Make this available in maven central

[francisdb]

If you want people to use this you better release the project on maven central.

I also suggest a cross-scala0version sbt build for 2.10 and 2.11

this guide should help:
http://www.scala-sbt.org/0.13/docs/Using-Sonatype.html
or
http://www.cakesolutions.net/teamblogs/publishing-artefacts-to-oss-sonatype-nexus-using-sbt-and-travis-ci

[danielkorzekwa]

will do it next week.

[francisdb]

Cool, if you want I can set up the base sbt build?

[danielkorzekwa]

Sure, I was actually thinking about switching the build from maven to sbt, feel free to do as much of the work as you want, I'm on holiday now and will be back on Monday next week.

[francisdb]

see pull request #15

[danielkorzekwa]

Completed

• bayes-scala snaphot is deployed to Sonatype repo on every commit

Further work (raise new issues if needed)

• deploy bayes-scala for Scala 2.11 to maven repo
• deploy non-snapshot versions to public maven repo

Notes:
Only bayes-scala for 2.10.4 is built at the moment. Supporting cross scala /cross jdk build and deployment from travis requires some additional work e.g. snaphot deployment to maven repo should be made just for one jdk version and not all java versions configured in .travis.yml file.

[francisdb]

Well, I suppose release builds will be performed manually so those can be crossversion

[danielkorzekwa]

The first release I plan doing manually, later on with sbt-release. Even snapshot could be cross versioned, with some scripting work to sbt and travis I think.

For playing with bayes-scala, the snapshot verion in sonatype shall be sufficient, therefore I'm not going to create release build unless someone actually wants to use bayes-scala in some production deployed application.

[francisdb]

unless someone actually wants to use bayes-scala in some production deployed application

That would be me :-) We are currently evaluating some bayes libs...
Was the reason for creating this ticket ;-)

[danielkorzekwa]

:), good, raise a ticket for release built once you are happy with bayes-scala and then I will create it for you!

[francisdb]

Cool, I'll have a go with the snapshot. Do you know how your lib compares to other scala bayes libs, like FACTORIE and Figaro?

[danielkorzekwa]

I started this project, because at that time, three years ago, there was
nothing really in java/scala for deterministic inference for discrete and
hybrid networks. If Infer.Net was available for commercial use I would
probably go for it, I also tried a few other libraries, but there were
always some features I was missing.

I had 3 requirements:

1. Implent skill based model with discrete variables, learning latent
   models with EM
2. Implement skill based model (true skill), hybrid model, expectation
   propagation,
3. Extend true skill model and build multi skill model (dynamic bayesian
   networks) and in the end, build infinite skill model (gaussian process)

I tried:
SamIam - cool for toy problems, but just discrete networks
Bnt - Matlab, discrete and gaussian networks, not java/scala
Factorie seemed to be too complicated and focused more on nlp, tokens,
etc., poor doc,
Figaro at that time wasn't open source I think, and it didn't get my
attention
Mallet - very good, but just discrete factor graphs
Infer.Net - best I've ever seen, not java/scala, a bit weird API
Smile - strange API, didn't try it in the end

First I started with building very specific APIs for my needs, e.g. api for
cluster graphs + loopy BP, api for factor graphs + EP, api for Gaussian
processes, later on I put a high level DSL (examples in README), which
makes the details of particular inference algorithms hidden from the user.

Describe your problem your want to solve and I will tell if the current
bayes-scala will work.

On 4 March 2015 at 20:14, Francis De Brabandere notifications@github.com
wrote:

Cool, I'll have a go with the snapshot. Do you know how your lib compares
to other scala bayes libs, like FACTORIE and Figaro?

—
Reply to this email directly or view it on GitHub
#12 (comment)
.

Daniel Korzekwa
Machine Learning Engineer
priv: https://www.linkedin.com/in/danielkorzekwa http://danmachine.com/
blog: http://blog.danmachine.com

[francisdb]

@danielkorzekwa we're doing pretty basic stuff so the current features should more than suffice

[danielkorzekwa]

do you build discrete or continous/hybrid networks?

On 5 March 2015 at 08:35, Francis De Brabandere notifications@github.com
wrote:

@danielkorzekwa https://github.com/danielkorzekwa we're doing pretty
basic stuff so the current features should more than suffice

—
Reply to this email directly or view it on GitHub
#12 (comment)
.

Daniel Korzekwa
Machine Learning Engineer
priv: https://www.linkedin.com/in/danielkorzekwa http://danmachine.com/
blog: http://blog.danmachine.com

[francisdb]

discrete

[francisdb]

maybe we should move this discussion to gitter or something similar?

[danielkorzekwa]

I'm sorry but I can't really chat online, too busy.

[francisdb]

np, thanks for sharing the lib