Victor Rodriguez-Fernandez

See the time slider of Figure 1 of this paper: https://ieeexplore.ieee.org/document/7192717/figures#figures

There should be a point-based selector to see the time windows around one point. The range of sorrounding points, both in the future and in the past and in the...

That way, the artifacts tree in wandb would show first which dataset it is (now the name, with this it would be the type), and then elements for the learner...

Currently, clustering is done on the projections. It shoud be done directly on the embeddings

Pintar, encima de la serie temporal, rectangulitos que representen los clusters que hay en cada instante temporal. La altura de los rectángulos vendrá dada por el numero total de clusters.

Upload new datasets from the app

3

Things to consider: - The app should call the notebook `01_create_artifact` to register the user dataset in wandb. This can be done with [papermill](https://github.com/nteract/papermill) - Since the app is going...

I do not think that training the encoder from the app makes sense given the time needed for that...however, we may offer the user with a configuration panel to select...

Also, a button to come back to that value if the user changes it would be nice

Zooming and panning the projections plot with click-drag and scroll

1

Currently the way of zooming and panning the projections is quite clunky. It would be great to have something eaiser as the [tensorflow embedding projector](https://projector.tensorflow.org/)