[PD-dev] HMM in PD

Jamie Bullock jamie at jamiebullock.com
Sat Dec 15 13:00:22 CET 2018


Hi Frodo,

Thanks for your interest :)

There are some errors in the help files generally. See for example: https://github.com/irllabs/ml-lib/issues/148#issuecomment-447277967 <https://github.com/irllabs/ml-lib/issues/148#issuecomment-447277967>

However, in the case of HMM, I don’t think it’s currently set up to receive time series (like ml.dtw, which *does* work), so this needs addressing.

I have a day booked on Monday to work on this, so there should be a new release on Monday. I will also try to get it into Deken on the other platforms (including Bela if possible) to make installation easier.

Cheers,

Jamie


> On 15 Dec 2018, at 11:48, Frodo Jedi <frodojedi.mailinglist at gmail.com> wrote:
> 
> Hi Jamie,
> thanks. I know well your externals and also your publication.
> I had tried your HMM external for PD, but it didn't work. The help file seems indeed incorrect and I also cannot save or read data.
> Is it only the help file wrong or there are other bugs you are aware of?
> 
> I would be immensely grateful if you could fix it!
> Could you please push a fix and let me know? More or less how long do you think this will take?
> 
> I am using linux, actually the bela board. 
> 
> Please let me know if I can contribute with testing for debugging
> 
> Cheers
> 
> FJ
> 
> On Fri, Dec 14, 2018 at 9:48 AM Jamie Bullock <jamie at jamiebullock.com <mailto:jamie at jamiebullock.com>> wrote:
> Hi Frodo,
> 
> There is GMM and HMM support in ml-lib, which is currently available via Deken on macOS, or the GitHub releases page here: https://github.com/irllabs/ml-lib/releases <https://github.com/irllabs/ml-lib/releases>
> 
> The GMM *should* just work. Currently the HMM help file is incorrect, but if you are interested in using this I can work on pushing a fix.
> 
> Cheers,
> 
> Jamie
> 
>> On 13 Dec 2018, at 19:09, Frodo Jedi <frodojedi.mailinglist at gmail.com <mailto:frodojedi.mailinglist at gmail.com>> wrote:
>> 
>> Thanks, Joseph, I will check it.
>> 
>> Anyone else has some other ideas?
>> 
>> Cheers
>> 
>> FJ
>> 
>> 
>> 
>> On Thu, Dec 13, 2018 at 7:04 PM Joseph Larralde <joseph.larralde at gmail.com <mailto:joseph.larralde at gmail.com>> wrote:
>> Hey,
>> 
>> I've been working on Jules Françoise's XMM library at IRCAM during a couple of years.
>> It's a really cool GMM and HMM library, in GPL 3 license.
>> I made wrappers for unity and nodejs, and several demo projects based on this library.
>> It's also there in Max as part of IRCAM's MuBu free package.
>> 
>> I planned to port it to pd as well, but don't have enough free time to do it right now (as it's a personal initiative).
>> If someone is willing to take care of this before I can schedule it, I'm available for any advice, discussion etc
>> Otherwise I'll do it, someday ...
>> 
>> Joseph
>> 
>> PS : it's here : https://github.com/ircam-rnd/xmm <https://github.com/ircam-rnd/xmm>
>> 
>> Le 13/12/18 à 18:18, Frodo Jedi a écrit :
>>> Hi all,
>>> is there any working implementation of the hidden markov model in pure data?
>>> Any recommendation? Anyone who is working on the same topic?
>>> 
>>> Cheers
>>> 
>>> FJ
>>> 
>>> 
>>> _______________________________________________
>>> Pd-dev mailing list
>>> Pd-dev at lists.iem.at <mailto:Pd-dev at lists.iem.at>
>>> https://lists.puredata.info/listinfo/pd-dev <https://lists.puredata.info/listinfo/pd-dev>
>> 
>> _______________________________________________
>> Pd-dev mailing list
>> Pd-dev at lists.iem.at <mailto:Pd-dev at lists.iem.at>
>> https://lists.puredata.info/listinfo/pd-dev <https://lists.puredata.info/listinfo/pd-dev>
>> _______________________________________________
>> Pd-dev mailing list
>> Pd-dev at lists.iem.at <mailto:Pd-dev at lists.iem.at>
>> https://lists.puredata.info/listinfo/pd-dev <https://lists.puredata.info/listinfo/pd-dev>
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puredata.info/pipermail/pd-dev/attachments/20181215/ee9438a8/attachment-0001.html>


More information about the Pd-dev mailing list