My analysis on why Temporal Memory prediction doesn't work on sequential data

mrcslws · December 5, 2017, 7:46pm

The backtracking_tm_shim.py wraps the pure TM. It doesn’t use the Backtracking TM. It mimics the interface of the Backtracking TM. Here’s the line where this class wraps the pure TM:

github.com

numenta/nupic/blob/1aea72abde4457878a16288d6786ffb088f69164/src/nupic/algorithms/backtracking_tm_shim.py#L165


  state = numpy.zeros([self.numberOfColumns(), self.getCellsPerColumn()])
  return state






class TMShim(TMShimMixin, TemporalMemory):
pass






class TMCPPShim(TMShimMixin, TemporalMemoryCPP):
pass






class MonitoredTMShim(MonitoredTemporalMemory):
"""
TM => Monitored Temporal Memory shim class.


TODO: This class is not very DRY. This whole file needs to be replaced by a
pure TemporalMemory region

Topic		Replies	Views
Exploring the "Repeating Inputs" problem Tangential Theories temporal-memory	36	3206	February 2, 2020
How does the HTM model learn about sequences? Numenta Theory	4	692	May 30, 2017
TemporalMemory for prediction Engineering question	35	1653	September 24, 2019
Exploring htm.core and the TM parameters NuPIC Community Fork	11	812	January 23, 2023
How Temporal Memory make t+n predictions, if his operate only t & t-1 timeframes? Getting Started sequence-memory , question , temporal-pooling	8	653	August 24, 2020

My analysis on why Temporal Memory prediction doesn't work on sequential data

Related Topics