s4753993-GCN #477
s4753993-GCN #477
Conversation
The previous datatypes for adjacency matrix, and especially features were creating problems during testing phase.
Had trouble returning all four(validation_accuracies,validation_losses,train_accuracies,train_losses) together from one function, so decided to save in np arrays and load the arrays during driver code.
|
Looks good. TSNE also fine. |
Good Practice (Design/Commenting, TF/Torch Usage)Adequate use and implementation Recognition ProblemSolves problem Commit LogMeaningful commit messages, could be more informative -1 DocumentationReadMe acceptable, background/architecture missing -2 Pull RequestSuccessful Pull Request (Working Algorithm Delivered on Time in Wrong Branch) -1 |
|
Hey Shakes, I believe I have fixed the branch issue, also just for knowledge could I know what you mean by README missing background/architecture? Does it mean architecture of the GCN model itself? |
Hello Shakes,
This PR, to the best of my knowledge contains files required to run a 2 layer GCN along with TSNE mapping on the facebook.npz dataset. All the work was done on google colab, and I tried my best to create individual modules for the various functions including functions to load data, the model itself as well as training and testing function.
However I was not able to successfully implement a driver function in python, hence I have submitted the jupyter file containing all the functions together.
To successfully run the code please open GCN_final.ipynb, and run all commands in sequential order. Please just ensure file_path to load data matches local environment.
Thank you
Arsh Upadhyaya
(s4753993)