Decompose VS kotlindl

I use Decompose, although I also have a kmp project which that supports. I find the compose extensions providing a very easy API when integrating with compose.

For one of our apps we used https://arkivanov.github.io/Decompose/ for navigation. It is mentioned in the Compose Multiplatform repo, but we found it pretty obscure to use and it has some confusing concepts built in.

View model is Googles recommended approach, but there are other approaches out their you can use like decompose

When you are building your app in KMM, sometimes you need to pack some libraries along with project files in the shared module. For example if you are using Decompose you need to access it's Value class in Swift.

I don't have much experience with conductor, so I'm curious what your concerns about lifecycle is lacking compared to fragments? But if your team already knows that framework I might just stick with views over fragments. I see there is a compose integration if you do ever plan on picking that up with conductor. Otherwise compose makes fragments obsolete and your team already knows conductor. I do agree the navigation story in compose is not mature especially after jetpack compose navigation. But there are other 3rd party libraries like compose destinations or decompose, but would be nice to see something better 1st party.

I personally like decompose for its testability, clear separation of concerns, and apis for compose animations. Plus it's multiplatform so you can use it for other kmp targets.

Full discretion, all my new apps are built using mvi with MVIKotlin and decompose for the compose extensions for navigation animations. A sample project I have built with that architecture can be found here.

Check out Decompose, it's multiplatform - https://github.com/arkivanov/Decompose

Decompose

This is a really nice sample app (works out of the box) that Jetbrains have made to showcase the capabilities of this Kotlin Deep Learning library https://github.com/Kotlin/kotlindl

There are Java bindings for TensorFlow, but that's quite low level. I tried to see if I can get some Keras API for Scala, but I'm no expert and haven't had enough time to invest in this, so it's stuck in alpha. Maybe I develop it slow burning over the next year. A bit envious that Kotlin has a Keras-like library.

KotlinDL 0.3 is available now on Maven Central with a variety of new features — check out all the changes that are coming to the new release!

Could you please add an issue with your use-case and proposed solution? Or write here more details. Do you need just serialization to objects in memory? or just in bytes?

While KotlinDL seems to be a good solution by Jetbrains, I would personally stick to Java frameworks like DL4J for a better community support and likely more features.

For deep learning this is your best bet: https://github.com/jetbrains/kotlindl

So, the KotlinDL library could solve classification and regression tasks on tabular data and image data.

Thank you for your feedback! Kotlin∇ is a research prototype for building differentiable programs from basic primitives. If you’re looking for a higher-level API, KotlinDL [1] or KInference [2] might be a better fit for building end-user applications such as image, speech or language processing.

[1]: https://github.com/JetBrains/KotlinDL

[2]: https://github.com/JetBrains-Research/kinference/