search

hakaru-CS4ZP6 / hakaru

A probabilistic programming language
BSD 3-Clause "New" or "Revised" License
0 stars 0 forks source link

Hakaru Paper Discussion #39

Closed staplejw closed 4 years ago

staplejw commented 6 years ago

We are beginning to tackle major sections of our paper. We have created this google doc for brainstorming ideas. https://docs.google.com/document/d/18redrh2HjrG1Y62bn7KhUCnyKR6yi7JZEwF85q0O3OQ/edit?usp=sharing

JacquesCarette commented 6 years ago

I can't access it. Can you add jacques.carette@gmail.com to the people who can see this?

JacquesCarette commented 6 years ago

BTW, issues are best assigned to at most 2 people (and even then). Sometimes it makes sense to create a milestone which is composed of a bunch of issues; each issue is assigned a single 'owner'.

staplejw commented 6 years ago

@JacquesCarette I have given you access rights for this file. There is also a 'report.tex' in the making, which I will share once more content has been transferred over.

JacquesCarette commented 6 years ago

Ok, got it.

Do you just want me to generally comment on it, or did you have some specific questions?

JacquesCarette commented 6 years ago

@staplejw asking again: what are you expecting from me on this?

staplejw commented 6 years ago

Since nothing is exactly finalized at this point, general comments/critiques would be great!

JacquesCarette commented 6 years ago

The organization of the paper is good. The outlined material is good too. Though there is a danger that some of the pieces will get drowned or feel out of place in the 'finished' paper (i.e. the Sublime plug-in is the most obvious thing here). Just make sure to keep this in mind as you evolve things.

You may find Google Docs to be a poor environment for writing about Hakaru -- the math will soon strain things quite severely. LaTeX would actually be easier...

You should try to sample from some distributions (especially the new ones), and then do plots of histograms. Including some of the equivalent ones, to show 'visually' that they do seem related. Of course, sampling is inexact, so histograms are 'impressionistic' and prove nothing!

JacquesCarette commented 6 years ago

Is this still a 'live' issue?

staplejw commented 6 years ago

Yes. With the poster now out of the way, the paper will become my main focus for the next week.

JacquesCarette commented 6 years ago

When you need me to look at things, make sure to explicitly ping me.

staplejw commented 6 years ago

Noted.