert it to 3d model using GNN do you have any videos or references that can help?

[@code4AI]

To add dimensions to objects: currently no videos available. But think about it, you need a vision memory of all 3d objects, then you should figure out the perspective of the photo taken to get the perspective factors right, then you need to get the distance with normed (1m high) length objects you can find in the image, and from that info you can deduct the average size of objects and compare this info of overall size (to exclude a lot of 3d objects) with 3d shape projections to your photo. Sounds fun to implement!

[@harkishanpatil5723]

And also I am not getting to how to visualize the results and predict results between two nodes.

[@nadaal-jarrah8670]

Hello, How can I use a different dataset, I use colab, and the dataset I read from the drive, the dataset contains 9 folder label and each label contain a .txt document, I don't have a problem reading the dataset but how can apply your code with my dataset. Thank you

[@code4AI]

Your description is too general, but it would be a good idea to build one data object, which you could address, eg in your input pipeline.

[@TheMishupo]

Can you please add links of the code and documents?

[@code4AI]

I was inspired by those code snippets here (the official DGL documentation): docs.dgl.ai/en/0.9.x/guide/training-link.html Always a good idea to go to the original source ...

[@MrAnsu007]

no graphical illustration is just so frustrating

[@code4AI]

Absolutely agree. What would you pay for it?