You guys are right, the CLA is a good fit for this, especially if we have a temporal pooler running. Coming up with a solution using the CLA would be a high profile win for the community and would showcase NuPic to the rest of the world in a very media friendly way.
There are some data sets available: The EPILEPSIAE project data is available and contains "well-documented meta data, highly annotated raw data as well as several calculated features. In total the database will contain continuous long term recordings of 275 patients, i.e. 225 scalp recordings and 50 intracranial recordings. The data acquisition and database are approved by the local ethics committees." http://www.epilepsiae.eu/project_outputs/european_database_on_epilepsy Here is another data set: http://eeg.pl/epi There are others on the NuPic list here interested feeding NuPic EEG data, not specifically for seizure prediction though. I wonder if Grok would be interested in hosting a Kaggle competition with the data for all-comers to take a shot at predicting with a prize? I would expect the NuPic OSS community would form one or more teams if that were the case and we could then see if the CLA outperforms other algos on messy temporal data. -Doug On Sat, Feb 8, 2014 at 12:53 AM, Chetan Surpur <[email protected]> wrote: > Hi Anubhav, > > That's a great idea! The CLA would be a good fit for this kind of anomaly > detection problem. > > I would start by collecting labeled data to begin experimenting with the > CLA and determine its performance in this task. Does anyone on this list > have access to such data? > > - Chetan > On Feb 8, 2014 12:44 AM, "Anubhav Chaturvedi" <[email protected]> > wrote: > >> Hi >> >> I am new to NuPIC platform have found it particularly interesting because >> of its inspiration from human brain model. I would the communities advice >> on an idea. >> >> I have read that the eeg signals start showing signs of an epileptic >> seizure way before its physiological impact. Also such eeg patterns vary a >> lot from person to person. Since NuPIC is an online learning platform, >> can't we develop a code that to learn its user's eeg patterns and predict a >> seizure so that they can be warned. If such an approach has been considered >> ( on NuPIC and not machine learning in general ), what were the challenges >> faced? >> >> I have had trouble finding people well versed in the fields of >> neuroscience to discuss about it. I hope you all find this area interesting >> and guide me. >> >> *Regards,* >> *Anubhav Chaturvedi* >> >> *Birla Institute of Technology & Science, Pilani* >> KK Birla Goa Campus >> +91-9637399150 >> >> *" You can do anything if you stop doing everything "* >> >> _______________________________________________ >> nupic mailing list >> [email protected] >> http://lists.numenta.org/mailman/listinfo/nupic_lists.numenta.org >> >> > _______________________________________________ > nupic mailing list > [email protected] > http://lists.numenta.org/mailman/listinfo/nupic_lists.numenta.org > >
_______________________________________________ nupic mailing list [email protected] http://lists.numenta.org/mailman/listinfo/nupic_lists.numenta.org
