commiting notebook for modelling of count data using ELFI

[adhaka]

Summary:

This is a notebook where I try out a few different things. To summarise I have tried to:

1. See how basic ELFI can be used for modelling highly non linear temporal count data(Ricker stochastic).

2. Add more summary measures in order to reduce the variance of the samples from the original example.
3. Implement the DTW(Dynamic Time Warping) algorithm which can be used for comparing unequal time series.
4. Model a data downloaded from the internet(Geyer data) count data on visits to a shop over 24 hours.
5. Use a simple linear Gaussian model for the data generating process
6. Carry out likelihood free inference by both SMC and Rejection sampling.

Copyright and Licensing

Please list the copyright holder for the work you are submitting (this will be you or your assignee, such as a university or company):
Akash Kumar Dhaka

By submitting this pull request, the copyright holder is agreeing to license the submitted work under the following licenses:

• Code: BSD3 (https://opensource.org/licenses/BSD-3-Clause)
• Documentation: CC-BY 4.0 (https://creativecommons.org/licenses/by/4.0/)

[ivan-afonichkin]

Very nice notebook, I like it a lot. Some things:

1. Is it a good idea to download data inside the code? Or is it better just to provide the data next to your ipython notebook? Because I pulled your code and then was looking into your code without internet, so I had to wait until I will have an internet in order to look into your code.
2. I'm not saying it's necessary, but sometimes it's a good idea to comment the results you've got (and your code).

Otherwise, it's an excellent notebook.

[SidRama]

I found the notebook to be quite interesting and it demonstrated some nice concepts. To add to the above comments:

1. Many cells can be merged together. This would make executing parts of the code together easier and also helps enhance the overall aesthetics.
2. There appears to be some runtime errors in the latter part of the notebook. I am not sure if this is a genuine problem and if it effects the validity of the results.

Overall, good effort! 👍

[vuolleko]

Very good notebook. Some requests:

• more explanations (even brief ones)
• there's a remaining TODO. Either do it or remove it. :)
• there are titles "Ricker's Model" and "Specify the model(Simulator) for generating synthetic obervations" with no content?
• DistanceMeasures.DTW seems to recurse?
• the Ricker model is provided under elfi.examples; the code seems mostly identical. Coincidence? Source for copied code MUST be identified (after verifying its licence allows usage under BSD3!)
• what's model_trial2?
• no need for an "identity" summary node, just make the simulator a parent for distance directly
• please discuss results
• in Out [60] (plot_marginals), something is clearly wrong with t1
• the pools are unused
• what's with the commented-out lines?
• overall cleaning
• move the notebook to zoo/notebooks/elfi-Poisson-Milestone2.ipynb (what's the milestone?)