Graph visualisation & embedding dimensionality reduction

[KristinaUlicna]

PR contribution summary

Why is this PR useful / good for? Please describe the problem(s) you're trying to address.

• This PR improves the evaluation of the trained GCN model:
• Allows to visualise the node features / embeddings before & after the GCN
• Allows to localise where in the graph the classifier makes errors, on per-node & per-edge basis.

List of proposed changes / linked issues & discussions

• Resolves Return node & edge embeddings from semi-last GCN layer #192
• Resolves [FEATURE REQUEST] Allow user to perform dimensionality reduction on extracted patch features #161
• Resolves Visualise GCN predictions of node/edge probabilities on graph #172
• Resolves [DEVELOPMENT] Repurpose logging to LOGGER to recycle its config #230

What should a reviewer concentrate their feedback on?

• 🏃 Notebooks to run - particularly infer_predictions.ipynb
• 💻 Code quality
• 📝 Everything looks OK?

What type of PR is this? (check all applicable)

• 🪄 Feature
• 🧑‍💻 Code refactor / style
• 🔥 Performance Improvements

Added tests?

• 🙋 no, because I need some help

---

PR review summary

Describe what this PR does & how you reviewed the individual items, where needed:

Some helper checks to tick off:

• Focus on image annotation
• Focus on model training
• Could any optimization be applied?
• Is there any redundant code?
• Are there any spelling errors?

In conclusion, after my review, I'd like to:

• 🙋 ask some clarifying questions
• 🙅 suggest some specific changes

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[crangelsmith]

Is this branched out of inference? I see a lot of similar changes in the files changed. If so, maybe we could change the base for easier reviewing, or should I wait until PR #223 is merged?

[crangelsmith]

Looks good, notebooks run as expected and make very useful graphs for improving the model!

Happy for this to be merged as long as it is added to the pipeline soon. I think all plots can be saved in the run directory (maybe under a subdirectory plots). A function evaluation(graph, classifier) could be written more generally so it can be used to evaluate performance at epoch level (if that is useful), after training, as well as in an unseen graph.