US 11,657,278 B2
Location processor for inferencing and learning based on sensorimotor input data
Jeffrey C. Hawkins, Atherton, CA (US); and Marcus Anthony Lewis, San Francisco, CA (US)
Assigned to Numenta, Inc., Redwood City, CA (US)
Filed by Numenta, Inc., Redwood City, CA (US)
Filed on Jun. 25, 2020, as Appl. No. 16/912,415.
Application 16/912,415 is a continuation of application No. 15/934,795, filed on Mar. 23, 2018, granted, now 10,733,436.
Claims priority of provisional application 62/569,379, filed on Oct. 6, 2017.
Claims priority of provisional application 62/476,409, filed on Mar. 24, 2017.
Prior Publication US 2020/0327322 A1, Oct. 15, 2020
This patent is subject to a terminal disclaimer.
Int. Cl. G06N 3/08 (2023.01); G06N 5/04 (2023.01); G06V 20/80 (2022.01); G06V 10/94 (2022.01); G06V 10/20 (2022.01); G06V 10/44 (2022.01); G06F 18/22 (2023.01); G06N 3/045 (2023.01)
CPC G06N 3/08 (2013.01) [G06F 18/22 (2023.01); G06N 3/045 (2023.01); G06N 5/04 (2013.01); G06V 10/255 (2022.01); G06V 10/451 (2022.01); G06V 10/454 (2022.01); G06V 10/95 (2022.01); G06V 20/80 (2022.01)] 19 Claims
OG exemplary drawing
 
1. A computer-implemented method of performing inference, comprising:
receiving, by a first input processor, first input data derived from a first feature of an object from a sensor;
generating, by the first input processor, a first input representation indicating a combination of the first feature and first potential locations on the first candidates of the object associated with the first feature;
determining, by a location processor, a first candidate-location representation from the first input representation, the first candidate-location representation indicating the first candidates and the first potential locations on the first candidates, and the first candidate-location representation being an activation state of a first subset of data elements in the location processor;
receiving, by a second input processor, second input data derived from a second feature of the same object from another sensor;
generating, by the second input processor, a second input representation indicating a combination of the second feature and second potential locations on second candidates of the object associated with the second feature;
determining, by the location processor, a second candidate-location representation from the second input representation, the second candidate-location representation indicating the second candidates and the second potential locations on the second candidates;
updating the first candidate-location representation to be consistent with the second candidate-location representation, the updated first candidate-location representation indicating a subset of the first candidates and the first potential locations on the subset of the first candidates; and
determining the object based at least on the updated first candidate-location representation.