New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
earthgecko_skyline_detector #330
Conversation
IssueID #2802: earthgecko_skyline_detector IssueID #2800: earthgecko_skyline - Added earthgecko_skyline Added: nab/detectors/earthgecko_skyline/ results/earthgeckoSkyline/ Modified: README.md config/thresholds.json results/final_results.json run.py
@subutai If this is simply a better implementation of the skyline detector, is there any reason not to just use the version that performs better? I would prefer if we only had one version on the leaderboard. What do you think @earthgecko? |
I think the main reason to keep the old implementation is that there have been published results with that version, so from a paper reproducibility standpoint it may be worth keeping. @earthgecko makes some really good points as to why that earlier implementation was insufficient, so I'm all for merging this version in (once any other review comments have been addressed). |
Hi @rhyolight and @subutai there may be another option. Seeing as the paper However I would suggest that if it is kept in, it is disabled by default and has This may be especially true if more time series are added to the corpus as the And on that note, I think that NAB may benefit from more real world time series |
That's an excellent idea! 👍
That would be great.
Wow, that would be a major improvement to NAB and warrant a version change. I have some nice data from the NAB competition as well (blood pressure data) but it was not enough to create a new version. However, if you can add some more diverse datasets, we might have enough to do a 1.1, 1.5 or even 2.0. |
There are different opinions about this ;) |
@smirmik I think that at some point, we will all come to realise, that we are all actually on the same side |
Wait until I add numenta and CAD OSE algorithms to Skyline and then we will see ;) We are all anomaly hunters in an information dimension. And to be honest, I am not sure that you lot and Jeff should be allowed to play with digital brains however you see fit :) However it must be said that it is amazing what you have achieved to date with such a port, such an abstract endeavour! I was glad to see that some of nupic scores have cyclonic complexity score of F, I would honesty expect nothing less. Sincere kudos to the team, as always, a remarkable achievement. @smirmik not to down play CAD OSE in anyway :) It just that numenta must be acknowledged for having done something really, really unique and innovative on a much grander canvas then simple anomaly detection. This took me much longer to post than expected, due to a github anomaly :) |
Would this be a change to this PR or are you talking about creating a new PR for it? |
Hi @rhyolight it will be a separate PR. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@earthgecko Looks nice and clean, but please run pylint with our pylintrc file and fix errors. Mostly spacing / indentation.
For example:
pylint --rcfile=$NUPIC/pylintrc nab/detectors/earthgecko_skyline/algorithms.py
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I just realized that most all the pylint warnings are coming from the copied code from the skyline detector. I can't really expect you do update all that code, nor do I want you to. I'll take a manual look over it and ignore the minor stuff.
@rhyolight I do not mind fixing all that actually. If I have to look at it and work with it, it always bothers me when things are not linted and it makes inline liniting in an IDE inefficient too. I will take a look. |
Typo Co-Authored-By: earthgecko <gary.wilson@of-networks.co.uk>
Ok, but don't change the skyline function names that were copied over. I want them to be exactly the same for search/find later. IF you want to make a PR after this that lints the entire NAB codebase I would be cool with that. |
I ran locally and got the same results as this PR, aside from expected rounding errors. |
Aside from any outstanding code review issues.... Before I do anything at this point, I want to make it clear what we have agreed to. Here is the path forward I prefer, please review and comment before I take action:
If you guys agree to that, I approve this PR. 👍 |
@rhyolight sounds good to me. If you would like any further changes with regards to aligning with the |
@rhyolight Sounds good to me! I assume we will later disable the old Skyline as the default case. |
@rhyolight the funny thing is ... (NAB-py2715-cp27mu) $ pylint --rcfile=$NUPIC/pylintrc nab/detectors/earthgecko_skyline/algorithms.py
Using config file ~/sandbox/of/github/numenta/nupic/pylintrc
------------------------------------
Your code has been rated at 10.00/10
(NAB-py2715-cp27mu) $ pylint --rcfile=$NUPIC/pylintrc nab/detectors/earthgecko_skyline/earthgecko_skyline_detector.py
Using config file ~/sandbox/of/github/numenta/nupic/pylintrc
************* Module nab.detectors.earthgecko_skyline.earthgecko_skyline_detector
C0412, 84: 2 - Imports from package nab are not grouped (ungrouped-imports)
-----------------------------------
Your code has been rated at 9.94/10
(NAB-py2715-cp27mu) $ pylint --rcfile=$NUPIC/pylintrc nab/detectors/earthgecko_skyline/skyline_algorithms.py
Using config file ~/sandbox/of/github/numenta/nupic/pylintrc
------------------------------------
Your code has been rated at 10.00/10
(NAB-py2715-cp27mu) $ I got it pylintrc'd for you :) However I had to change skyline function names that were in |
Adding earthgecko_skyline detector code, results, etc.