790 add scientific use case frb to dashboard

[laurasootes]

This PR adds the FRB use case to the dashboard.

The FRB case and data is quite different from the weather example, and therefore requires a lot of FRB-specific data conversions and such, making the code not that pretty. It only works for RISE (as is the case in the notebook). Since it is not straightforward, I made an issue to get it to also work for LIME #836 .
Fixes #790

Additionally, I split the left menu; it now starts with a choice for example or own data, and only then shows the choices/inputs required depending on example/own data. I think this makes the dashboard less confising and less visually overwhelming: you do not see choices and buttons you do not need. Extension to #818

I updated the tests in accordance with the new input data layout

Also, added the DIANNA.utils.downloader for the data used in the dashboard. Fixes #839

[SarahAlidoost]

@laurasootes nice work! 👍 I like the new interface with separate sections in the left menu, much cleaner now. I was able to run the dashboard, and its tests. Just a few comments on the changed file. Also, I noticed that the quality of frb images is not the same as those in the notebook. can you check it? see here:

from dashboard

from notebook

[laurasootes]

@SarahAlidoost Thanks for your review!
I have implemented your suggestions:

• for the FRB data, I used the downloader function now, and also implemented these for the other examples
• for the naming of the variables related to the FRB adjustments; I renamed them and added some inline comments as you suggested. I hope the code is clearer now

As for the result; the difference is caused by the parameters used. Changing the model parameters gives me this:

(The difference in darkness is caused by the overlay, which was not done in the notebook)

[SarahAlidoost]

As for the result; the difference is caused by the parameters used. Changing the model parameters gives me this:

@laurasootes I amnot sure if the difference is caused by the parameters. Here is the screenshot of my dashboard with the same parameters as yours. There are two main differences:

1. the figure's aspect ratio: yours is square, while mine is rectangular. I checked the explanation[0, :, ::-1].T.shape that is (32, 64).
2. the x and y axis labels are missing in my figure. To check this, I saved the fig returned by plot_image() as fig.savefig('...') or in the notebook after 4 - Plot the explanation and input data next to each other visualization.plot_image(heatmap) returns the same plot as mine. Can you please check these two issues?

[laurasootes]

As for the result; the difference is caused by the parameters used. Changing the model parameters gives me this:

@laurasootes I amnot sure if the difference is caused by the parameters. Here is the screenshot of my dashboard with the same parameters as yours. There are two main differences:

1. the figure's aspect ratio: yours is square, while mine is rectangular. I checked the explanation[0, :, ::-1].T.shape that is (32, 64).
2. the x and y axis labels are missing in my figure. To check this, I saved the fig returned by plot_image() as fig.savefig('...'). Can you please check these two issues?

To compare with the one in the notebook I temporarily fixed the aspect radio and added the axis labels in the dashboard (and removed the background original image), so that I could check how far the red part of the explanation reaches as the ’shape’ of the explanation was my concern. (I did not add the axis labels in the dashboard as I think it does not add any information for the explanation). Besides these formatting things I believe your result is the same as mine right?

[SarahAlidoost]

As for the result; the difference is caused by the parameters used. Changing the model parameters gives me this:

@laurasootes I amnot sure if the difference is caused by the parameters. Here is the screenshot of my dashboard with the same parameters as yours. There are two main differences:

1. the figure's aspect ratio: yours is square, while mine is rectangular. I checked the explanation[0, :, ::-1].T.shape that is (32, 64).
2. the x and y axis labels are missing in my figure. To check this, I saved the fig returned by plot_image() as fig.savefig('...'). Can you please check these two issues?

To compare with the one in the notebook I temporarily fixed the aspect radio and added the axis labels in the dashboard (and removed the background original image), so that I could check how far the red part of the explanation reaches as the ’shape’ of the explanation was my concern. (I did not add the axis labels in the dashboard as I think it does not add any information for the explanation). Besides these formatting things I believe your result is the same as mine right?

thanks for the explanations 👍 , now I understand what causes the differences. Yes, if I add the aspect ratio and the lables, the figures are the same. 😄

[laurasootes]

@SarahAlidoost thanks for your suggestions. I removed the background from the plot, which also removed the normalisation problem :)

[SarahAlidoost]

@laurasootes thanks for addressing the comments 👍