Replies: 2 comments 1 reply
-
I never thought about it like that and to me that sounds very good 👍 Also, in my experience, trying to get everything right immediately will most likely ensure that nothing gets done since it is too overwhelming and the fear of pushing breaking releases is too big. You can either please the few people that use the package already by not introducing breaking releases or please future users by making the API simple
You can add MLJModelInterface to this package. It is very lightweight so shouldn't add to much loading time to people who don't want to use it. For example, see |
Beta Was this translation helpful? Give feedback.
-
It depends probably. I was having overly conservative developers in mind which I think is likely to cause a continuous problems in the long run. Hard decisions; easy life. Easy decisions; hard life. Your idea of expanding rapidly and then pruning sounds really clever.
I love such big plans 🚀🚀 |
Beta Was this translation helpful? Give feedback.
-
Currently the way development is going I'm going to grow the package very quickly and then prune back, which may be sub-optimal. So instead I'm thinking about ways to more effectively grow. I'd love to hear feedback from the community about what would be most useful. One suggestion I have is the following:
I also planning on integrating the package with MLJ so some components will lie there. I'm also planning on adding a Bayesian interface which could either go in the main package (as it should be fairly agnostic to individual choice), or in the models package, or in it's own SurvivalBayes.jl.
Beta Was this translation helpful? Give feedback.
All reactions