The multiencoder can encode different fields of the same data record into the same SDR, but what is the output of SP?

[helena_Thielen]

@flash59

I want to find out the differences between HTM and other unsupervised learning algorithms? Could you give me some advices?

sorry I am not familiar with other models… I just used supervised NN and HTM so far

@flash59

but now, I am confused that the two fields will be encoded into a same SDR, which means the HTM will process them with no differences (the same learning process in the SP), but, the system can output the results of each field separately. I want to know what happened underground?

The HTM uses a NN for the classification/prediction. This is what we call the SIR-classifier.
As far as I know, there is no paper about all the details. But you can check these links to find out more about this classifier:

SDR Classifier

http://nupic.docs.numenta.org/0.7.0.dev0/api/algorithms/classifiers.html?highlight=sdr#module-nupic.algorithms.sdr_classifier

And there is also a paper about the prediction that you can check… but the part about the classifier itself is quite short… but the paper is really interesting anyway.
and finally I also asked about the classifiers in this forum.
Maybe you get an better idea about the different classifiers when reading this. It is quite short, but IMO helpful