Hangouts Link: https://meet.google.com/gzm-wmum-pfm
Instructions: Ask to attend and someone should let you in in the first 10 minutes of the meeting. Email breck@mc-stan.org if you have problems or want to attend the physical meeting in New York City.
Meeting Agenda:
Breck: Responsiveness of question answering on Discourse?
Please add agenda items in responses.
Breck: Creation of model/truth repo for inference research evaluation. Yuling interested in creating such, wants to know if anyone else is doing something similar.
I invited Alexey from Google who is also interested in such an evaluation framework and model / dataset.
Stats trying to measure how responsive we are on discourse:
new topics in May: 199
Unanswered topics in May: 20
So we’re at a 10% response rate for May at least. You can fiddle with those reports to figure out other stats.
R on StackOverflow looks like 26% unanswered for all time (didn’t find monthly data)
Duco: Doing Shiny Stan work.
Alexey R: From Tensor Flow Prob. Hear to learn about benchmarking.
Charles: Working on Kinsol solver.
Ben Goodrich: Still trying to get RStanArm to CRAN. Working paper with Blablan??Sp?? Working on getting OCaml working on windows.
Steve B: Stan GPU paper.
Jonah: Fixing RStanArm bug. Adding beta binomials to RStanArm. Mr. T paper working on.
Andrew: Submitted paper, spending time on grants and allocation. Teach Bayesian Workflow class this fall. Working with Shira on fitting.
Breck: Still working on lead paper with Jonathan. Fundraising, elections for Stan Governing Body (SGB). StanCon.
Lauren: Paper submitted to Archive, paper on MRP in psychology. Working to do MRP in on some interesting Australian data.
Mitzi: Working on CmdStanPy, cleaning up and making more pythonic. Getting search going into mc-stan.org/users/documentation.
Bob: Working on compilation issues. Working on getting book done.
Sean: Working on developer meeting.
Matthijs: Autodiff semantics coverage in new compiler.
Andrew reporting that Ryan wrote draft of static analysis research.
General Meeting:
Breck: What is the status of standing up an evaluation set of models to support inference research. Action: Send around email to interested parties to coordinate efforts.
Breck: How responsive are we on discourse? Do we need to allocate more resources? Result is to think about it some more. We are ok, maybe need to hire someone.
General: What is the electorate for coming elections? It is the membership of the “Stan Developers Only” group on Discourse. This could be improved, apparently there is a proposal circulating to do just that.
And the new compiler is passing 86/87 tests (and the other one is a tiny off-by-2e-8 error on a single parameter that seems like a weird cpu-specific floating point issue). Waiting to hook up optimizations or faster compile times before we do our first beta release!