Mat and David, Thanks for your positive feedbacks.
Agreed. The 2nd idea will take some more time. Another app idea on top of the 1st idea is sar(sysstat) anomaly detection. This is our daily challenge, so is for the any other IT support staffs. But this is not a new idea because Grok for IT is a flagship sample app of HTM. Anyway, OK, I will prepare my submission based on the 1st idea. Btw, David, I am going to use HTM.java. Thanks for your efforts so far! Thanks, Takenori On Fri, Oct 16, 2015 at 2:30 AM, David Ray <[email protected]> wrote: > Takenori, those are great ideas! > > Sent from my iPhone > > On Oct 15, 2015, at 10:58 AM, Matthew Taylor <[email protected]> wrote: > > Hi Takenori, > > I like both your ideas! The first idea may not be an application, but it > seems like it would provide a platform for others creating application, and > it has a potential business use-case. The 2nd idea is very interesting, but > I'm no sure that you will be able to collect enough data. You'll need a lot > of people with smartphones sending data, an app that collects the data you > need, and probably weeks worth of data to have enough for HTM to start > recognizing patterns. > > But I suggest you submit your first idea, it has a lot of potential. Then > maybe while you are building it, you will come up with another app idea? > > Thanks, > > --------- > Matt Taylor > OS Community Flag-Bearer > Numenta > > On Thu, Oct 15, 2015 at 3:05 AM, Takenori Sato <[email protected]> wrote: > >> Hello, >> >> I would like to join the HTM Challenge, >> but am not sure at all if mine can meet the requirements. >> >> In the first place, let me explain about my idea. >> >> 1. Amazon S3 extension API for anomaly detection >> >> * PUT bucket anomaly detection >> (parameters) >> - target bucket to store results(anomaly scores/predictions) >> - input information >> - model information >> - output information >> >> We develop and sell S3 compatible object storage software and appliance. >> In Japan, major service providers use our product to offer cloud storage >> service to their end users. >> >> So, imagine an end user uploads a sensor data periodically to the bucket >> on a cloud. In the cloud, a job to produce anomaly scores/predictions get >> scheduled, executed with HTM. And the result becomes available on the >> target bucket. An end user keeps polling the latest anomaly scores on the >> target bucket to see if there's an anomaly, and takes an action as >> required. Of course, it is possible to get historical data as well. >> >> This is a generic APIs for anomaly detection. >> >> 2. Low/Super Low frequency wave detector >> >> Low/Super Low frequency wave is becoming more problematic to human >> health(especially to sleep). >> >> It is not easy to measure, nor impossible to hear. Having difficulty in >> sleep on one night could be because of low frequency wave coming from a >> nearby location. >> >> So I guess it is not a bad idea to measure and record low/super low >> frequency waves with a smart phone, and keep uploading to a cloud. >> >> On the other hand, more and more disasters(earthquakes, landslides, >> eruptions) happen today in Japan. According to some researchers, they are >> observed with low frequency waves. If such an app is used by many people, >> it would detect a disaster before it happens. >> >> In summary, such an low/super low frequency wave data(geospatial temporal >> data) is put on a bucket with anomaly detection enabled, being analyzed by >> HTM as a whole. Then anomaly scores to indicate (unknown)disasters are >> generated, and notified as needed. >> >> >> 1 fully utilizes HTM, but not an application. 2 is an application, but >> does not directly use HTM. >> >> Besides, perhaps it's impossible to get samples of a disaster for a demo >> in one month. >> >> Is this qualified for the challenge? If yes, which part? >> >> Thanks, >> Takenori >> > >
