Stan ODE paper?

[Aki Vehtari]

Hi,

I couldn't find by quick search, but I guess there is already some Stan ODE paper illustrating how easy it use ODEs in Stan?

Aki

[Daniel Lee]

I believe it's in the reverse mode autodiff paper on arXiv.

> --
> You received this message because you are subscribed to the Google Groups "stan development mailing list" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

[Daniel Lee]

See Chapter 13: http://arxiv.org/abs/1509.07164


Daniel

> On Apr 13, 2016, at 8:50 AM, Aki Vehtari <aki.v...@aalto.fi> wrote:
>

[Aki Vehtari]

Daniel, thanks for this, although I was thinking something suitable for a journal club not a booklet :)

Could you add table of contents in the beginning?

Aki

[Bob Carpenter]

The current version of the arXiv paper is a bit sketchy
on the nested autodiff and also doesn't cover the Jacobians
required for the stiff ODE case that Michael worked out.

I think it might make sense to have a dedicated paper on
this topic that presupposed the longer tome. Otherwise,
we'd have to either gloss over the details or repeat a lot
of material.

- Bob

[Aki Vehtari]

For a potential journal club, I'm not interested in autodiff (although I'm certain that DOE autodiff is useful, too), but more of a Stan ODE application example with bit of background on ODE solvers to advertise Stan ODE capabilities, but I guess there is not such paper yet.

Aki

[Daniel Lee]

How's Bob's case study on soil carbon modeling?

http://mc-stan.org/documentation/case-studies/soil-knit.html

[Bob Carpenter]

We need another one of these for PK/PD, but the soil carbon's
definitely an easy example and it comes with working R code.

- Bob

[Daniel Lee]

Yup. Working on the monster model, but need to find more time!


Daniel

[Aki Vehtari]

Thanks! I knew Bob had made such analysis, but I couldn't find this.

Aki

[Sebastian Weber]

I hate to admit if Stan is not good at something, but ODEs have definetly not been a strong field for Stan up till now. That is, Stan oftentimes just ran against the wall by freezing the boost odeint integrator. With the robust RK45 patch and CVODES this is history.

I am happy to file a PK/PD example soon with the shiny new CVODES integrator.

Sebastian

On Wednesday, April 13, 2016 at 6:28:26 PM UTC+2, Bob Carpenter wrote:
> We need another one of these for PK/PD, but the soil carbon's
> definitely an easy example and it comes with working R code.
>
> - Bob
>
> > On Apr 13, 2016, at 12:23 PM, Daniel Lee
> >

> > How's Bob's case study on soil carbon modeling?
> >
> > http://mc-stan.org/documentation/case-studies/soil-knit.html
> >
> >
> >
> >> On Apr 13, 2016, at 12:15 PM, Aki Vehtari
> >>

> >> For a potential journal club, I'm not interested in autodiff (although I'm certain that DOE autodiff is useful, too), but more of a Stan ODE application example with bit of background on ODE solvers to advertise Stan ODE capabilities, but I guess there is not such paper yet.
> >>
> >> Aki
> >>
> >>> On Wednesday, April 13, 2016 at 7:00:37 PM UTC+3, Bob Carpenter wrote:
> >>> The current version of the arXiv paper is a bit sketchy
> >>> on the nested autodiff and also doesn't cover the Jacobians
> >>> required for the stiff ODE case that Michael worked out.
> >>>
> >>> I think it might make sense to have a dedicated paper on
> >>> this topic that presupposed the longer tome. Otherwise,
> >>> we'd have to either gloss over the details or repeat a lot
> >>> of material.
> >>>
> >>> - Bob
> >>>
> >>>
> >>>> On Apr 13, 2016, at 10:16 AM, Aki Vehtari
> >>>>

> >>>> Daniel, thanks for this, although I was thinking something suitable for a journal club not a booklet :)
> >>>>
> >>>> Could you add table of contents in the beginning?
> >>>>
> >>>> Aki
> >>>>
> >>>>> On Wednesday, April 13, 2016 at 4:07:13 PM UTC+3, Daniel Lee wrote:
> >>>>> See Chapter 13: http://arxiv.org/abs/1509.07164
> >>>>>
> >>>>>
> >>>>> Daniel
> >>>>>
> >>>>>
> >>>>>> On Apr 13, 2016, at 8:50 AM, Aki Vehtari
> >>>>>>