astrohackweek / astrohackweek2015 Goto Github PK

This was cool last year. I'll try and get to NY in time to join, and will need beer and company. Where shall we go? Answers on this postcard please! :-)

@davidwhogg @dhuppenkothen @kbarbary
Can you please think of a hack or two that either you plan to work on, or think might make a good AHW project, and then post it to the Astro Hack Week 2015 Hack Index here:

https://hackpad.com/Astro-Hack-Week-2015-Hack-Index-RLfELNMgCSC

Thanks!

The repo got a bit messy during the week, so I'm going to reorganize it as such:

• day1-data-exporation
• day2-big-data
• day3-machine-learning
• day4-day5-inference
• breakouts/...
• hacks/...

Before I do, wanted to know what the status of the bayes-cv branch is. I assume it should be merged into master? I can move the CV part that was a hack into hacks. @drphilmarshall @eggplantbren

Hey @j-faria !

Hope you got home OK, and have recovered a bit :-)

I think you should start a new repo to house our investigation of Bayesian cross-validation. Why don't you start one, extracting [the work you have done so far] from this repo, and invite me, @davidwhogg @jakevdp and @glouppe as collaborators? We are all interested in comparing machine learning-style CV with Bayesian evidence (and other quantities), and some short demonstrations could make a nice contribution to the astroph arxiv for everyone to learn from. What do you think?

Hi all!

I just adapted last year's Astro Hack Week wiki into this year's, and started a new hackpad collection for us to use. It's almost ready to be emailed to the participants to get them started thinking about hacks, but first I think some of the logistics details need fixing.

@dhuppenkothen can you please do some browsing of this site, starting from the README at https://github.com/AstroHackWeek/AstroHackWeek2015 , and fill in whatever missing details you can, please?

Cheers

Phil

I'll dig out last year's, edit and for comment before @dhuppenkothen or I sends it out.

Slated for Friday at 11am. @dfm, if you lay out your plan in a reply, I can provide (links to) notebooks that can (probably) be adapted...

We should find a place to keep track of this... where is best? Git repo? Website?

• move content on 2015 out to a list of "past events" which should also have 2014 link.
• make bare-bones content for 2016

I think we need to update it to match the wiki.

Should we send out an email with info telling people how to submit blog entries? I can manage the Nikola blog site.

If so, we should also create a schedule of when the entries will go live.

@eggplantbren - is there anything you want or need to check in to support your Thursday session 1 tutorial?

Hi @dfm - this one's for me to do I think, so you can see what I'm planning - and so we can issue ourselves with more prep. From my notes, we're roughly aiming for:

• Set up example, show analytic solution
• Implement MH sampler, run
• Visual model checks: fit looks bad
• Extension to model, visual check. Fit looks better.
• Quantified model check by PPP. Brief discussion, of sensitivity, and comparison to classical hypothesis test.
• Quantified model check by CV. Brief discussion, of question being answered.
• Quantified model comparison by Evidence - involves diversion to SMC. Brief discussion, of sensitivity in general and prior volume in particular.

@kbarbary: are we ready to merge day1 into master?

we need to choose dates now

@eggplantbren, do you have anything to check in for Friday's second (i.e. after coffee) session?

We need to fix this, even retrospectively!