search

ICB-DCM / pyABC

distributed, likelihood-free inference
https://pyabc.rtfd.io
BSD 3-Clause "New" or "Revised" License
206 stars 44 forks source link

Semi-automatic selection of summary statistics #108

Open yannikschaelte opened 5 years ago

yannikschaelte commented 5 years ago

Implement a scheme to semi-automatically select good summary statistics, based on https://arxiv.org/pdf/1004.1112.pdf / https://arxiv.org/pdf/1302.5624.pdf.

yannikschaelte commented 5 years ago

Will probably be only of limited use for typical biological types of problems / data, where more specific methods may be required.

Havaldar commented 4 years ago

@yannikschaelte is there still interest in implementing this enhancement? If so could I take a crack at it? Wanted to check before starting anything.

yannikschaelte commented 4 years ago

Hi, first of all, thanks for your interest. Yes, we still plan to have this implemented at some point, as a base/reference of more tailored approaches. Are you personally interested in using this method? A problem is that currently we have low capacity to look into this, or to help you in doing so. But if you are interested, that would really be nice, and we could also have a short telco.

Conceptually, it should be possible to keep the algorithms presented in the paper quite decoupled from the main pyABC ABCSMC routine, so that would be good.

EmadAlamoudi commented 4 years ago

Dear @Havaldar, thanks for your interest. As @yannikschaelte already said that there is still an interest to implement this feature. If you still interested in working on this, I would be more than happy to be involved in the development process. As an initial step, I would recommend a virtual meeting to discuss the different possibilities and exchange ideas.

Havaldar commented 4 years ago

@EmadAlamoudi @yannikschaelte I am so sorry for the sudden disappearance (was working on finishing my thesis and got sidetracked). I would love to help work on this enhancement and can get on a virtual chat whenever is convenient for you.

EmadAlamoudi commented 4 years ago

Dear @Havaldar, we wish you all the best in your thesis. I will post multiple options for the meeting and please select what is best for you. @yannikschaelte you can also select the time that suits you. options are:

(all times in CET)

If all did not agree with your agenda, please feel free to suggest additional times.

Havaldar commented 4 years ago

October 8th sounds great!

EmadAlamoudi commented 4 years ago

Great, I will prepare a zoom link and share it with you for the meeting. Can you please give me your email address so we can communicate easier.

Havaldar commented 4 years ago

ah2717@nyu.edu @EmadAlamoudi

yannikschaelte commented 3 years ago

Hi all, any developments here?

Havaldar commented 3 years ago

@yannikschaelte I am sorry for such a delayed response! I have been stuck due to immigration issues and haven't been able to get to this and am not sure I will be able to in the near future. I am relly sorry about not responding earlier!

yannikschaelte commented 3 years ago

Oh, sorry to hear that, no problem! Hope the immigration issues work out. I was just asking because at some point we will investigate in a similar direction. I'll post here once we do that (not in the next few weeks probably), to not duplicate things.

yannikschaelte commented 3 years ago

Hi @Havaldar , just to let you know that we are now using automatic summary statistic regression in an own project and thus implemented a version for pyABC. Happy to discuss if you want to follow up on the topic.