Hi Nick,

Thanks for the explanations. All this sounds great. With regards to
documentation for anomaly detection, there is a wiki page, the hackathon
video, and the anomaly likelihood code.  Is that sufficient for now? Maybe
if you write a section on anomaly detection we can make that our
documentation? You are a good writer!

Thanks,

--Subutai


On Thu, Sep 4, 2014 at 4:32 AM, Nicholas Mitri <[email protected]> wrote:

> Hi Subutai,
>
> Thanks again for the comments. I had a chance to go over them and will be
> editing the doc accordingly. Considering your last comment:
>
> "I don’t know if you were planning to do this or not, but it might be nice
> to have a short illustrative example and show how each algorithm component
> works within the context of that example. Also, were you planning on
> discussing anomaly detection and how anomaly scores are computed?"
>
>
> I do want to discuss both anomaly and swarming but the lack of
> documentation has been a hindrance. The vids covering those topics are
> great and informative but they don’t provide enough detail for me to be
> able to write a meaningful and sufficiently technical overview of their
> functions. I’ll go back to the wiki and vids and try to extract as much
> info to include them in the overview.
>
> As for the first part of the comment, I’ve assigned that discussion to a
> different chapter in the thesis (outline below)
>
>
> 5 Operation
>
>
>    - SP for clustering
>       - SP for feature extraction (Might be deprecated)
>       - TP for sequence learning
>       - TP for high level temporal stability (Might be deprecated)
>       - SP+TP
>       - Anomaly detection
>       - Classification
>
> I’m attaching the section covering the first point which I’ve shared with
> the ML some time ago. It has been updated since but includes some
> redundancies with the Overview chapter that will be removed later.
>
> I’ll be devising experiments for each section and carrying them out
> hopefully soon.
>
> Finally, the last chapter of the thesis is going to be about using HTM for
> Apnea prediction in infants. I’ve procured data from a local hospital
> collected from CPAP devices monitoring the breathing of multiple infants
> that suffer from apnea. I’m hoping that the data can be leveraged to
> predict anomalies in the breathing patterns.
>
>
> best,
> Nick
>
>
>
> On Sep 3, 2014, at 7:13 PM, Subutai Ahmad <[email protected]> wrote:
>
> Hi Nick,
>
> Overall this is a great succinct description of the algorithm. Nice job!
>  I really like the diagrams.
>
> I sent you some specific comments separately.  Hope that is useful!
>  Please send the final thesis if you are able to do that - I'm sure it
> would be helpful to lots of people on the list.
>
> Cheers,
>
> --Subutai
>
>
> On Tue, Sep 2, 2014 at 5:44 AM, Nicholas Mitri <[email protected]>
> wrote:
>
>> Hey all,
>>
>> I tried compiling most of the information I could gather from the white
>> paper, documentation, and vids into a succinct chapter for my thesis.
>> I’m attaching it here. It’s a very rough draft and I wouldn’t be
>> surprised if it were rife with inaccuracies.
>>
>> I’d highly appreciate any comments or feedback to refine it and iron out
>> any mistakes.
>>
>> thanks,
>> Nick
>>
>>
>
>
>

Reply via email to