Any chance of core.logic getting extended with probKanren?

[Henrik Larsson]

I have started to play around with ProbLog2 and find the concept of probabilistic logic programming to be super fun. When googeling miniKanren and probabilistic logic programming the following came up:

https://github.com/webyrd/probKanren

So my question now is what are the chances that something like probKanren getting implemented in core.logic and how advance is probKanren vs ProbLog2? What im after is the conditional probabilites that ProbLog2 can handle.

There are some documentation on core.logic (https://github.com/clojure/core.logic/wiki/CLP(Prob)) but it is dated at 2013 and im not sure what the roadmap is for core.logic or if it even has a roadmap.

Thanks for any input regarding this.

Best regards Henrik

[David Nolen]

Definitely interested in it! But I don't have time to work on it myself. So the likelihood of this feature landing is entirely dependent on contributions :)

HTH,

David

--
You received this message because you are subscribed to the Google
Groups "Clojure" group.
To post to this group, send email to clo...@googlegroups.com
Note that posts from new members are moderated - please be patient with your first post.
To unsubscribe from this group, send email to
clojure+u...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/clojure?hl=en
---
You received this message because you are subscribed to the Google Groups "Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email to clojure+u...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Russell Mull]

On Friday, November 20, 2015 at 5:51:05 AM UTC-8, Henrik Larsson wrote:

I have started to play around with ProbLog2 and find the concept of probabilistic logic programming to be super fun. When googeling miniKanren and probabilistic logic programming the following came up:

https://github.com/webyrd/probKanren

Do you know if there have been any publications on this work? I've looked around a little bit, but I can only find the source code. 

- Russell 

[Dragan Djuric]

I am working on something related to probabilistic programming/inference/learning. Not yet ready for use but I hope to get it there the next year.

Although some key building blocks are the same (MCMC), I really cannot see how such things could be integrated to core.logic, and even why. So, I would like very much to hear some of the needed use cases for this (without reading hundreds of pages of someone's PhD dissertation, please :)

Can you write a few "hello world" examples of what you would like to see in core.logic related to probabilistic programming?

[Carl Cotner]

> Although some key building blocks are the same (MCMC), I really cannot see
> how such things could be integrated to core.logic, and even why.

This is not a concrete answer to your question, but probability can be
thought of as a continuous extension of logic where 0 = False and 1 =
True (and .5 = "half true", etc.).

This extension is unique given a few natural conditions.

Carl

[Dragan Djuric]

I know well how probabilistic logic works. It is a superset of the true/false logic, and goes much beyond "half true" etc. My question was about how this would look like when integrated with core.logic vs just using full-blown probabilistic logic-based library/language.

[Carl Cotner]

Sorry, didn't mean to offend. I was just offering a reason for the
"even why". Perhaps I misinterpreted ...

[Dragan Djuric]

I am sorry if my reply sounded like I was offended. I was not, I was just going straight to the point without beating around the bush :)

[David Nolen]

The rationale is simply that Prolog-like systems have embraced it as it appears to be a very natural fit - https://dtai.cs.kuleuven.be/problog/

David

[Carl Cotner]

No worries. I was just trying to point out that at least conceptually
it is a very natural extension.