Add tutorial notebook for RISE on FRB timeseries data

[loostrum]

Also includes the FRB data itself, so this PR relates to both #707 and #710.

The FRB data and model nearly work with DIANNA timeseries as-is. The only issue is that the data needs some reshaping, as DIANNA expects data in (time, channel) format while the FRB model uses (channel, time). This is further explained in the notebook.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[loostrum]

To Do:

• add to README in tutorials folder (need to find a logo to represent FRB data)

[loostrum]

Tutorial added to README, logo will come later (see #719).
I've also added an explainer on a few modified versions of the input, to make the notebook a bit more interesting.

[loostrum]

#718 is handled in this PR as well - I considered doing it separately but when reviewing notebooks I thought it easier to just have everything in one go.

[elboyran]

Ah, meanwhile, I have messed up with the structure affecting mist the notebooks and their data, labels, and models in #699 (currently up to reviewing)... I wonder what order of merging is best?

[cwmeijer]

Great notebook with the extra's about shifting and adding other sources. One of the coolest tutorials that we have now! Thanks for giving this some love ;-)

I made some comments about some details.

[loostrum]

Thanks for your comments @elboyran and @cwmeijer, I'll get back to them early next week!

[elboyran]

Thanks for your comments @elboyran and @cwmeijer, I'll get back to them early next week!

@loostrum , at the last stand-up, only myself and @cwmeijer were present and I had put a "stand-up" label for my issue (and branch) #667 (PR #699). We think, that if you have to address some comments anyway, maybe you can also apply the small changes according to the new folder structure and we merge my PR first, what do you think? I'll put soon the folder structure in "main" and in "667-reorganize" for convenience.

[elboyran]

Old and proposed folder structures of dianna/dianna and dianna/tutorials :

main branch

dianna (6 directories, 56 files)

├── init.py
├── ...
├── dashboard
│   ├── ...
├── data
│   ├── bee_2.png
│   ├── binary-mnist.npz
│   ├── digit0.jpeg
│   ├── digit0.jpg
│   ├── digit0.png
│   ├── digit1.png
│   ├── labels_mnist.txt
│   ├── labels_resnet50.txt
│   ├── labels_text.txt
│   ├── leafsnap_model.onnx
│   ├── logo.png
│   ├── mnist_model_tf.onnx
│   ├── movie_review_model.onnx
│   ├── movie_reviews_word_vectors.txt
│   ├── season_prediction_model_temp_max_binary.onnx
│   ├── weather_data.npy
│   └── weather_data_labels.txt
├── methods
│   ├── ...
├── utils
│   ├── ...
└── visualization
│   ├── ...

tutorials (4 directories, 38 files)

├── README.md
├── conversion_onnx
│   ├── ...
├── data
│   ├── binary-mnist.npz
│   ├── coffee_test.csv
│   ├── coffee_train.csv
│   ├── leafsnap_classes.csv
│   ├── leafsnap_example_acer_rubrum.jpg
│   ├── movie_reviews_word_vectors.txt
│   └── shapes.npz
├── demo.ipynb
├── img
│   ├── bee.jpg
│   └── digit0.jpg
├── kernelshap_geometric_shapes.ipynb
├── kernelshap_mnist.ipynb
├── kernelshap_tabular_penguin.ipynb
├── kernelshap_tabular_weather.ipynb
├── lime_images.ipynb
├── lime_tabular_penguin.ipynb
├── lime_tabular_weather.ipynb
├── lime_text.ipynb
├── lime_timeseries_coffee.ipynb
├── lime_timeseries_weather.ipynb
├── models
│   ├── coffee.onnx
│   ├── geometric_shapes_model.onnx
│   ├── leafsnap_model.onnx
│   ├── mnist_model.onnx
│   ├── mnist_model_tf.onnx
│   ├── movie_review_model.onnx
│   ├── penguin_model.onnx
│   ├── season_prediction_model_temp_max_binary.onnx
│   └── sunshine_hours_regression_model.onnx
├── rise_imagenet.ipynb
├── rise_mnist.ipynb
├── rise_text.ipynb
└── rise_timeseries_weather.ipynb

---

667-reorganize branch

dianna (8 directories, 66 files)

├── init.py
├── ...
├── dashboard
│   ├── ...
├── data
│   ├── bee.jpg
│   ├── binary-mnist.npz
│   ├── digit0.jpeg
│   ├── digit0.jpg
│   ├── digit0.png
│   ├── digit1.png
│   ├── leafsnap_example_acer_rubrum.jpg
│   ├── logo.png
│   ├── shapes.npz
│   └── weather_data.npy
├── labels
│   ├── coffee_test.csv
│   ├── coffee_train.csv
│   ├── labels_mnist.txt
│   ├── labels_resnet50.txt
│   ├── labels_text.txt
│   ├── leafsnap_classes.csv
│   ├── movie_reviews_word_vectors.txt
│   └── weather_data_labels.txt
├── methods
│   ├── ...
├── models
│   ├── coffee.onnx
│   ├── geometric_shapes_model.onnx
│   ├── leafsnap_model.onnx
│   ├── mnist_model.onnx
│   ├── mnist_model_tf.onnx
│   ├── movie_review_model.onnx
│   ├── penguin_model.onnx
│   ├── season_prediction_model_temp_max_binary.onnx
│   └── sunshine_hours_regression_model.onnx
├── utils
│   ├── ...
└── visualization
├── ...

tutorials (5 directories, 20 files)

├── README.md
├── conversion_onnx
│   ├── ...
├── demo.ipynb
└── explainers
├── KernelSHAP
│   ├── kernelshap_geometric_shapes.ipynb
│   ├── kernelshap_mnist.ipynb
│   ├── kernelshap_tabular_penguin.ipynb
│   └── kernelshap_tabular_weather.ipynb
├── LIME
│   ├── lime_images.ipynb
│   ├── lime_tabular_penguin.ipynb
│   ├── lime_tabular_weather.ipynb
│   ├── lime_text.ipynb
│   ├── lime_timeseries_coffee.ipynb
│   └── lime_timeseries_weather.ipynb
└── RISE
├── rise_imagenet.ipynb
├── rise_mnist.ipynb
├── rise_text.ipynb
└── rise_timeseries_weather.ipynb

[loostrum]

Thanks for your comments @elboyran and @cwmeijer, I'll get back to them early next week!

@loostrum , at the last stand-up, only myself and @cwmeijer were present and I had put a "stand-up" label for my issue (and branch) #667 (PR #699). We think, that if you have to address some comments anyway, maybe you can also apply the small changes according to the new folder structure and we merge my PR first, what do you think? I'll put soon the folder structure in "main" and in "667-reorganize" for convenience.

Yes that seems like a good idea! When #699 is merged I'll check that this PR then matches the new layout

[loostrum]

Most review comments handled, logo added. I'm not sure what to do about the FRB dataset section. We identify a tutorial by a logo and then explain that logo by connecting it to a dataset, but for the FRB case there is no public dataset. I've left out a hyperlink for now. I can also link it to the paper where the model and its training are explained?

[loostrum]

Note: the Mac OS on Python 3.11 CI is failing for reasons unrelated to this PR

[elboyran]

Small suggestions in notebook.

[loostrum]

@elboyran could you check if your comments are now adequately addressed? Then I think this is ready for merging

[elboyran]

Looks nice. Only some MD links checks should pass.