Replies: 15 comments 33 replies
-
The new website at https://fslab.org is just currently a generic display of the projects in the incubation space. @muehlhaus feel free to populate with alternative material as things come together using whatever web tech you like :-) The current projects in this incubation space are
Note FSharp.Charting is windows-only and should be renamed and moved to fsprojects - we don't want any windows-only stuff in FsLab going forward. XPlot will probably soon be better replaced by Plotly.NET So this is really Quite An Empty Space. Let's fill it with goodness and make sure to also embrace other assets like FSharp.Data, Giraffe etc. which live outside this space. |
Beta Was this translation helpful? Give feedback.
-
Thank you @dsyme for quickly clearing the way and taking care of old stuff. |
Beta Was this translation helpful? Give feedback.
-
I think flips would be an excellent inclusion. |
Beta Was this translation helpful? Give feedback.
-
Awesome progress! I'll be glad to help contribute tutorials, samples, or anything else that may be needed to help get folks started. |
Beta Was this translation helpful? Give feedback.
-
I'd encourage everyone involved in FsLab to watch the SciSharp discussions, and one or more of these repos
These are implemented in C# but we want them to have good F#/C# docs and experience, and we should still get involved since they do form an important entry point for Python-aware users into the F#/.NET data science ecosystem. We want SciSharp and FsLab to work well together and that means good communication, cooperation and co-participation so lpease do watch those repos if you're interested in them |
Beta Was this translation helpful? Give feedback.
-
So, I took this of Hadley Wickham book: it pretty much specifies a normal data science workflow. For Import, Tidy, Transform, Visualize, and Communication steps, we got some tools: TypeProviders, Deedle, Plotly.NET, FSharp.Stats, etc. The sore thumb would be the Model step. More specifically, simplifying and making prototyping easy with quick turn around. |
Beta Was this translation helpful? Give feedback.
-
@dsyme To get people started adding documentation and tutorials, I think for these kinds of F#-first data science projects it is really important to get an easy way to include plots in the docs. We mentioned this first in fsprojects/FSharp.Formatting#611 (cc @HLWeil), and we have a workaround that postprocesses the results to generate inline html, but that won't work with the awesome new |
Beta Was this translation helpful? Give feedback.
-
I'll review the issue - thanks for the context |
Beta Was this translation helpful? Give feedback.
-
@kMutagene I've been considering accepting .ipynb as input to the |
Beta Was this translation helpful? Give feedback.
-
@siavash-babaei That is a great idea. We would start creating a landing page. My suggestion would be to follow your (or Hadley Wickham) schema for a normal data science workflow:
|
Beta Was this translation helpful? Give feedback.
-
I'd like to welcome @luisquintanilla to FsLab. Luis works on .NET Documentation for F# machine learning and much more at Microsoft and is passionate about F# in the data-science and machine learning space. Luis is happy to help co-admin in FsLab |
Beta Was this translation helpful? Give feedback.
-
Thanks @dsyme Looking forward to working with the rest of the community on these projects 🙂 |
Beta Was this translation helpful? Give feedback.
-
I'd like the organization to help with consolidating the experience of the libraries in the different environments (.net interactive, fsi, etc.). I'm taking the exemple of Flips which I'm getting acquainted with, it would be good to be able to print the decisions and equations into those environments in ways that make them more appealing to end users. Those efforts can be consolidated if some of the people working at microsoft on the projects where this integrates / adds value, by contributing to individual projects with the proposed extensions. There are also few other things that would be good for making fslab more attractive:
Those things would need to function on their own, not consume too much of the libraries maintainers time, wether or not the library is incubating or has a separate org. |
Beta Was this translation helpful? Give feedback.
-
One important thing that is sorely lacking is proper Vidro Tutorials. Checking DataCamp or Udemy, one can find thousands of hours of courses for say python. They make adoption and uptake real easy. Furthermore, say when a general programmer already familiar with core language and concepts, wants to start doing data science, it would take him/her tens of hours rather than hundreds to start being really productive in the field. I reckon around 150 hours of course would cover most thing from basics to advanced, concurrency, parallelism, design, backend, frontend, type providers, computation expressions, data science, etc. What few we currently have, scattered over Udemy, Pluralsight, FsharpTV, etc., are commonly very short, old, often only covering the very basics, and sometimes incomplete. Producing material is neither cheap nor easy, but those who can, are already donating significant time and effort. This has just as much effect - if not more - than code contribution or whatever: DSyme's Expert F# or Scott Wlaschin's site is as an important asset as any library or framework out there and I would feel just as giddy with an F# Spark as if "DDD with F#" was turned into an applied course. A good FsLab I think should include neatly organized tutorials for the various tasks and workflows, hopefully in Video as well as textual formats. |
Beta Was this translation helpful? Give feedback.
-
I saw this somewhere and thought "Hoom, Interesting Point": "The problem with the F# library ecosystem is that it is VERY insular. People write libraries in F# for F#. Writing in F# is great. However most libraries should be written for .Net instead: the traction would be much larger and .Net is a perfectly good API target. While F#-F# sometimes results in better APIs (e.g. Fabulous), the restriction to .Net APIs would improve most libraries by preventing excesses (a common culprit is crazy operators). Libraries which are written for F# which should be written for .Net include: FAKE, Farmer, most type providers (the erasing ones). If these were retargeted it would 1. help them remain actively maintained projects 2. integrate F# by bringing in contributors who normally work in other languages and turning F# into a useful producer language not just a consumer one." |
Beta Was this translation helpful? Give feedback.
-
This is to continue the discussion with @muehlhaus, @siavash-babaei, @panesofglass, @matthewcrews, @Shmew, @kMutagene, @cartermp, @nhirschey, @bvenn, @WalternativE and others at fslaborg/zzarchive-FsLab#137
Welcome all!
Beta Was this translation helpful? Give feedback.
All reactions