Anomaly likelihood

[breznak]

Likelihood is an alternative method of anomaly computation, ported from Python repo.

Beta work, please review for high level functionality! 👍

TODO:

• port all required methods/functionality
  • fix syntax
  • improve "API" design, methods' signatures,...
  • fix formatting
• integrate into nupic's cmake build system
• port tests from tests/unit/.... in nupic-PY Anomaly Likelihood: port additional tests from python  #124
• create a header file (at the end, when everything is cleaned up)
• integrate into Anomaly class - almost, some PROBLEMS *)
• resolve CI/platform errors
  • Win/Bamboo: 1
• at the very last, rebase commits and squash them to more compact pieces
• use SlidingWindow add Sliding window implementation numenta/nupic.core-legacy#1277 instead of boost::CircularBuffer after it's merged
• consider inheriting from Anomaly and defining proper constructor, Anomaly likelihood numenta/nupic.core-legacy#1268 (comment)
• consider reset() for Likelihood, Port Anomaly Likelihood code from Python numenta/nupic.core-legacy#1267 (comment)
• honoring the timestamp, as in Anomaly likelihood numenta/nupic.core-legacy#1268 (comment)
• discuss removing aggregation functionality from Likelihood Anomaly likelihood numenta/nupic.core-legacy#1268 (comment)

Fixes numenta#1267
Moved from numenta#1268
! the issues/links in this description won't work, please follow the original PR for the references.

[breznak]

Any idea why the OSX Py tests are failing?
https://travis-ci.org/htm-community/nupic.cpp/builds/459707811#L5468

I didn't modify any Py/Swig code here

[dkeeney]

I don't know the background on this module but this seems like a nice algorithm.
I did not see anything I object to so I will approve it.

[dkeeney]

Hmmm, missed the fact that it did not pass the Python tests. This has something to do with Cells4. Did this indirectly change something in Cells4 somehow?

[breznak]

Tests pass ok locally, This is something in the CI with the bindings, happens with all other PRs as well. I’m suspecting some caching/artifacts

…

-- Marek Otahal :o)

[ctrl-z-9000-times]

Great work breznak!

My only worry is the unit tests aren't all there yet: //TODO port likelihood tests. If you don't want to do this task now then I would recommend documenting the task and leaving it for later. Just open an issue for this TODO note after this PR is merged.

[breznak]

My only worry is the unit tests aren't all there yet: //TODO port likelihood tests. If you don't want to do this task now then I would recommend documenting the task and leaving it for later

Thanks for review @ctrl-z-9000-times ! Yes, it's an old PR, I'm not using it currently, just want it merged so other people can. Opened #124 with TODO tests

[breznak]

Do you have an idea why this is failing? The PR doesn't touch any Region/python stuff. Passes OK an my linux locally, could you please test?

https://travis-ci.org/htm-community/nupic.cpp/builds/460796412#L5470

[breznak]

Assuming prev OK, this is finally merged too!