45616756 - GCN on Facebook dataset

[wildanazz]

Hello, this is my submission for the report assignment. In this report, I use GCN for node classification on Facebook Large Page-Page Network. There are three main files: driver.py, model.py, and plot.py. Data preparation, model creation, and training is in the driver.py. GCN layer and model class are in model.py. The plots are generated from plot.py. Thank you.

Muhammad Wildan Aziz, 45616756

[gayanku]

• PR Nov 1

• Commits span 3 days ( 30,31,1)

• Non descriptive commit messages.

• Readme is well done with background, results, diagrams.

• multi layer (2) layer GCN, Good val accuracy (+85%) though test accuracy not seem to be provided, it can be expected to be similar. Semi supervised split.

• Valid for NORMAL ( subject to no TSNE plot or code for plotting submitted)

• Code does not plot a TSNE as needed. While you are saving TSNE embeddings ( which a 2 dimensional) there is no plot code to plot the node (x,y) on a visualizable graph. This is a requirement.

• Code structure is ok ( has model file, driver file), but can be improved. Move pre processing, train etc from main() as you have done when calling tsne() from main rather than have a huge monolithic main().

[shakes76]

TF/Torch Usage

Adequate use and implementation

Good Practice (Design/Commenting)

Good spacing and comments
No Header blocks -1

Algorithm

Driver Script present
Shows Usage & Demo & Visualisation & Data usage
Module present
No Data leakage
Difficulty: Normal -5

Commit Log

Meaningful commit messages
Progressive commits used

Documentation

Good Description and more Comments could've been used -1
Markdown used PDF submitted

Pull Request

Successful Pull Request (Working Algorithm Delivered on Time in Correct Branch)
Please remove model files, these will clog up the repository. Your pull request will not be merged until they are removed. Feedback marks will be awarded when these are removed. -2
Request Description could be more informative -1

[wildanazz]

Hi @shakes76, I've attempted to remove the unnecessary files that may clog up the repository. Thank you

[shakes76]

Feedback marks awarded.