Tutorial notebooks

[vdutor]

What do we want to demonstrate in notebooks? (Please edit/update this issue description as appropriate.)

• [x] General introduction notebook (Vincent)
• [x] Efficient sampling (Vincent)
• [x] GPflux features: monitoring, tensorboard, saving (Vincent)
• [x] GPflux with neural net layers (Vincent)
• [ ] Convolutional DGP (Artem?)
• [ ] Latent Variable model (CDE paper) with extension to Importance Weighting? (Hugh?)

[tingyuansen]

Hey there! Sorry to interrupt. For some reason, it seems like I do not have the authorization to view the tutorial/documentation. I wonder if that is expected? Thank you so much!

[vdutor]

Hi @tingyuansen, thanks for bringing this to our attention. The documentation and tutorial links should all have been updated by now. Definitely let us know if you spot more broken ones.

[tingyuansen]

Thank you!

[dtchang]

On "GPflux with neural net layers", please discuss usage patterns and usage constraints when building hybrid models.

For example, can one put a GPLayer in the middle of a neural net model, preceded and followed by Keras layers? The existing code example (only one) has a GPLayer as the last layer but preceded by a linear Dense layer: tf.keras.layers.Dense(1, activation="linear"), gp_layer, Is the linear Dense layer with output dimension 1 required?

[hkoohy]

Hi, Thanks for releasing this exciting package. Would you be able to add a classification task example to the tutorial? thanks Hashem

[tomiesm]

Hi, Thanks for releasing this exciting package. Would you be able to add a classification task example to the tutorial? thanks Hashem

I would really appreciate if an example of a classification would be provided. Especially for a multivariate data and multiclass problem. I tried to come up with a gpflux code to do that. I failed. The error outputs are extremely unhelpful. Thus, an example would be highly appreciated.