Compare to mlr3proba param spaces and implementation

[pfistfl]

[RaphaelS1]

In particular:

https://github.com/mlr-org/mlr3extralearners/blob/master/R/learner_survivalmodels_surv_dnnsurv.R

And useful upstream functionality:

https://github.com/RaphaelS1/survivalmodels/blob/master/R/dnnsurv.R https://github.com/RaphaelS1/survivalmodels/blob/master/R/helpers_keras.R

[pfistfl]

@RaphaelS1 We should take some time and try to come up with a better design:

• There are a few tabular architectures that might be worthwhile to implement which might work for surv regr and classif (and many others)
• We should have an abstraction for building the architecture which allows 1) composition of architectures and 2) using architectures across domains such as survival, classif, multilabel etc.

[RaphaelS1]

Sure happy to spend some time discussing. Ping me on MM if you want to set up a call