Rasmus F. Ørsøe

Should we rename `GraphDefinition` to `DataRepresentation`?

1

As we are transitioning into a deep learning library, I think we should strongly consider to rename one of our key modules `GraphDefinition` to a less GNN-centric name, e.g. `DataRepresentation`....

Refactors `StandardModel` such that the easy syntax (`.fit`, `.predict_as_dataframe`, etc) is stored in a seperate class `EasySyntax`, allowing different `Model` s to use it without the restrictions of `StandardModel`. Refactor...

Make `DataConverter` accept a list of files to convert instead of only a directory

1

**Is your feature request related to a problem? Please describe.** `DataConverter` accepts a single argument `input_dir: Union[str, List[str]]` which point to one or multiple directories. These directories are searched using...

Add an interactive event viewer

1

Would be nice if we could display events quickly and take snapshots for presentations. Perhaps this can be used as a starting point: https://towardsdatascience.com/guide-to-real-time-visualisation-of-massive-3d-point-clouds-in-python-ea6f00241ee0

NormalizingFlow

4

This PR adds compatibility for normalizing flows to graphnet. Specifically, this PR does the following: 1. Adds `NormalizingFlow` in `graphnet.models.flows` as a new generic model class which all normalizing flows...

Update the example scripts to use the new `GraphNeTDataModule`

2

Since #642 we've gotten a much more elegant way of producing dataloaders for training and inference. We should update our training scripts to use this new method instead, and strongly...

Deprecation List for GraphNeT 2.0

2

This thread tracks the items marked for deprecation for the release of GraphNeT 2.0. - [ ] - ´gnn´ argument in `StandardModel`. See #646 - [ ] - call to...

Add functionality to derive graph-specific scalar quantities

1

PyTorch Geometric have a pretty low focus on graph-specific quantities that can be derived off of graphs. There exists many additional quantities we could extract from events - both on...

Re-think sim_type

2

**Is your feature request related to a problem? Please describe.** In I3TruthExtractor we rely on the variable `sim_type` to modify the extractor's behavior. The variable is inferred by the extractor...

This target value is regressed by RetroReco and the prediction itself is available already in the RetroReco table. However, the truth value is not. I fear this value is not...