Here is the code: https://github.com/numenta/nupic/blob/master/nupic/algorithms/anomaly.py#L33
"The raw anomaly score is the fraction of active columns not predicted." And if you don't know what that means, watch this: https://www.youtube.com/watch?v=z6r3ekreRzY Hope that helps, --------- Matt Taylor OS Community Flag-Bearer Numenta On Fri, May 8, 2015 at 5:59 PM, Tom Tan <tom....@me.com> wrote: > Hi Matt and other experts, > > I have two questions. First question is again to understand how anomaly > likelihood is generated. I have a time series data. After running swarm, I > ran new data (not used in swarm) to the model. The prediction tracked the > actual reasonably well, however, I got wild anomaly likelihood (see diagram > below). What I might have done wrong? > > > > > 2nd question: Any doc/example showing persist and retrieve running active > cell states? Our data stream is one data point per hour or 30 minutes. I > can't keep CLA running all the time. In stead, when the new data is ready, > a new process is folked and Nupic is invoked. Therefore I need to save the > “state” when the previous run finishes and restore it before processing next > data point. > > > Regards, > Tom >
