I am new here. I have some trouble with HTM algorithm. Hope to get help from you.
My question starts from after temporal pooling. Once I got the layer learning several pattern of the streaming data, the layer is now correctly doing multiple prediction. How could I be informed which pattern the data is currently undergoing.
I’ve watch the “A whole new world” tutorial on Youtube. Once the environment is successfully classified, there is a stable representation on L3, regardless the current position of the motor and L4 activities. How to achieve that?
The application lead to this question is a voice recognition program. I can encode the acoustic signal into SDR and put into HTM learning model. I’d like the program to report the syllables it heard.
Thanks very much!
Temporal pooling is not entirely implemented in NuPIC. Only sequence memory. Our current sensorimotor work will make progress on temporal pooling, but at this point with NuPIC you cannot identify all sequences like you want.